Compare commits

...

36 Commits

Author SHA1 Message Date
jif-oai
6e936f92a4 R3 2025-10-31 14:24:09 +01:00
jif-oai
423c40d9b8 R2 2025-10-31 14:07:53 +01:00
jif-oai
bed7f051a4 R1 2025-10-31 14:03:07 +01:00
jif-oai
e770917d9b Some clippy 2025-10-31 13:37:11 +01:00
jif-oai
cfe7fb825c Q1 2025-10-31 13:28:33 +01:00
jif-oai
69341a04f4 R3 2025-10-30 21:16:14 +00:00
jif-oai
3315336366 R2 2025-10-30 21:05:46 +00:00
jif-oai
7a9c344bd5 R1 2025-10-30 20:44:42 +00:00
jif-oai
482e6d7fad Restore all comments 2025-10-30 19:07:30 +00:00
jif-oai
a1eee44844 Merge remote-tracking branch 'origin/main' into jif/client-work
# Conflicts:
#	codex-rs/core/tests/suite/client.rs
2025-10-30 19:02:00 +00:00
Owen Lin
89c00611c2 [app-server] remove serde(skip_serializing_if = "Option::is_none") annotations (#5939)
We had this annotation everywhere in app-server APIs which made it so
that fields get serialized as `field?: T`, meaning if the field as
`None` we would omit the field in the payload. Removing this annotation
changes it so that we return `field: T | null` instead, which makes
codex app-server's API more aligned with the convention of public OpenAI
APIs like Responses.

Separately, remove the `#[ts(optional_fields = nullable)]` annotations
that were recently added which made all the TS types become `field?: T |
null` which is not great since clients need to handle undefined and
null.

I think generally it'll be best to have optional types be either:
- `field: T | null` (preferred, aligned with public OpenAI APIs)
- `field?: T` where we have to, such as types generated from the MCP
schema:
https://github.com/modelcontextprotocol/modelcontextprotocol/blob/main/schema/2025-06-18/schema.ts
(see changes to `mcp-types/`)

I updated @etraut-openai's unit test to check that all generated TS
types are one or the other, not both (so will error if we have a type
that has `field?: T | null`). I don't think there's currently a good use
case for that - but we can always revisit.
2025-10-30 18:18:53 +00:00
Anton Panasenko
9572cfc782 [codex] add developer instructions (#5897)
we are using developer instructions for code reviews, we need to pass
them in cli as well.
2025-10-30 11:18:31 -07:00
Dylan Hurd
4a55646a02 chore: testing on freeform apply_patch (#5952)
## Summary
Duplicates the tests in `apply_patch_cli.rs`, but tests the freeform
apply_patch tool as opposed to the function call path. The good news is
that all the tests pass with zero logical tests, with the exception of
the heredoc, which doesn't really make sense in the freeform tool
context anyway.

@jif-oai since you wrote the original tests in #5557, I'd love your
opinion on the right way to DRY these test cases between the two. Happy
to set up a more sophisticated harness, but didn't want to go down the
rabbit hole until we agreed on the right pattern

## Testing
- [x] These are tests
2025-10-30 10:40:48 -07:00
jif-oai
209af68611 nit: log rmcp_client (#5978) 2025-10-30 17:40:38 +00:00
jif-oai
16b5f62b31 V7 2025-10-30 16:48:22 +00:00
jif-oai
bae5341585 V6 2025-10-30 16:24:41 +00:00
jif-oai
43f7733ad7 V4 2025-10-30 15:48:24 +00:00
jif-oai
0bba25eff3 V3 2025-10-30 15:34:21 +00:00
jif-oai
f4f9695978 feat: compaction prompt configurable (#5959)
```
 codex -c compact_prompt="Summarize in bullet points"
 ```
2025-10-30 14:24:24 +00:00
Ahmed Ibrahim
5fcc380bd9 Pass initial history as an optional to codex delegate (#5950)
This will give us more freedom on controlling the delegation. i.e we can
fork our history and run `compact`.
2025-10-30 07:22:42 -07:00
jif-oai
bb8f963b08 V2 2025-10-30 13:27:59 +00:00
jif-oai
7c48a0b717 V1 2025-10-30 13:25:17 +00:00
jif-oai
aa76003e28 chore: unify config crates (#5958) 2025-10-30 10:28:32 +00:00
Ahmed Ibrahim
fac548e430 Send delegate header (#5942)
Send delegate type header
2025-10-30 09:49:40 +00:00
Ahmed Ibrahim
9bd3453592 Add debug-only slash command for rollout path (#5936)
## Summary
- add a debug-only `/rollout` slash command that prints the rollout file
path or reports when none is known
- surface the new command in the slash command metadata and cover it
with unit tests

<img width="539" height="99" alt="image"
src="https://github.com/user-attachments/assets/688e1334-8a06-4576-abb8-ada33b458661"
/>
2025-10-30 03:51:00 +00:00
zhao-oai
b34efde2f3 asdf (#5940)
.
2025-10-30 01:10:41 +00:00
Ahmed Ibrahim
7aa46ab5fc ignore agent message deltas for the review mode (#5937)
The deltas produce the whole json output. ignore them.
2025-10-30 00:47:55 +00:00
pakrym-oai
bf35105af6 Re-enable SDK image forwarding test (#5934)
## Summary
- re-enable the TypeScript SDK test that verifies local images are
forwarded to `codex exec`

## Testing
- `pnpm test` *(fails: unable to download pnpm 10.8.1 because external
network access is blocked in the sandbox)*

------
https://chatgpt.com/codex/tasks/task_i_690289cb861083209fd006867e2adfb1
2025-10-29 23:18:26 +00:00
pakrym-oai
3429e82e45 Add item streaming events (#5546)
Adds AgentMessageContentDelta, ReasoningContentDelta,
ReasoningRawContentDelta item streaming events while maintaining
compatibility for old events.

---------

Co-authored-by: Owen Lin <owen@openai.com>
2025-10-29 22:33:57 +00:00
pakrym-oai
815ae4164a [exec] Add MCP tool arguments and results (#5899)
Extends mcp_tool_call item to include arguments and results.
2025-10-29 14:23:57 -07:00
Ahmed Ibrahim
13e1d0362d Delegate review to codex instance (#5572)
In this PR, I am exploring migrating task kind to an invocation of
Codex. The main reason would be getting rid off multiple
`ConversationHistory` state and streamlining our context/history
management.

This approach depends on opening a channel between the sub-codex and
codex. This channel is responsible for forwarding `interactive`
(`approvals`) and `non-interactive` events. The `task` is responsible
for handling those events.

This opens the door for implementing `codex as a tool`, replacing
`compact` and `review`, and potentially subagents.

One consideration is this code is very similar to `app-server` specially
in the approval part. If in the future we wanted an interactive
`sub-codex` we should consider using `codex-mcp`
2025-10-29 21:04:25 +00:00
jif-oai
db31f6966d chore: config editor (#5878)
The goal is to have a single place where we actually write files

In a follow-up PR, will move everything config related in a dedicated
module and move the helpers in a dedicated file
2025-10-29 20:52:46 +00:00
jif-oai
2b20cd66af fix: icu_decimal version (#5919) 2025-10-29 20:46:45 +00:00
Rasmus Rygaard
39e09c289d Add a wrapper around raw response items (#5923)
We currently have nested enums when sending raw response items in the
app-server protocol. This makes downstream schemas confusing because we
need to embed `type`-discriminated enums within each other.

This PR adds a small wrapper around the response item so we can keep the
schemas separate
2025-10-29 20:32:40 +00:00
Eric Traut
069a38a06c Add missing "nullable" macro to protocol structs that contain optional fields (#5901)
This PR addresses a current hole in the TypeScript code generation for
the API server protocol. Fields that are marked as "Optional<>" in the
Rust code are serialized such that the value is omitted when it is
deserialized — appearing as `undefined`, but the TS type indicates
(incorrectly) that it is always defined but possibly `null`. This can
lead to subtle errors that the TypeScript compiler doesn't catch. The
fix is to include the `#[ts(optional_fields = nullable)]` macro for all
protocol structs that contain one or more `Optional<>` fields.

This PR also includes a new test that validates that all TS protocol
code containing "| null" in its type is marked optional ("?") to catch
cases where `#[ts(optional_fields = nullable)]` is omitted.
2025-10-29 12:09:47 -07:00
jif-oai
3183935bd7 feat: add output even in sandbox denied (#5908) 2025-10-29 18:21:18 +00:00
139 changed files with 8491 additions and 4960 deletions

160
codex-rs/Cargo.lock generated
View File

@@ -592,9 +592,9 @@ checksum = "bef38d45163c2f1dde094a7dfd33ccf595c92905c8f8f4fdc18d06fb1037718a"
[[package]]
name = "bitflags"
version = "2.9.1"
version = "2.10.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1b8e56985ec62d17e9c1001dc89c88ecd7dc08e47eba5ec7c29c7b5eeecde967"
checksum = "812e12b5285cc515a9c72a5c1d3b6d46a19dac5acfef5265968c166106e31dd3"
[[package]]
name = "block-buffer"
@@ -829,6 +829,29 @@ dependencies = [
"tracing",
]
[[package]]
name = "codex-api-client"
version = "0.0.0"
dependencies = [
"async-trait",
"bytes",
"codex-app-server-protocol",
"codex-otel",
"codex-protocol",
"eventsource-stream",
"futures",
"maplit",
"regex-lite",
"reqwest",
"serde",
"serde_json",
"thiserror 2.0.16",
"tokio",
"tokio-util",
"toml",
"tracing",
]
[[package]]
name = "codex-app-server"
version = "0.0.0"
@@ -1039,6 +1062,7 @@ name = "codex-common"
version = "0.0.0"
dependencies = [
"clap",
"codex-api-client",
"codex-app-server-protocol",
"codex-core",
"codex-protocol",
@@ -1059,6 +1083,7 @@ dependencies = [
"base64",
"bytes",
"chrono",
"codex-api-client",
"codex-app-server-protocol",
"codex-apply-patch",
"codex-async-utils",
@@ -1131,6 +1156,7 @@ dependencies = [
"anyhow",
"assert_cmd",
"clap",
"codex-api-client",
"codex-arg0",
"codex-common",
"codex-core",
@@ -1206,7 +1232,6 @@ name = "codex-git"
version = "0.0.0"
dependencies = [
"assert_matches",
"once_cell",
"pretty_assertions",
"regex",
"schemars 0.8.22",
@@ -1296,6 +1321,7 @@ dependencies = [
"assert_matches",
"async-stream",
"bytes",
"codex-api-client",
"codex-core",
"futures",
"reqwest",
@@ -1343,6 +1369,7 @@ dependencies = [
"codex-utils-image",
"icu_decimal",
"icu_locale_core",
"icu_provider",
"mcp-types",
"mime_guess",
"schemars 0.8.22",
@@ -1436,6 +1463,7 @@ dependencies = [
"chrono",
"clap",
"codex-ansi-escape",
"codex-api-client",
"codex-app-server-protocol",
"codex-arg0",
"codex-common",
@@ -1669,6 +1697,7 @@ version = "0.0.0"
dependencies = [
"anyhow",
"assert_cmd",
"codex-api-client",
"codex-core",
"codex-protocol",
"notify",
@@ -1738,7 +1767,7 @@ version = "0.28.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "829d955a0bb380ef178a640b91779e3987da38c9aea133b20614cfed8cdea9c6"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"crossterm_winapi",
"futures-core",
"mio",
@@ -2082,7 +2111,7 @@ version = "0.3.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "89a09f22a6c6069a18470eb92d2298acf25463f14256d24778e1230d789a2aec"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"objc2",
]
@@ -2984,9 +3013,9 @@ dependencies = [
[[package]]
name = "icu_collections"
version = "2.0.0"
version = "2.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "200072f5d0e3614556f94a9930d5dc3e0662a652823904c3a75dc3b0af7fee47"
checksum = "4c6b649701667bbe825c3b7e6388cb521c23d88644678e83c0c4d0a621a34b43"
dependencies = [
"displaydoc",
"potential_utf",
@@ -2997,34 +3026,31 @@ dependencies = [
[[package]]
name = "icu_decimal"
version = "2.0.0"
version = "2.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "fec61c43fdc4e368a9f450272833123a8ef0d7083a44597660ce94d791b8a2e2"
checksum = "a38c52231bc348f9b982c1868a2af3195199623007ba2c7650f432038f5b3e8e"
dependencies = [
"displaydoc",
"fixed_decimal",
"icu_decimal_data",
"icu_locale",
"icu_locale_core",
"icu_provider",
"tinystr",
"writeable",
"zerovec",
]
[[package]]
name = "icu_decimal_data"
version = "2.0.0"
version = "2.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b70963bc35f9bdf1bc66a5c1f458f4991c1dc71760e00fa06016b2c76b2738d5"
checksum = "2905b4044eab2dd848fe84199f9195567b63ab3a93094711501363f63546fef7"
[[package]]
name = "icu_locale"
version = "2.0.0"
version = "2.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "6ae5921528335e91da1b6c695dbf1ec37df5ac13faa3f91e5640be93aa2fbefd"
checksum = "532b11722e350ab6bf916ba6eb0efe3ee54b932666afec989465f9243fe6dd60"
dependencies = [
"displaydoc",
"icu_collections",
"icu_locale_core",
"icu_locale_data",
@@ -3036,12 +3062,13 @@ dependencies = [
[[package]]
name = "icu_locale_core"
version = "2.0.0"
version = "2.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0cde2700ccaed3872079a65fb1a78f6c0a36c91570f28755dda67bc8f7d9f00a"
checksum = "edba7861004dd3714265b4db54a3c390e880ab658fec5f7db895fae2046b5bb6"
dependencies = [
"displaydoc",
"litemap",
"serde",
"tinystr",
"writeable",
"zerovec",
@@ -3049,17 +3076,16 @@ dependencies = [
[[package]]
name = "icu_locale_data"
version = "2.0.0"
version = "2.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4fdef0c124749d06a743c69e938350816554eb63ac979166590e2b4ee4252765"
checksum = "f03e2fcaefecdf05619f3d6f91740e79ab969b4dd54f77cbf546b1d0d28e3147"
[[package]]
name = "icu_normalizer"
version = "2.0.0"
version = "2.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "436880e8e18df4d7bbc06d58432329d6458cc84531f7ac5f024e93deadb37979"
checksum = "5f6c8828b67bf8908d82127b2054ea1b4427ff0230ee9141c54251934ab1b599"
dependencies = [
"displaydoc",
"icu_collections",
"icu_normalizer_data",
"icu_properties",
@@ -3070,42 +3096,40 @@ dependencies = [
[[package]]
name = "icu_normalizer_data"
version = "2.0.0"
version = "2.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "00210d6893afc98edb752b664b8890f0ef174c8adbb8d0be9710fa66fbbf72d3"
checksum = "7aedcccd01fc5fe81e6b489c15b247b8b0690feb23304303a9e560f37efc560a"
[[package]]
name = "icu_properties"
version = "2.0.1"
version = "2.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "016c619c1eeb94efb86809b015c58f479963de65bdb6253345c1a1276f22e32b"
checksum = "e93fcd3157766c0c8da2f8cff6ce651a31f0810eaa1c51ec363ef790bbb5fb99"
dependencies = [
"displaydoc",
"icu_collections",
"icu_locale_core",
"icu_properties_data",
"icu_provider",
"potential_utf",
"zerotrie",
"zerovec",
]
[[package]]
name = "icu_properties_data"
version = "2.0.1"
version = "2.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "298459143998310acd25ffe6810ed544932242d3f07083eee1084d83a71bd632"
checksum = "02845b3647bb045f1100ecd6480ff52f34c35f82d9880e029d329c21d1054899"
[[package]]
name = "icu_provider"
version = "2.0.0"
version = "2.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "03c80da27b5f4187909049ee2d72f276f0d9f99a42c306bd0131ecfe04d8e5af"
checksum = "85962cf0ce02e1e0a629cc34e7ca3e373ce20dda4c4d7294bbd0bf1fdb59e614"
dependencies = [
"displaydoc",
"icu_locale_core",
"serde",
"stable_deref_trait",
"tinystr",
"writeable",
"yoke",
"zerofrom",
@@ -3212,7 +3236,7 @@ version = "0.11.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f37dccff2791ab604f9babef0ba14fbe0be30bd368dc541e2b08d07c8aa908f3"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"inotify-sys",
"libc",
]
@@ -3275,7 +3299,7 @@ version = "0.7.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d93587f37623a1a17d94ef2bc9ada592f5465fe7732084ab7beefabe5c77c0c4"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"cfg-if",
"libc",
]
@@ -3521,7 +3545,7 @@ version = "0.1.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4488594b9328dee448adb906d8b126d9b7deb7cf5c22161ee591610bb1be83c0"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"libc",
]
@@ -3531,7 +3555,7 @@ version = "0.2.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "761e49ec5fd8a5a463f9b84e877c373d888935b71c6be78f3767fe2ae6bed18e"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"libc",
]
@@ -3799,7 +3823,7 @@ version = "0.28.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ab2156c4fce2f8df6c499cc1c763e4394b7482525bf2a9701c9d79d215f519e4"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"cfg-if",
"cfg_aliases 0.1.1",
"libc",
@@ -3811,7 +3835,7 @@ version = "0.29.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "71e2746dc3a24dd78b3cfcb7be93368c6de9963d30f43a6a73998a9cf4b17b46"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"cfg-if",
"cfg_aliases 0.2.1",
"libc",
@@ -3824,7 +3848,7 @@ version = "0.30.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "74523f3a35e05aba87a1d978330aef40f67b0304ac79c1c00b294c9830543db6"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"cfg-if",
"cfg_aliases 0.2.1",
"libc",
@@ -3852,7 +3876,7 @@ version = "8.2.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4d3d07927151ff8575b7087f245456e549fea62edf0ec4e565a5ee50c8402bc3"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"fsevent-sys",
"inotify",
"kqueue",
@@ -4022,7 +4046,7 @@ version = "0.3.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e6f29f568bec459b0ddff777cec4fe3fd8666d82d5a40ebd0ff7e66134f89bcc"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"objc2",
"objc2-core-graphics",
"objc2-foundation",
@@ -4034,7 +4058,7 @@ version = "0.3.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1c10c2894a6fed806ade6027bcd50662746363a9589d3ec9d9bef30a4e4bc166"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"dispatch2",
"objc2",
]
@@ -4045,7 +4069,7 @@ version = "0.3.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "989c6c68c13021b5c2d6b71456ebb0f9dc78d752e86a98da7c716f4f9470f5a4"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"dispatch2",
"objc2",
"objc2-core-foundation",
@@ -4064,7 +4088,7 @@ version = "0.3.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "900831247d2fe1a09a683278e5384cfb8c80c79fe6b166f9d14bfdde0ea1b03c"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"objc2",
"objc2-core-foundation",
]
@@ -4075,7 +4099,7 @@ version = "0.3.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7282e9ac92529fa3457ce90ebb15f4ecbc383e8338060960760fa2cf75420c3c"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"objc2",
"objc2-core-foundation",
]
@@ -4107,7 +4131,7 @@ version = "0.10.73"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8505734d46c8ab1e19a1dce3aef597ad87dcb4c37e7188231769bd6bd51cebf8"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"cfg-if",
"foreign-types",
"libc",
@@ -4439,7 +4463,7 @@ version = "0.18.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "97baced388464909d42d89643fe4361939af9b7ce7a31ee32a168f832a70f2a0"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"crc32fast",
"fdeflate",
"flate2",
@@ -4498,11 +4522,12 @@ dependencies = [
[[package]]
name = "potential_utf"
version = "0.1.2"
version = "0.1.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e5a7c30837279ca13e7c867e9e40053bc68740f988cb07f7ca6df43cc734b585"
checksum = "b73949432f5e2a09657003c25bca5e19a0e9c84f8058ca374f49e0ebe605af77"
dependencies = [
"serde",
"serde_core",
"writeable",
"zerovec",
]
@@ -4628,7 +4653,7 @@ version = "0.10.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "76979bea66e7875e7509c4ec5300112b316af87fa7a252ca91c448b32dfe3993"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"getopts",
"memchr",
"pulldown-cmark-escape",
@@ -4809,7 +4834,7 @@ name = "ratatui"
version = "0.29.0"
source = "git+https://github.com/nornagon/ratatui?branch=nornagon-v0.29.0-patch#9b2ad1298408c45918ee9f8241a6f95498cdbed2"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"cassowary",
"compact_str",
"crossterm",
@@ -4839,7 +4864,7 @@ version = "0.5.15"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7e8af0dde094006011e6a740d4879319439489813bd0bcdc7d821beaeeff48ec"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
]
[[package]]
@@ -5068,7 +5093,7 @@ version = "0.38.44"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "fdb5bc1ae2baa591800df16c9ca78619bf65c0488b41b96ccec5d11220d8c154"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"errno",
"libc",
"linux-raw-sys 0.4.15",
@@ -5081,7 +5106,7 @@ version = "1.0.8"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "11181fbabf243db407ef8df94a6ce0b2f9a733bd8be4ad02b4eda9602296cac8"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"errno",
"libc",
"linux-raw-sys 0.9.4",
@@ -5147,7 +5172,7 @@ version = "14.0.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7803e8936da37efd9b6d4478277f4b2b9bb5cdb37a113e8d63222e58da647e63"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"cfg-if",
"clipboard-win",
"fd-lock",
@@ -5346,7 +5371,7 @@ version = "2.11.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "897b2245f0b511c87893af39b033e5ca9cce68824c4d7e7630b5a1d339658d02"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"core-foundation 0.9.4",
"core-foundation-sys",
"libc",
@@ -5359,7 +5384,7 @@ version = "3.5.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b3297343eaf830f66ede390ea39da1d462b6b0c1b000f420d0a83f898bbbe6ef"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"core-foundation 0.10.1",
"core-foundation-sys",
"libc",
@@ -6058,7 +6083,7 @@ version = "0.6.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3c879d448e9d986b661742763247d3693ed13609438cf3d006f51f5368a5ba6b"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"core-foundation 0.9.4",
"system-configuration-sys",
]
@@ -6516,7 +6541,7 @@ version = "0.6.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "adc82fd73de2a9722ac5da747f12383d2bfdb93591ee6c58486e0097890f05f2"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
"bytes",
"futures-util",
"http",
@@ -7591,14 +7616,14 @@ version = "0.39.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "6f42320e61fe2cfd34354ecb597f86f413484a798ba44a8ca1165c58d42da6c1"
dependencies = [
"bitflags 2.9.1",
"bitflags 2.10.0",
]
[[package]]
name = "writeable"
version = "0.6.1"
version = "0.6.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ea2f10b9bb0928dfb1b42b65e1f9e36f7f54dbdf08457afefb38afcdec4fa2bb"
checksum = "9edde0db4769d2dc68579893f2306b26c6ecfbe0ef499b013d731b7b9247e0b9"
[[package]]
name = "x11rb"
@@ -7793,10 +7818,11 @@ dependencies = [
[[package]]
name = "zerovec"
version = "0.11.2"
version = "0.11.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4a05eb080e015ba39cc9e23bbe5e7fb04d5fb040350f99f34e338d5fdd294428"
checksum = "6c28719294829477f525be0186d13efa9a3c602f7ec202ca9e353d310fb9a002"
dependencies = [
"serde",
"yoke",
"zerofrom",
"zerovec-derive",

View File

@@ -38,7 +38,7 @@ members = [
"utils/pty",
"utils/readiness",
"utils/string",
"utils/tokenizer",
"utils/tokenizer", "api-client",
]
resolver = "2"
@@ -87,6 +87,7 @@ codex-utils-pty = { path = "utils/pty" }
codex-utils-readiness = { path = "utils/readiness" }
codex-utils-string = { path = "utils/string" }
codex-utils-tokenizer = { path = "utils/tokenizer" }
codex-api-client = { path = "api-client" }
core_test_support = { path = "core/tests/common" }
mcp-types = { path = "mcp-types" }
mcp_test_support = { path = "mcp-server/tests/common" }
@@ -122,8 +123,9 @@ escargot = "0.5"
eventsource-stream = "0.2.3"
futures = { version = "0.3", default-features = false }
http = "1.3.1"
icu_decimal = "2.0.0"
icu_locale_core = "2.0.0"
icu_decimal = "2.1"
icu_provider = { version = "2.1", features = ["sync"] }
icu_locale_core = "2.1"
ignore = "0.4.23"
image = { version = "^0.25.8", default-features = false }
indexmap = "2.6.0"
@@ -253,7 +255,7 @@ unwrap_used = "deny"
# cargo-shear cannot see the platform-specific openssl-sys usage, so we
# silence the false positive here instead of deleting a real dependency.
[workspace.metadata.cargo-shear]
ignored = ["openssl-sys", "codex-utils-readiness", "codex-utils-tokenizer"]
ignored = ["icu_provider", "openssl-sys", "codex-utils-readiness", "codex-utils-tokenizer"]
[profile.release]
lto = "fat"

View File

@@ -0,0 +1,28 @@
[package]
name = "codex-api-client"
version.workspace = true
edition.workspace = true
[dependencies]
async-trait = { workspace = true }
bytes = { workspace = true }
codex-app-server-protocol = { workspace = true }
codex-otel = { workspace = true }
codex-protocol = { path = "../protocol" }
eventsource-stream = { workspace = true }
futures = { workspace = true, default-features = false, features = ["std"] }
regex-lite = { workspace = true }
reqwest = { workspace = true, features = ["json", "stream"] }
serde = { workspace = true, features = ["derive"] }
serde_json = { workspace = true }
thiserror = { workspace = true }
tokio = { workspace = true, features = ["sync", "time", "rt", "rt-multi-thread", "macros", "io-util"] }
tokio-util = { workspace = true }
tracing = { workspace = true }
[dev-dependencies]
maplit = "1.0.2"
toml = { workspace = true }
[lints]
workspace = true

View File

@@ -0,0 +1,16 @@
use async_trait::async_trait;
use crate::error::Error;
use crate::prompt::Prompt;
use crate::stream::ResponseStream;
#[async_trait]
pub trait ApiClient: Send + Sync {
type Config: Send + Sync;
async fn new(config: Self::Config) -> Result<Self, Error>
where
Self: Sized;
async fn stream(&self, prompt: Prompt) -> Result<ResponseStream, Error>;
}

View File

@@ -0,0 +1,15 @@
use async_trait::async_trait;
use codex_app_server_protocol::AuthMode;
#[derive(Debug, Clone)]
pub struct AuthContext {
pub mode: AuthMode,
pub bearer_token: Option<String>,
pub account_id: Option<String>,
}
#[async_trait]
pub trait AuthProvider: Send + Sync {
async fn auth_context(&self) -> Option<AuthContext>;
async fn refresh_token(&self) -> Result<Option<String>, String>;
}

View File

@@ -0,0 +1,866 @@
use std::collections::VecDeque;
use std::pin::Pin;
use std::task::Context;
use std::task::Poll;
use std::time::Duration;
use async_trait::async_trait;
use bytes::Bytes;
use codex_otel::otel_event_manager::OtelEventManager;
use codex_protocol::models::ContentItem;
use codex_protocol::models::FunctionCallOutputContentItem;
use codex_protocol::models::ReasoningItemContent;
use codex_protocol::models::ResponseItem;
use codex_protocol::protocol::SessionSource;
use codex_protocol::protocol::SubAgentSource;
use eventsource_stream::Eventsource;
use futures::Stream;
use futures::StreamExt;
use futures::TryStreamExt;
use serde_json::Value;
use serde_json::json;
use tokio::sync::mpsc;
use tokio::time::timeout;
use tracing::debug;
use tracing::trace;
use crate::api::ApiClient;
use crate::error::Error;
use crate::model_provider::ModelProviderInfo;
use crate::prompt::Prompt;
use crate::stream::ResponseEvent;
use crate::stream::ResponseStream;
pub type Result<T> = std::result::Result<T, Error>;
#[derive(Clone, Copy, Debug)]
pub enum ChatAggregationMode {
AggregatedOnly,
Streaming,
}
#[derive(Clone)]
pub struct ChatCompletionsApiClientConfig {
pub http_client: reqwest::Client,
pub provider: ModelProviderInfo,
pub model: String,
pub otel_event_manager: OtelEventManager,
pub session_source: SessionSource,
pub aggregation_mode: ChatAggregationMode,
}
#[derive(Clone)]
pub struct ChatCompletionsApiClient {
config: ChatCompletionsApiClientConfig,
}
#[async_trait]
impl ApiClient for ChatCompletionsApiClient {
type Config = ChatCompletionsApiClientConfig;
async fn new(config: Self::Config) -> Result<Self> {
Ok(Self { config })
}
async fn stream(&self, prompt: Prompt) -> Result<ResponseStream> {
Self::validate_prompt(&prompt)?;
let payload = self.build_payload(&prompt)?;
let (tx_event, rx_event) = mpsc::channel::<Result<ResponseEvent>>(1600);
let mut attempt = 0u64;
let max_retries = self.config.provider.request_max_retries();
loop {
attempt += 1;
let mut req_builder = self
.config
.provider
.create_request_builder(&self.config.http_client, &None)
.await?;
if let SessionSource::SubAgent(sub) = &self.config.session_source {
let subagent = if let SubAgentSource::Other(label) = sub {
label.clone()
} else {
serde_json::to_value(sub)
.ok()
.and_then(|v| v.as_str().map(std::string::ToString::to_string))
.unwrap_or_else(|| "other".to_string())
};
req_builder = req_builder.header("x-openai-subagent", subagent);
}
let res = self
.config
.otel_event_manager
.log_request(attempt, || {
req_builder
.header(reqwest::header::ACCEPT, "text/event-stream")
.json(&payload)
.send()
})
.await;
match res {
Ok(resp) if resp.status().is_success() => {
let stream = resp
.bytes_stream()
.map_err(|err| Error::ResponseStreamFailed {
source: err,
request_id: None,
});
let idle_timeout = self.config.provider.stream_idle_timeout();
let otel = self.config.otel_event_manager.clone();
let mode = self.config.aggregation_mode;
tokio::spawn(process_chat_sse(
stream,
tx_event.clone(),
idle_timeout,
otel,
mode,
));
return Ok(ResponseStream { rx_event });
}
Ok(resp) => {
if attempt >= max_retries {
let status = resp.status();
let body = resp
.text()
.await
.unwrap_or_else(|_| "<failed to read response>".to_string());
return Err(Error::UnexpectedStatus { status, body });
}
let retry_after = resp
.headers()
.get(reqwest::header::RETRY_AFTER)
.and_then(|v| v.to_str().ok())
.and_then(|s| s.parse::<u64>().ok())
.map(Duration::from_secs);
tokio::time::sleep(retry_after.unwrap_or_else(|| backoff(attempt))).await;
}
Err(error) => {
if attempt >= max_retries {
return Err(Error::Http(error));
}
tokio::time::sleep(backoff(attempt)).await;
}
}
}
}
}
impl ChatCompletionsApiClient {
fn validate_prompt(prompt: &Prompt) -> Result<()> {
if prompt.output_schema.is_some() {
return Err(Error::UnsupportedOperation(
"output_schema is not supported for Chat Completions API".to_string(),
));
}
Ok(())
}
fn build_payload(&self, prompt: &Prompt) -> Result<serde_json::Value> {
let mut messages = Vec::<serde_json::Value>::new();
messages.push(json!({ "role": "system", "content": prompt.instructions }));
let mut reasoning_by_anchor_index: std::collections::HashMap<usize, String> =
std::collections::HashMap::new();
let mut last_emitted_role: Option<&str> = None;
for item in &prompt.input {
match item {
ResponseItem::Message { role, .. } => last_emitted_role = Some(role.as_str()),
ResponseItem::FunctionCall { .. } | ResponseItem::LocalShellCall { .. } => {
last_emitted_role = Some("assistant");
}
ResponseItem::FunctionCallOutput { .. } => last_emitted_role = Some("tool"),
ResponseItem::Reasoning { .. }
| ResponseItem::Other
| ResponseItem::CustomToolCall { .. }
| ResponseItem::CustomToolCallOutput { .. }
| ResponseItem::WebSearchCall { .. }
| ResponseItem::GhostSnapshot { .. } => {}
}
}
let mut last_user_index: Option<usize> = None;
for (idx, item) in prompt.input.iter().enumerate() {
if let ResponseItem::Message { role, .. } = item
&& role == "user"
{
last_user_index = Some(idx);
}
}
if !matches!(last_emitted_role, Some("user")) {
for (idx, item) in prompt.input.iter().enumerate() {
if let Some(u_idx) = last_user_index
&& idx <= u_idx
{
continue;
}
if let ResponseItem::Reasoning {
content: Some(items),
..
} = item
{
let mut text = String::new();
for entry in items {
match entry {
ReasoningItemContent::ReasoningText { text: segment }
| ReasoningItemContent::Text { text: segment } => {
text.push_str(segment);
}
}
}
if text.trim().is_empty() {
continue;
}
let mut attached = false;
if idx > 0
&& let ResponseItem::Message { role, .. } = &prompt.input[idx - 1]
&& role == "assistant"
{
reasoning_by_anchor_index
.entry(idx - 1)
.and_modify(|v| v.push_str(&text))
.or_insert(text.clone());
attached = true;
}
if !attached && idx + 1 < prompt.input.len() {
match &prompt.input[idx + 1] {
ResponseItem::FunctionCall { .. }
| ResponseItem::LocalShellCall { .. } => {
reasoning_by_anchor_index
.entry(idx + 1)
.and_modify(|v| v.push_str(&text))
.or_insert(text.clone());
}
ResponseItem::Message { role, .. } if role == "assistant" => {
reasoning_by_anchor_index
.entry(idx + 1)
.and_modify(|v| v.push_str(&text))
.or_insert(text.clone());
}
_ => {}
}
}
}
}
}
let mut last_assistant_text: Option<String> = None;
for (idx, item) in prompt.input.iter().enumerate() {
match item {
ResponseItem::Message { role, content, .. } => {
let mut text = String::new();
let mut items: Vec<serde_json::Value> = Vec::new();
let mut saw_image = false;
for c in content {
match c {
ContentItem::InputText { text: t }
| ContentItem::OutputText { text: t } => {
text.push_str(t);
items.push(json!({"type":"text","text": t}));
}
ContentItem::InputImage { image_url } => {
saw_image = true;
items.push(
json!({"type":"image_url","image_url": {"url": image_url}}),
);
}
}
}
if role == "assistant" {
if let Some(prev) = &last_assistant_text
&& prev == &text
{
continue;
}
last_assistant_text = Some(text.clone());
}
let content_value = if role == "assistant" {
json!(text)
} else if saw_image {
json!(items)
} else {
json!(text)
};
let mut message = json!({
"role": role,
"content": content_value,
});
if let Some(reasoning) = reasoning_by_anchor_index.get(&idx)
&& let Some(obj) = message.as_object_mut()
{
obj.insert("reasoning".to_string(), json!({"text": reasoning}));
}
messages.push(message);
}
ResponseItem::FunctionCall {
name,
arguments,
call_id,
..
} => {
messages.push(json!({
"role": "assistant",
"tool_calls": [{
"id": call_id,
"type": "function",
"function": {
"name": name,
"arguments": arguments,
},
}],
}));
}
ResponseItem::FunctionCallOutput { call_id, output } => {
let content_value = if let Some(items) = &output.content_items {
let mapped: Vec<serde_json::Value> = items
.iter()
.map(|item| match item {
FunctionCallOutputContentItem::InputText { text } => {
json!({"type":"text","text": text})
}
FunctionCallOutputContentItem::InputImage { image_url } => {
json!({"type":"image_url","image_url": {"url": image_url}})
}
})
.collect();
json!(mapped)
} else {
json!(output.content)
};
messages.push(json!({
"role": "tool",
"tool_call_id": call_id,
"content": content_value,
}));
}
ResponseItem::LocalShellCall {
id,
call_id,
action,
..
} => {
let tool_id = call_id
.clone()
.filter(|value| !value.is_empty())
.or_else(|| id.clone())
.unwrap_or_default();
messages.push(json!({
"role": "assistant",
"tool_calls": [{
"id": tool_id,
"type": "function",
"function": {
"name": "shell",
"arguments": serde_json::to_string(action).unwrap_or_default(),
},
}],
}));
}
ResponseItem::CustomToolCall {
call_id,
name,
input,
..
} => {
messages.push(json!({
"role": "assistant",
"tool_calls": [{
"id": call_id.clone(),
"type": "function",
"function": {
"name": name,
"arguments": input,
},
}],
}));
}
ResponseItem::CustomToolCallOutput { call_id, output } => {
messages.push(json!({
"role": "tool",
"tool_call_id": call_id,
"content": output,
}));
}
ResponseItem::WebSearchCall { .. }
| ResponseItem::Reasoning { .. }
| ResponseItem::Other
| ResponseItem::GhostSnapshot { .. } => {}
}
}
let tools_json = create_tools_json_for_chat_completions_api(&prompt.tools)?;
let payload = json!({
"model": self.config.model,
"messages": messages,
"stream": true,
"tools": tools_json,
});
trace!("chat completions payload: {}", payload);
Ok(payload)
}
}
async fn append_assistant_text(
tx_event: &mpsc::Sender<Result<ResponseEvent>>,
assistant_item: &mut Option<ResponseItem>,
text: String,
) {
if assistant_item.is_none() {
let item = ResponseItem::Message {
id: None,
role: "assistant".to_string(),
content: vec![],
};
*assistant_item = Some(item.clone());
let _ = tx_event
.send(Ok(ResponseEvent::OutputItemAdded(item)))
.await;
}
if let Some(ResponseItem::Message { content, .. }) = assistant_item {
content.push(ContentItem::OutputText { text: text.clone() });
let _ = tx_event
.send(Ok(ResponseEvent::OutputTextDelta(text.clone())))
.await;
}
}
async fn append_reasoning_text(
tx_event: &mpsc::Sender<Result<ResponseEvent>>,
reasoning_item: &mut Option<ResponseItem>,
text: String,
) {
if reasoning_item.is_none() {
let item = ResponseItem::Reasoning {
id: String::new(),
summary: Vec::new(),
content: Some(vec![]),
encrypted_content: None,
};
*reasoning_item = Some(item.clone());
let _ = tx_event
.send(Ok(ResponseEvent::OutputItemAdded(item)))
.await;
}
if let Some(ResponseItem::Reasoning {
content: Some(content),
..
}) = reasoning_item
{
content.push(ReasoningItemContent::ReasoningText { text: text.clone() });
let _ = tx_event
.send(Ok(ResponseEvent::ReasoningContentDelta(text.clone())))
.await;
}
}
async fn process_chat_sse<S>(
stream: S,
tx_event: mpsc::Sender<Result<ResponseEvent>>,
idle_timeout: Duration,
otel_event_manager: OtelEventManager,
_aggregation_mode: ChatAggregationMode,
) where
S: Stream<Item = Result<Bytes>> + Unpin,
{
let mut stream = stream.eventsource();
#[derive(Default)]
struct FunctionCallState {
name: Option<String>,
arguments: String,
call_id: Option<String>,
}
let mut function_call_state = FunctionCallState::default();
let mut assistant_item: Option<ResponseItem> = None;
let mut reasoning_item: Option<ResponseItem> = None;
loop {
let response = timeout(idle_timeout, stream.next()).await;
otel_event_manager.log_sse_event(&response, idle_timeout);
let sse = match response {
Ok(Some(Ok(sse))) => sse,
Ok(Some(Err(e))) => {
debug!("SSE Error: {e:#}");
let event = Error::Stream(e.to_string(), None);
let _ = tx_event.send(Err(event)).await;
return;
}
Ok(None) => {
if let Some(item) = assistant_item.take() {
let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
}
if let Some(item) = reasoning_item.take() {
let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
}
let _ = tx_event
.send(Ok(ResponseEvent::Completed {
response_id: String::new(),
token_usage: None,
}))
.await;
return;
}
Err(_) => {
let _ = tx_event
.send(Err(Error::Stream(
"idle timeout waiting for SSE".into(),
None,
)))
.await;
return;
}
};
trace!("chat_completions received SSE chunk: {}", sse.data);
if sse.data.trim() == "[DONE]" {
if let Some(item) = assistant_item.take() {
let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
}
if let Some(item) = reasoning_item.take() {
let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
}
let _ = tx_event
.send(Ok(ResponseEvent::Completed {
response_id: String::new(),
token_usage: None,
}))
.await;
return;
}
let chunk: serde_json::Value = match serde_json::from_str(&sse.data) {
Ok(v) => v,
Err(_) => continue,
};
let choice_opt = chunk.get("choices").and_then(|c| c.get(0));
if let Some(choice) = choice_opt {
if let Some(content) = choice
.get("delta")
.and_then(|d| d.get("content"))
.and_then(|c| c.as_str())
&& !content.is_empty()
{
append_assistant_text(&tx_event, &mut assistant_item, content.to_string()).await;
}
if let Some(reasoning_val) = choice.get("delta").and_then(|d| d.get("reasoning")) {
let mut maybe_text = reasoning_val
.as_str()
.map(str::to_string)
.filter(|s| !s.is_empty());
if maybe_text.is_none() && reasoning_val.is_object() {
if let Some(s) = reasoning_val
.get("text")
.and_then(|t| t.as_str())
.filter(|s| !s.is_empty())
{
maybe_text = Some(s.to_string());
} else if let Some(s) = reasoning_val
.get("content")
.and_then(|t| t.as_str())
.filter(|s| !s.is_empty())
{
maybe_text = Some(s.to_string());
}
}
if let Some(reasoning) = maybe_text {
append_reasoning_text(&tx_event, &mut reasoning_item, reasoning).await;
}
}
if let Some(message_reasoning) = choice.get("message").and_then(|m| m.get("reasoning"))
{
if let Some(s) = message_reasoning.as_str() {
if !s.is_empty() {
append_reasoning_text(&tx_event, &mut reasoning_item, s.to_string()).await;
}
} else if let Some(obj) = message_reasoning.as_object()
&& let Some(s) = obj
.get("text")
.and_then(|v| v.as_str())
.or_else(|| obj.get("content").and_then(|v| v.as_str()))
&& !s.is_empty()
{
append_reasoning_text(&tx_event, &mut reasoning_item, s.to_string()).await;
}
}
if let Some(tool_calls) = choice
.get("delta")
.and_then(|d| d.get("tool_calls"))
.and_then(|v| v.as_array())
{
for call in tool_calls {
if let Some(index) = call.get("index").and_then(serde_json::Value::as_u64)
&& index == 0
&& let Some(function) = call.get("function")
{
if let Some(name) = function.get("name").and_then(|n| n.as_str()) {
function_call_state.name = Some(name.to_string());
}
if let Some(arguments) = function.get("arguments").and_then(|a| a.as_str())
{
function_call_state.arguments.push_str(arguments);
}
if let Some(id) = call.get("id").and_then(|i| i.as_str()) {
function_call_state.call_id = Some(id.to_string());
}
if let Some(finish) = choice.get("finish_reason").and_then(|f| f.as_str())
&& finish == "tool_calls"
&& let Some(name) = function_call_state.name.take()
{
let call_id = function_call_state.call_id.take().unwrap_or_default();
let arguments = std::mem::take(&mut function_call_state.arguments);
let item = ResponseItem::FunctionCall {
id: None,
name,
arguments,
call_id,
};
let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
}
}
}
}
}
}
}
pub trait AggregateStreamExt: Stream<Item = Result<ResponseEvent>> + Sized {
fn aggregate(self) -> AggregatedChatStream<Self>
where
Self: Unpin,
{
AggregatedChatStream::new(self, AggregateMode::AggregatedOnly)
}
fn streaming_mode(self) -> AggregatedChatStream<Self>
where
Self: Unpin,
{
AggregatedChatStream::new(self, AggregateMode::Streaming)
}
}
impl<S> AggregateStreamExt for S where S: Stream<Item = Result<ResponseEvent>> + Sized + Unpin {}
enum AggregateMode {
AggregatedOnly,
Streaming,
}
pub struct AggregatedChatStream<S> {
inner: S,
cumulative: String,
cumulative_reasoning: String,
pending: VecDeque<ResponseEvent>,
mode: AggregateMode,
}
impl<S> AggregatedChatStream<S>
where
S: Stream<Item = Result<ResponseEvent>> + Unpin,
{
fn new(inner: S, mode: AggregateMode) -> Self {
Self {
inner,
cumulative: String::new(),
cumulative_reasoning: String::new(),
pending: VecDeque::new(),
mode,
}
}
}
impl<S> Stream for AggregatedChatStream<S>
where
S: Stream<Item = Result<ResponseEvent>> + Unpin,
{
type Item = Result<ResponseEvent>;
fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<Self::Item>> {
if let Some(ev) = self.pending.pop_front() {
return Poll::Ready(Some(Ok(ev)));
}
loop {
match Pin::new(&mut self.inner).poll_next(cx) {
Poll::Pending => return Poll::Pending,
Poll::Ready(None) => return Poll::Ready(None),
Poll::Ready(Some(Err(e))) => return Poll::Ready(Some(Err(e))),
Poll::Ready(Some(Ok(ResponseEvent::OutputItemDone(item)))) => {
let is_assistant_message = matches!(
&item,
ResponseItem::Message { role, .. } if role == "assistant"
);
if is_assistant_message {
match self.mode {
AggregateMode::AggregatedOnly => {
if self.cumulative.is_empty()
&& let ResponseItem::Message { content, .. } = &item
&& let Some(text) = content.iter().find_map(|c| match c {
ContentItem::OutputText { text } => Some(text),
_ => None,
})
{
self.cumulative.push_str(text);
}
continue;
}
AggregateMode::Streaming => {
if self.cumulative.is_empty() {
return Poll::Ready(Some(Ok(ResponseEvent::OutputItemDone(
item,
))));
} else {
continue;
}
}
}
}
return Poll::Ready(Some(Ok(ResponseEvent::OutputItemDone(item))));
}
Poll::Ready(Some(Ok(ResponseEvent::RateLimits(snapshot)))) => {
return Poll::Ready(Some(Ok(ResponseEvent::RateLimits(snapshot))));
}
Poll::Ready(Some(Ok(ResponseEvent::Completed {
response_id,
token_usage,
}))) => {
let mut emitted_any = false;
if !self.cumulative_reasoning.is_empty()
&& matches!(self.mode, AggregateMode::AggregatedOnly)
{
let aggregated_reasoning = ResponseItem::Reasoning {
id: String::new(),
summary: Vec::new(),
content: Some(vec![ReasoningItemContent::ReasoningText {
text: std::mem::take(&mut self.cumulative_reasoning),
}]),
encrypted_content: None,
};
self.pending
.push_back(ResponseEvent::OutputItemDone(aggregated_reasoning));
emitted_any = true;
}
if !self.cumulative.is_empty() {
let aggregated_message = ResponseItem::Message {
id: None,
role: "assistant".to_string(),
content: vec![ContentItem::OutputText {
text: std::mem::take(&mut self.cumulative),
}],
};
self.pending
.push_back(ResponseEvent::OutputItemDone(aggregated_message));
emitted_any = true;
}
if emitted_any {
self.pending.push_back(ResponseEvent::Completed {
response_id: response_id.clone(),
token_usage: token_usage.clone(),
});
if let Some(ev) = self.pending.pop_front() {
return Poll::Ready(Some(Ok(ev)));
}
}
return Poll::Ready(Some(Ok(ResponseEvent::Completed {
response_id,
token_usage,
})));
}
Poll::Ready(Some(Ok(ResponseEvent::Created))) => continue,
Poll::Ready(Some(Ok(ResponseEvent::OutputTextDelta(delta)))) => {
self.cumulative.push_str(&delta);
if matches!(self.mode, AggregateMode::Streaming) {
return Poll::Ready(Some(Ok(ResponseEvent::OutputTextDelta(delta))));
}
}
Poll::Ready(Some(Ok(ResponseEvent::ReasoningContentDelta(delta)))) => {
self.cumulative_reasoning.push_str(&delta);
if matches!(self.mode, AggregateMode::Streaming) {
return Poll::Ready(Some(Ok(ResponseEvent::ReasoningContentDelta(delta))));
}
}
Poll::Ready(Some(Ok(ResponseEvent::ReasoningSummaryDelta(_)))) => continue,
Poll::Ready(Some(Ok(ResponseEvent::ReasoningSummaryPartAdded))) => continue,
Poll::Ready(Some(Ok(ResponseEvent::OutputItemAdded(item)))) => {
return Poll::Ready(Some(Ok(ResponseEvent::OutputItemAdded(item))));
}
}
}
}
}
fn create_tools_json_for_chat_completions_api(
tools: &[serde_json::Value],
) -> Result<Vec<serde_json::Value>> {
let tools_json = tools
.iter()
.filter_map(|tool| {
if tool.get("type") != Some(&serde_json::Value::String("function".to_string())) {
return None;
}
let function_value = if let Some(function) = tool.get("function") {
function.clone()
} else if let Some(map) = tool.as_object() {
let mut function = map.clone();
function.remove("type");
Value::Object(function)
} else {
return None;
};
Some(json!({
"type": "function",
"function": function_value,
}))
})
.collect::<Vec<serde_json::Value>>();
Ok(tools_json)
}
fn backoff(attempt: u64) -> Duration {
let capped = attempt.min(6);
Duration::from_millis(100 * 2u64.pow(capped as u32))
}

View File

@@ -0,0 +1,42 @@
use std::time::Duration;
use thiserror::Error;
pub type Result<T> = std::result::Result<T, Error>;
#[derive(Debug, Error)]
pub enum Error {
#[error("{0}")]
UnsupportedOperation(String),
#[error(transparent)]
Http(#[from] reqwest::Error),
#[error("{source}")]
ResponseStreamFailed {
#[source]
source: reqwest::Error,
request_id: Option<String>,
},
#[error("{0}")]
Stream(String, Option<Duration>),
#[error("unexpected status {status}: {body}")]
UnexpectedStatus {
status: reqwest::StatusCode,
body: String,
},
#[error("retry limit reached (status {status}, request id: {request_id:?})")]
RetryLimit {
status: reqwest::StatusCode,
request_id: Option<String>,
},
#[error("missing environment variable {var}")]
MissingEnvVar {
var: String,
instructions: Option<String>,
},
#[error("{0}")]
Auth(String),
#[error(transparent)]
Json(#[from] serde_json::Error),
#[error("{0}")]
Other(String),
}

View File

@@ -0,0 +1,35 @@
pub mod api;
pub mod auth;
pub mod chat;
pub mod error;
pub mod model_provider;
pub mod prompt;
pub mod responses;
pub mod stream;
pub use crate::api::ApiClient;
pub use crate::auth::AuthContext;
pub use crate::auth::AuthProvider;
pub use crate::chat::AggregateStreamExt;
pub use crate::chat::ChatAggregationMode;
pub use crate::chat::ChatCompletionsApiClient;
pub use crate::chat::ChatCompletionsApiClientConfig;
pub use crate::error::Error;
pub use crate::error::Result;
pub use crate::model_provider::BUILT_IN_OSS_MODEL_PROVIDER_ID;
pub use crate::model_provider::ModelProviderInfo;
pub use crate::model_provider::WireApi;
pub use crate::model_provider::built_in_model_providers;
pub use crate::model_provider::create_oss_provider;
pub use crate::model_provider::create_oss_provider_with_base_url;
pub use crate::prompt::Prompt;
pub use crate::responses::ResponsesApiClient;
pub use crate::responses::ResponsesApiClientConfig;
pub use crate::responses::stream_from_fixture;
pub use crate::stream::EventStream;
pub use crate::stream::Reasoning;
pub use crate::stream::ResponseEvent;
pub use crate::stream::ResponseStream;
pub use crate::stream::TextControls;
pub use crate::stream::TextFormat;
pub use crate::stream::TextFormatType;

View File

@@ -5,17 +5,18 @@
//! 2. User-defined entries inside `~/.codex/config.toml` under the `model_providers`
//! key. These override or extend the defaults at runtime.
use crate::CodexAuth;
use crate::default_client::CodexHttpClient;
use crate::default_client::CodexRequestBuilder;
use codex_app_server_protocol::AuthMode;
use serde::Deserialize;
use serde::Serialize;
use std::collections::HashMap;
use std::env::VarError;
use std::time::Duration;
use crate::error::EnvVarError;
use codex_app_server_protocol::AuthMode;
use serde::Deserialize;
use serde::Serialize;
use crate::auth::AuthContext;
use crate::error::Error;
use crate::error::Result;
const DEFAULT_STREAM_IDLE_TIMEOUT_MS: u64 = 300_000;
const DEFAULT_STREAM_MAX_RETRIES: u64 = 5;
const DEFAULT_REQUEST_MAX_RETRIES: u64 = 4;
@@ -23,19 +24,19 @@ const DEFAULT_REQUEST_MAX_RETRIES: u64 = 4;
const MAX_STREAM_MAX_RETRIES: u64 = 100;
/// Hard cap for user-configured `request_max_retries`.
const MAX_REQUEST_MAX_RETRIES: u64 = 100;
const DEFAULT_OLLAMA_PORT: u32 = 11434;
/// Wire protocol that the provider speaks. Most third-party services only
/// implement the classic OpenAI Chat Completions JSON schema, whereas OpenAI
/// itself (and a handful of others) additionally expose the more modern
/// *Responses* API. The two protocols use different request/response shapes
/// and *cannot* be auto-detected at runtime, therefore each provider entry
/// Responses API. The two protocols use different request/response shapes
/// and cannot be auto-detected at runtime, therefore each provider entry
/// must declare which one it expects.
#[derive(Debug, Clone, Copy, Default, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "lowercase")]
pub enum WireApi {
/// The Responses API exposed by OpenAI at `/v1/responses`.
Responses,
/// Regular Chat Completions compatible with `/v1/chat/completions`.
#[default]
Chat,
@@ -50,87 +51,79 @@ pub struct ModelProviderInfo {
pub base_url: Option<String>,
/// Environment variable that stores the user's API key for this provider.
pub env_key: Option<String>,
/// Optional instructions to help the user get a valid value for the
/// variable and set it.
pub env_key_instructions: Option<String>,
/// Value to use with `Authorization: Bearer <token>` header. Use of this
/// config is discouraged in favor of `env_key` for security reasons, but
/// this may be necessary when using this programmatically.
pub experimental_bearer_token: Option<String>,
/// Which wire protocol this provider expects.
#[serde(default)]
pub wire_api: WireApi,
/// Optional query parameters to append to the base URL.
pub query_params: Option<HashMap<String, String>>,
/// Additional HTTP headers to include in requests to this provider where
/// the (key, value) pairs are the header name and value.
pub http_headers: Option<HashMap<String, String>>,
/// Optional HTTP headers to include in requests to this provider where the
/// (key, value) pairs are the header name and _environment variable_ whose
/// (key, value) pairs are the header name and environment variable whose
/// value should be used. If the environment variable is not set, or the
/// value is empty, the header will not be included in the request.
pub env_http_headers: Option<HashMap<String, String>>,
/// Maximum number of times to retry a failed HTTP request to this provider.
pub request_max_retries: Option<u64>,
/// Number of times to retry reconnecting a dropped streaming response before failing.
pub stream_max_retries: Option<u64>,
/// Idle timeout (in milliseconds) to wait for activity on a streaming response before treating
/// the connection as lost.
pub stream_idle_timeout_ms: Option<u64>,
/// Does this provider require an OpenAI API Key or ChatGPT login token? If true,
/// user is presented with login screen on first run, and login preference and token/key
/// are stored in auth.json. If false (which is the default), login screen is skipped,
/// and API key (if needed) comes from the "env_key" environment variable.
/// the user is presented with a login screen on first run, and login preference and token/key
/// are stored in auth.json. If false (which is the default), the login screen is skipped,
/// and the API key (if needed) comes from the `env_key` environment variable.
#[serde(default)]
pub requires_openai_auth: bool,
}
impl ModelProviderInfo {
/// Construct a `POST` RequestBuilder for the given URL using the provided
/// [`CodexHttpClient`] applying:
/// provider-specific headers (static + env based)
/// Bearer auth header when an API key is available.
/// Auth token for OAuth.
/// Construct a `POST` request builder for the given URL using the provided
/// [`reqwest::Client`] applying:
/// - provider-specific headers (static and environment based)
/// - Bearer auth header when an API key is available
/// - Auth token for OAuth
///
/// If the provider declares an `env_key` but the variable is missing/empty, returns an [`Err`] identical to the
/// one produced by [`ModelProviderInfo::api_key`].
pub async fn create_request_builder<'a>(
&'a self,
client: &'a CodexHttpClient,
auth: &Option<CodexAuth>,
) -> crate::error::Result<CodexRequestBuilder> {
/// If the provider declares an `env_key` but the variable is missing or empty, this returns an
/// error identical to the one produced by [`ModelProviderInfo::api_key`].
pub async fn create_request_builder(
&self,
client: &reqwest::Client,
auth: &Option<AuthContext>,
) -> Result<reqwest::RequestBuilder> {
let effective_auth = if let Some(secret_key) = &self.experimental_bearer_token {
Some(CodexAuth::from_api_key(secret_key))
Some(AuthContext {
mode: AuthMode::ApiKey,
bearer_token: Some(secret_key.clone()),
account_id: None,
})
} else {
match self.api_key() {
Ok(Some(key)) => Some(CodexAuth::from_api_key(&key)),
Ok(None) => auth.clone(),
Err(err) => {
if auth.is_some() {
auth.clone()
} else {
return Err(err);
}
}
match self.api_key()? {
Some(key) => Some(AuthContext {
mode: AuthMode::ApiKey,
bearer_token: Some(key),
account_id: None,
}),
None => auth.clone(),
}
};
let url = self.get_full_url(&effective_auth);
let url = self.get_full_url(effective_auth.as_ref());
let mut builder = client.post(url);
if let Some(auth) = effective_auth.as_ref() {
builder = builder.bearer_auth(auth.get_token().await?);
if let Some(context) = effective_auth.as_ref()
&& let Some(token) = context.bearer_token.as_ref()
{
builder = builder.bearer_auth(token);
}
Ok(self.apply_http_headers(builder))
@@ -149,10 +142,10 @@ impl ModelProviderInfo {
})
}
pub(crate) fn get_full_url(&self, auth: &Option<CodexAuth>) -> String {
pub fn get_full_url(&self, auth: Option<&AuthContext>) -> String {
let default_base_url = if matches!(
auth,
Some(CodexAuth {
Some(AuthContext {
mode: AuthMode::ChatGPT,
..
})
@@ -165,7 +158,7 @@ impl ModelProviderInfo {
let base_url = self
.base_url
.clone()
.unwrap_or(default_base_url.to_string());
.unwrap_or_else(|| default_base_url.to_string());
match self.wire_api {
WireApi::Responses => format!("{base_url}/responses{query_string}"),
@@ -173,7 +166,7 @@ impl ModelProviderInfo {
}
}
pub(crate) fn is_azure_responses_endpoint(&self) -> bool {
pub fn is_azure_responses_endpoint(&self) -> bool {
if self.wire_api != WireApi::Responses {
return false;
}
@@ -188,10 +181,9 @@ impl ModelProviderInfo {
.unwrap_or(false)
}
/// Apply provider-specific HTTP headers (both static and environment-based)
/// onto an existing [`CodexRequestBuilder`] and return the updated
/// builder.
fn apply_http_headers(&self, mut builder: CodexRequestBuilder) -> CodexRequestBuilder {
/// Apply provider-specific HTTP headers (both static and environment-based) onto an existing
/// [`reqwest::RequestBuilder`] and return the updated builder.
fn apply_http_headers(&self, mut builder: reqwest::RequestBuilder) -> reqwest::RequestBuilder {
if let Some(extra) = &self.http_headers {
for (k, v) in extra {
builder = builder.header(k, v);
@@ -210,10 +202,9 @@ impl ModelProviderInfo {
builder
}
/// If `env_key` is Some, returns the API key for this provider if present
/// (and non-empty) in the environment. If `env_key` is required but
/// cannot be found, returns an error.
pub fn api_key(&self) -> crate::error::Result<Option<String>> {
/// If `env_key` is `Some`, returns the API key for this provider if present (and non-empty) in
/// the environment. If `env_key` is required but cannot be found, returns an error.
pub fn api_key(&self) -> Result<Option<String>> {
match &self.env_key {
Some(env_key) => {
let env_value = std::env::var(env_key);
@@ -225,11 +216,9 @@ impl ModelProviderInfo {
Ok(Some(v))
}
})
.map_err(|_| {
crate::error::CodexErr::EnvVar(EnvVarError {
var: env_key.clone(),
instructions: self.env_key_instructions.clone(),
})
.map_err(|_| Error::MissingEnvVar {
var: env_key.clone(),
instructions: self.env_key_instructions.clone(),
})
}
None => Ok(None),
@@ -258,28 +247,23 @@ impl ModelProviderInfo {
}
}
const DEFAULT_OLLAMA_PORT: u32 = 11434;
pub const BUILT_IN_OSS_MODEL_PROVIDER_ID: &str = "oss";
/// Built-in default provider list.
pub fn built_in_model_providers() -> HashMap<String, ModelProviderInfo> {
use ModelProviderInfo as P;
// We do not want to be in the business of adjucating which third-party
// providers are bundled with Codex CLI, so we only include the OpenAI and
// open source ("oss") providers by default. Users are encouraged to add to
// `model_providers` in config.toml to add their own providers.
// We do not want to be in the business of adjudicating which third-party providers are bundled
// with Codex CLI, so we only include the OpenAI and open source ("oss") providers by default.
// Users are encouraged to add to `model_providers` in config.toml to add their own providers.
[
(
"openai",
P {
name: "OpenAI".into(),
// Allow users to override the default OpenAI endpoint by
// exporting `OPENAI_BASE_URL`. This is useful when pointing
// Codex at a proxy, mock server, or Azure-style deployment
// without requiring a full TOML override for the built-in
// OpenAI provider.
// Allow users to override the default OpenAI endpoint by exporting `OPENAI_BASE_URL`.
// This is useful when pointing Codex at a proxy, mock server, or Azure-style
// deployment without requiring a full TOML override for the built-in OpenAI provider.
base_url: std::env::var("OPENAI_BASE_URL")
.ok()
.filter(|v| !v.trim().is_empty()),
@@ -318,9 +302,10 @@ pub fn built_in_model_providers() -> HashMap<String, ModelProviderInfo> {
.collect()
}
/// Convenience helper for the built-in OSS provider.
pub fn create_oss_provider() -> ModelProviderInfo {
// These CODEX_OSS_ environment variables are experimental: we may
// switch to reading values from config.toml instead.
// These CODEX_OSS_ environment variables are experimental: we may switch to reading values from
// config.toml instead.
let codex_oss_base_url = match std::env::var("CODEX_OSS_BASE_URL")
.ok()
.filter(|v| !v.trim().is_empty())
@@ -366,23 +351,23 @@ fn matches_azure_responses_base_url(base_url: &str) -> bool {
"azure-api.",
"azurefd.",
];
AZURE_MARKERS.iter().any(|marker| base.contains(marker))
AZURE_MARKERS.iter().any(|needle| base.contains(needle))
}
#[cfg(test)]
mod tests {
use super::*;
use pretty_assertions::assert_eq;
use maplit::hashmap;
#[test]
fn test_deserialize_ollama_model_provider_toml() {
fn deserializes_defaults_without_optional_fields() {
let azure_provider_toml = r#"
name = "Ollama"
base_url = "http://localhost:11434/v1"
name = "Azure"
base_url = "https://xxxxx.openai.azure.com/openai"
"#;
let expected_provider = ModelProviderInfo {
name: "Ollama".into(),
base_url: Some("http://localhost:11434/v1".into()),
name: "Azure".into(),
base_url: Some("https://xxxxx.openai.azure.com/openai".into()),
env_key: None,
env_key_instructions: None,
experimental_bearer_token: None,
@@ -415,7 +400,7 @@ query_params = { api-version = "2025-04-01-preview" }
env_key_instructions: None,
experimental_bearer_token: None,
wire_api: WireApi::Chat,
query_params: Some(maplit::hashmap! {
query_params: Some(hashmap! {
"api-version".to_string() => "2025-04-01-preview".to_string(),
}),
http_headers: None,
@@ -447,10 +432,10 @@ env_http_headers = { "X-Example-Env-Header" = "EXAMPLE_ENV_VAR" }
experimental_bearer_token: None,
wire_api: WireApi::Chat,
query_params: None,
http_headers: Some(maplit::hashmap! {
http_headers: Some(hashmap! {
"X-Example-Header".to_string() => "example-value".to_string(),
}),
env_http_headers: Some(maplit::hashmap! {
env_http_headers: Some(hashmap! {
"X-Example-Env-Header".to_string() => "EXAMPLE_ENV_VAR".to_string(),
}),
request_max_retries: None,
@@ -516,16 +501,12 @@ env_http_headers = { "X-Example-Env-Header" = "EXAMPLE_ENV_VAR" }
};
assert!(named_provider.is_azure_responses_endpoint());
let negative_cases = [
"https://api.openai.com/v1",
"https://example.com/openai",
"https://myproxy.azurewebsites.net/openai",
];
let negative_cases = ["https://api.openai.com/v1", "https://example.com"];
for base_url in negative_cases {
let provider = provider_for(base_url);
assert!(
!provider.is_azure_responses_endpoint(),
"expected {base_url} not to be detected as Azure"
"expected {base_url} to be non-Azure"
);
}
}

View File

@@ -0,0 +1,49 @@
use codex_protocol::models::ResponseItem;
use codex_protocol::protocol::SessionSource;
use serde_json::Value;
use crate::Reasoning;
use crate::TextControls;
#[derive(Debug, Clone, Default)]
pub struct Prompt {
pub instructions: String,
pub input: Vec<ResponseItem>,
pub tools: Vec<Value>,
pub parallel_tool_calls: bool,
pub output_schema: Option<Value>,
pub reasoning: Option<Reasoning>,
pub text_controls: Option<TextControls>,
pub prompt_cache_key: Option<String>,
pub previous_response_id: Option<String>,
pub session_source: Option<SessionSource>,
}
impl Prompt {
#[allow(clippy::too_many_arguments)]
pub fn new(
instructions: String,
input: Vec<ResponseItem>,
tools: Vec<Value>,
parallel_tool_calls: bool,
output_schema: Option<Value>,
reasoning: Option<Reasoning>,
text_controls: Option<TextControls>,
prompt_cache_key: Option<String>,
previous_response_id: Option<String>,
session_source: Option<SessionSource>,
) -> Self {
Self {
instructions,
input,
tools,
parallel_tool_calls,
output_schema,
reasoning,
text_controls,
prompt_cache_key,
previous_response_id,
session_source,
}
}
}

View File

@@ -0,0 +1,742 @@
use std::io::BufRead;
use std::path::Path;
use std::sync::Arc;
use std::sync::OnceLock;
use std::time::Duration;
use async_trait::async_trait;
use bytes::Bytes;
use codex_app_server_protocol::AuthMode;
use codex_otel::otel_event_manager::OtelEventManager;
use codex_protocol::ConversationId;
use codex_protocol::models::ResponseItem;
use codex_protocol::protocol::RateLimitSnapshot;
use codex_protocol::protocol::RateLimitWindow;
use codex_protocol::protocol::SessionSource;
use codex_protocol::protocol::SubAgentSource;
use codex_protocol::protocol::TokenUsage;
use eventsource_stream::Eventsource;
use futures::Stream;
use futures::StreamExt;
use futures::TryStreamExt;
use regex_lite::Regex;
use reqwest::StatusCode;
use reqwest::header::HeaderMap;
use serde::Deserialize;
use serde::Serialize;
use serde_json::Value;
use serde_json::json;
use tokio::sync::mpsc;
use tokio::time::timeout;
use tokio_util::io::ReaderStream;
use tracing::debug;
use tracing::trace;
use crate::api::ApiClient;
use crate::auth::AuthProvider;
use crate::error::Error;
use crate::model_provider::ModelProviderInfo;
use crate::prompt::Prompt;
use crate::stream::ResponseEvent;
use crate::stream::ResponseStream;
type Result<T> = std::result::Result<T, Error>;
#[derive(Clone)]
pub struct ResponsesApiClientConfig {
pub http_client: reqwest::Client,
pub provider: ModelProviderInfo,
pub model: String,
pub conversation_id: ConversationId,
pub auth_provider: Option<Arc<dyn AuthProvider>>,
pub otel_event_manager: OtelEventManager,
}
#[derive(Clone)]
pub struct ResponsesApiClient {
config: ResponsesApiClientConfig,
}
#[async_trait]
impl ApiClient for ResponsesApiClient {
type Config = ResponsesApiClientConfig;
async fn new(config: Self::Config) -> Result<Self> {
Ok(Self { config })
}
async fn stream(&self, prompt: Prompt) -> Result<ResponseStream> {
if self.config.provider.wire_api != crate::model_provider::WireApi::Responses {
return Err(Error::UnsupportedOperation(
"ResponsesApiClient requires a Responses provider".to_string(),
));
}
let mut payload_json = self.build_payload(&prompt)?;
if self.config.provider.is_azure_responses_endpoint()
&& let Some(input_value) = payload_json.get_mut("input")
&& let Some(array) = input_value.as_array_mut()
{
attach_item_ids_array(array, &prompt.input);
}
let max_attempts = self.config.provider.request_max_retries();
for attempt in 0..=max_attempts {
match self
.attempt_stream_responses(attempt, &prompt, &payload_json)
.await
{
Ok(stream) => return Ok(stream),
Err(StreamAttemptError::Fatal(err)) => return Err(err),
Err(retryable) => {
if attempt == max_attempts {
return Err(retryable.into_error());
}
tokio::time::sleep(retryable.delay(attempt)).await;
}
}
}
unreachable!("attempt_stream_responses should always return");
}
}
impl ResponsesApiClient {
fn build_payload(&self, prompt: &Prompt) -> Result<Value> {
let azure_workaround = self.config.provider.is_azure_responses_endpoint();
let mut payload = json!({
"model": self.config.model,
"instructions": prompt.instructions,
"input": prompt.input,
"tools": prompt.tools,
"tool_choice": "auto",
"parallel_tool_calls": prompt.parallel_tool_calls,
"store": azure_workaround,
"stream": true,
"prompt_cache_key": prompt
.prompt_cache_key
.clone()
.unwrap_or_else(|| self.config.conversation_id.to_string()),
});
if let Some(reasoning) = prompt.reasoning.as_ref()
&& let Some(obj) = payload.as_object_mut()
{
obj.insert("reasoning".to_string(), serde_json::to_value(reasoning)?);
}
if let Some(text) = prompt.text_controls.as_ref()
&& let Some(obj) = payload.as_object_mut()
{
obj.insert("text".to_string(), serde_json::to_value(text)?);
}
if let Some(prev) = prompt.previous_response_id.as_ref()
&& let Some(obj) = payload.as_object_mut()
{
obj.insert(
"previous_response_id".to_string(),
Value::String(prev.clone()),
);
}
let include = if prompt.reasoning.is_some() {
vec!["reasoning.encrypted_content".to_string()]
} else {
Vec::new()
};
if let Some(obj) = payload.as_object_mut() {
obj.insert(
"include".to_string(),
Value::Array(include.into_iter().map(Value::String).collect()),
);
}
Ok(payload)
}
async fn attempt_stream_responses(
&self,
attempt: u64,
prompt: &Prompt,
payload_json: &Value,
) -> std::result::Result<ResponseStream, StreamAttemptError> {
let auth = match &self.config.auth_provider {
Some(provider) => provider.auth_context().await,
None => None,
};
trace!(
"POST to {}: {:?}",
self.config.provider.get_full_url(auth.as_ref()),
serde_json::to_string(payload_json)
.unwrap_or_else(|_| "<unable to serialize payload>".to_string())
);
let mut req_builder = self
.config
.provider
.create_request_builder(&self.config.http_client, &auth)
.await
.map_err(StreamAttemptError::Fatal)?;
if let Some(SessionSource::SubAgent(sub)) = prompt.session_source.as_ref() {
let subagent = match sub {
SubAgentSource::Other(label) => label.clone(),
other => serde_json::to_value(other)
.ok()
.and_then(|v| v.as_str().map(ToString::to_string))
.unwrap_or_else(|| "other".to_string()),
};
req_builder = req_builder.header("x-openai-subagent", subagent);
}
req_builder = req_builder
.header("conversation_id", self.config.conversation_id.to_string())
.header("session_id", self.config.conversation_id.to_string())
.header(reqwest::header::ACCEPT, "text/event-stream")
.json(payload_json);
if let Some(ctx) = auth.as_ref()
&& ctx.mode == AuthMode::ChatGPT
&& let Some(account_id) = ctx.account_id.as_ref()
{
req_builder = req_builder.header("chatgpt-account-id", account_id);
}
let res = self
.config
.otel_event_manager
.log_request(attempt, || req_builder.send())
.await;
let mut request_id = None;
if let Ok(resp) = &res {
request_id = resp
.headers()
.get("cf-ray")
.and_then(|v| v.to_str().ok())
.map(std::string::ToString::to_string);
}
match res {
Ok(resp) if resp.status().is_success() => {
let (tx_event, rx_event) = mpsc::channel::<Result<ResponseEvent>>(1600);
if let Some(snapshot) = parse_rate_limit_snapshot(resp.headers())
&& tx_event
.send(Ok(ResponseEvent::RateLimits(snapshot)))
.await
.is_err()
{
debug!("receiver dropped rate limit snapshot event");
}
let request_id_for_stream = request_id.clone();
let stream = resp
.bytes_stream()
.map_err(move |err| Error::ResponseStreamFailed {
source: err,
request_id: request_id_for_stream.clone(),
});
tokio::spawn(process_sse(
stream,
tx_event,
self.config.provider.stream_idle_timeout(),
self.config.otel_event_manager.clone(),
));
Ok(ResponseStream { rx_event })
}
Ok(res) => {
let status = res.status();
let retry_after_secs = res
.headers()
.get(reqwest::header::RETRY_AFTER)
.and_then(|v| v.to_str().ok())
.and_then(|s| s.parse::<u64>().ok());
let retry_after = retry_after_secs.map(|s| Duration::from_millis(s * 1_000));
if status == StatusCode::UNAUTHORIZED
&& let Some(provider) = self.config.auth_provider.as_ref()
&& let Some(ctx) = auth.as_ref()
&& ctx.mode == AuthMode::ChatGPT
{
provider
.refresh_token()
.await
.map_err(|err| StreamAttemptError::Fatal(Error::Auth(err)))?;
}
if !(status == StatusCode::TOO_MANY_REQUESTS
|| status == StatusCode::UNAUTHORIZED
|| status.is_server_error())
{
// Surface error body.
let body = res
.text()
.await
.unwrap_or_else(|_| "<failed to read response>".to_string());
return Err(StreamAttemptError::Fatal(Error::UnexpectedStatus {
status,
body,
}));
}
Err(StreamAttemptError::RetryableHttpError {
status,
retry_after,
request_id,
})
}
Err(err) => Err(StreamAttemptError::RetryableTransportError(Error::Http(
err,
))),
}
}
}
enum StreamAttemptError {
RetryableHttpError {
status: StatusCode,
retry_after: Option<Duration>,
request_id: Option<String>,
},
RetryableTransportError(Error),
Fatal(Error),
}
impl StreamAttemptError {
fn delay(&self, attempt: u64) -> Duration {
let backoff_attempt = attempt + 1;
match self {
StreamAttemptError::RetryableHttpError { retry_after, .. } => {
retry_after.unwrap_or_else(|| backoff(backoff_attempt))
}
StreamAttemptError::RetryableTransportError { .. } => backoff(backoff_attempt),
StreamAttemptError::Fatal(_) => Duration::from_secs(0),
}
}
fn into_error(self) -> Error {
match self {
StreamAttemptError::RetryableHttpError {
status, request_id, ..
} => Error::RetryLimit { status, request_id },
StreamAttemptError::RetryableTransportError(error) => error,
StreamAttemptError::Fatal(error) => error,
}
}
}
#[derive(Debug, Deserialize, Serialize)]
struct SseEvent {
#[serde(rename = "type")]
kind: String,
response: Option<Value>,
item: Option<Value>,
delta: Option<String>,
}
#[derive(Debug, Deserialize)]
struct ResponseCompleted {
id: String,
usage: Option<ResponseCompletedUsage>,
}
#[derive(Debug, Deserialize)]
struct ResponseCompletedUsage {
input_tokens: i64,
input_tokens_details: Option<ResponseCompletedInputTokensDetails>,
output_tokens: i64,
output_tokens_details: Option<ResponseCompletedOutputTokensDetails>,
total_tokens: i64,
}
impl From<ResponseCompletedUsage> for TokenUsage {
fn from(val: ResponseCompletedUsage) -> Self {
TokenUsage {
input_tokens: val.input_tokens,
cached_input_tokens: val
.input_tokens_details
.map(|d| d.cached_tokens)
.unwrap_or(0),
output_tokens: val.output_tokens,
reasoning_output_tokens: val
.output_tokens_details
.map(|d| d.reasoning_tokens)
.unwrap_or(0),
total_tokens: val.total_tokens,
}
}
}
#[derive(Debug, Deserialize)]
struct ResponseCompletedInputTokensDetails {
cached_tokens: i64,
}
#[derive(Debug, Deserialize)]
struct ResponseCompletedOutputTokensDetails {
reasoning_tokens: i64,
}
fn attach_item_ids_array(items: &mut [Value], original_items: &[ResponseItem]) {
for (value, item) in items.iter_mut().zip(original_items.iter()) {
if let ResponseItem::Reasoning { id, .. }
| ResponseItem::Message { id: Some(id), .. }
| ResponseItem::WebSearchCall { id: Some(id), .. }
| ResponseItem::FunctionCall { id: Some(id), .. }
| ResponseItem::LocalShellCall { id: Some(id), .. }
| ResponseItem::CustomToolCall { id: Some(id), .. }
| ResponseItem::CustomToolCallOutput { call_id: id, .. }
| ResponseItem::FunctionCallOutput { call_id: id, .. } = item
{
if id.is_empty() {
continue;
}
if let Some(obj) = value.as_object_mut() {
obj.insert("id".to_string(), Value::String(id.clone()));
}
}
}
}
fn parse_rate_limit_snapshot(headers: &HeaderMap) -> Option<RateLimitSnapshot> {
let primary = parse_rate_limit_window(
headers,
"x-codex-primary-used-percent",
"x-codex-primary-window-minutes",
"x-codex-primary-reset-at",
);
let secondary = parse_rate_limit_window(
headers,
"x-codex-secondary-used-percent",
"x-codex-secondary-window-minutes",
"x-codex-secondary-reset-at",
);
Some(RateLimitSnapshot { primary, secondary })
}
fn parse_rate_limit_window(
headers: &HeaderMap,
used_percent_header: &str,
window_minutes_header: &str,
resets_at_header: &str,
) -> Option<RateLimitWindow> {
let used_percent: Option<f64> = parse_header_f64(headers, used_percent_header);
used_percent.and_then(|used_percent| {
let window_minutes = parse_header_i64(headers, window_minutes_header);
let resets_at = parse_header_i64(headers, resets_at_header);
let has_data = used_percent != 0.0
|| window_minutes.is_some_and(|minutes| minutes != 0)
|| resets_at.is_some();
has_data.then_some(RateLimitWindow {
used_percent,
window_minutes,
resets_at,
})
})
}
fn parse_header_f64(headers: &HeaderMap, name: &str) -> Option<f64> {
parse_header_str(headers, name)?
.parse::<f64>()
.ok()
.filter(|v| v.is_finite())
}
fn parse_header_i64(headers: &HeaderMap, name: &str) -> Option<i64> {
parse_header_str(headers, name)?.parse::<i64>().ok()
}
fn parse_header_str<'a>(headers: &'a HeaderMap, name: &str) -> Option<&'a str> {
headers.get(name)?.to_str().ok()
}
async fn process_sse<S>(
stream: S,
tx_event: mpsc::Sender<Result<ResponseEvent>>,
idle_timeout: Duration,
otel_event_manager: OtelEventManager,
) where
S: Stream<Item = Result<Bytes>> + Unpin + Send + 'static,
{
let mut stream = stream.eventsource();
let mut response_completed: Option<ResponseCompleted> = None;
let mut response_error: Option<Error> = None;
loop {
let start = std::time::Instant::now();
let response = timeout(idle_timeout, stream.next()).await;
let duration = start.elapsed();
otel_event_manager.log_sse_event(&response, duration);
let sse = match response {
Ok(Some(Ok(sse))) => sse,
Ok(Some(Err(e))) => {
debug!("SSE Error: {e:#}");
let event = Error::Stream(e.to_string(), None);
let _ = tx_event.send(Err(event)).await;
return;
}
Ok(None) => {
match response_completed {
Some(ResponseCompleted {
id: response_id,
usage,
}) => {
if let Some(token_usage) = &usage {
otel_event_manager.sse_event_completed(
token_usage.input_tokens,
token_usage.output_tokens,
token_usage
.input_tokens_details
.as_ref()
.map(|d| d.cached_tokens),
token_usage
.output_tokens_details
.as_ref()
.map(|d| d.reasoning_tokens),
token_usage.total_tokens,
);
}
let event = ResponseEvent::Completed {
response_id,
token_usage: usage.map(Into::into),
};
let _ = tx_event.send(Ok(event)).await;
}
None => {
let error = response_error.unwrap_or(Error::Stream(
"stream closed before response.completed".into(),
None,
));
otel_event_manager.see_event_completed_failed(&error);
let _ = tx_event.send(Err(error)).await;
}
}
return;
}
Err(_) => {
let _ = tx_event
.send(Err(Error::Stream(
"idle timeout waiting for SSE".into(),
None,
)))
.await;
return;
}
};
let raw = sse.data.clone();
trace!("SSE event: {}", raw);
let event: SseEvent = match serde_json::from_str(&sse.data) {
Ok(event) => event,
Err(e) => {
debug!("Failed to parse SSE event: {e}, data: {}", &sse.data);
continue;
}
};
match event.kind.as_str() {
"response.output_item.done" => {
let Some(item_val) = event.item else { continue };
let Ok(item) = serde_json::from_value::<ResponseItem>(item_val) else {
debug!("failed to parse ResponseItem from output_item.done");
continue;
};
let event = ResponseEvent::OutputItemDone(item);
if tx_event.send(Ok(event)).await.is_err() {
return;
}
}
"response.output_text.delta" => {
if let Some(delta) = event.delta {
let event = ResponseEvent::OutputTextDelta(delta);
if tx_event.send(Ok(event)).await.is_err() {
return;
}
}
}
"response.reasoning_summary_text.delta" => {
if let Some(delta) = event.delta {
let event = ResponseEvent::ReasoningSummaryDelta(delta);
if tx_event.send(Ok(event)).await.is_err() {
return;
}
}
}
"response.reasoning_text.delta" => {
if let Some(delta) = event.delta {
let event = ResponseEvent::ReasoningContentDelta(delta);
if tx_event.send(Ok(event)).await.is_err() {
return;
}
}
}
"response.created" => {
if event.response.is_some() {
let _ = tx_event.send(Ok(ResponseEvent::Created)).await;
}
}
"response.failed" => {
if let Some(resp_val) = event.response {
response_error = Some(Error::Stream(
"response.failed event received".to_string(),
None,
));
if let Some(error) = resp_val.get("error") {
match serde_json::from_value::<ErrorResponse>(error.clone()) {
Ok(error) => {
if is_context_window_error(&error) {
response_error = Some(Error::UnsupportedOperation(
"context window exceeded".to_string(),
));
} else {
let delay = try_parse_retry_after(&error);
let message = error.message.clone().unwrap_or_default();
response_error = Some(Error::Stream(message, delay));
}
}
Err(e) => {
let error = format!("failed to parse ErrorResponse: {e}");
debug!(error);
response_error = Some(Error::Stream(error, None))
}
}
}
}
}
"response.completed" => {
if let Some(resp_val) = event.response {
match serde_json::from_value::<ResponseCompleted>(resp_val) {
Ok(r) => {
response_completed = Some(r);
}
Err(e) => {
let error = format!("failed to parse ResponseCompleted: {e}");
debug!(error);
response_error = Some(Error::Stream(error, None));
continue;
}
};
};
}
"response.output_item.added" => {
let Some(item_val) = event.item else { continue };
let Ok(item) = serde_json::from_value::<ResponseItem>(item_val) else {
debug!("failed to parse ResponseItem from output_item.done");
continue;
};
let event = ResponseEvent::OutputItemAdded(item);
if tx_event.send(Ok(event)).await.is_err() {
return;
}
}
"response.reasoning_summary_part.added" => {
let event = ResponseEvent::ReasoningSummaryPartAdded;
if tx_event.send(Ok(event)).await.is_err() {
return;
}
}
_ => {}
}
}
}
#[derive(Debug, Deserialize)]
struct ErrorResponse {
code: Option<String>,
message: Option<String>,
}
fn backoff(attempt: u64) -> Duration {
let exponent = attempt.min(6) as u32;
let base = 2u64.pow(exponent);
Duration::from_millis(base * 100)
}
fn rate_limit_regex() -> Option<&'static Regex> {
static RE: OnceLock<Option<Regex>> = OnceLock::new();
RE.get_or_init(|| Regex::new(r"Please try again in (\d+(?:\.\d+)?)(s|ms)").ok())
.as_ref()
}
fn try_parse_retry_after(err: &ErrorResponse) -> Option<Duration> {
if err.code.as_deref() != Some("rate_limit_exceeded") {
return None;
}
if let Some(re) = rate_limit_regex()
&& let Some(message) = &err.message
&& let Some(captures) = re.captures(message)
{
let seconds = captures.get(1);
let unit = captures.get(2);
if let (Some(value), Some(unit)) = (seconds, unit) {
let value = value.as_str().parse::<f64>().ok()?;
let unit = unit.as_str();
if unit == "s" {
return Some(Duration::from_secs_f64(value));
} else if unit == "ms" {
return Some(Duration::from_millis(value as u64));
}
}
}
None
}
fn is_context_window_error(error: &ErrorResponse) -> bool {
error.code.as_deref() == Some("context_length_exceeded")
}
/// used in tests to stream from a text SSE file
pub async fn stream_from_fixture(
path: impl AsRef<Path>,
provider: ModelProviderInfo,
otel_event_manager: OtelEventManager,
) -> Result<ResponseStream> {
let (tx_event, rx_event) = mpsc::channel::<Result<ResponseEvent>>(1600);
let display_path = path.as_ref().display().to_string();
let file = std::fs::File::open(path.as_ref())
.map_err(|e| Error::Other(format!("failed to open fixture {display_path}: {e}")))?;
let lines = std::io::BufReader::new(file).lines();
let mut content = String::new();
for line in lines {
let line =
line.map_err(|e| Error::Other(format!("failed to read fixture {display_path}: {e}")))?;
content.push_str(&line);
content.push_str("\n\n");
}
let rdr = std::io::Cursor::new(content);
let stream = ReaderStream::new(rdr).map_err(|e| Error::Other(e.to_string()));
tokio::spawn(process_sse(
stream,
tx_event,
provider.stream_idle_timeout(),
otel_event_manager,
));
Ok(ResponseStream { rx_event })
}

View File

@@ -0,0 +1,83 @@
use std::pin::Pin;
use std::task::Context;
use std::task::Poll;
use codex_protocol::config_types::ReasoningEffort as ReasoningEffortConfig;
use codex_protocol::config_types::ReasoningSummary as ReasoningSummaryConfig;
use codex_protocol::models::ResponseItem;
use codex_protocol::protocol::RateLimitSnapshot;
use codex_protocol::protocol::TokenUsage;
use futures::Stream;
use serde::Serialize;
use serde_json::Value;
use tokio::sync::mpsc;
use crate::error::Result;
#[derive(Debug, Serialize, Clone)]
pub struct Reasoning {
#[serde(skip_serializing_if = "Option::is_none")]
pub effort: Option<ReasoningEffortConfig>,
#[serde(skip_serializing_if = "Option::is_none")]
pub summary: Option<ReasoningSummaryConfig>,
}
#[derive(Debug, Serialize, Default, Clone)]
#[serde(rename_all = "snake_case")]
pub enum TextFormatType {
#[default]
JsonSchema,
}
#[derive(Debug, Serialize, Default, Clone)]
pub struct TextFormat {
pub r#type: TextFormatType,
pub strict: bool,
pub schema: Value,
pub name: String,
}
#[derive(Debug, Serialize, Default, Clone)]
pub struct TextControls {
#[serde(skip_serializing_if = "Option::is_none")]
pub verbosity: Option<String>,
#[serde(skip_serializing_if = "Option::is_none")]
pub format: Option<TextFormat>,
}
#[derive(Debug)]
pub enum ResponseEvent {
Created,
OutputItemDone(ResponseItem),
OutputItemAdded(ResponseItem),
Completed {
response_id: String,
token_usage: Option<TokenUsage>,
},
OutputTextDelta(String),
ReasoningSummaryDelta(String),
ReasoningContentDelta(String),
ReasoningSummaryPartAdded,
RateLimits(RateLimitSnapshot),
}
#[derive(Debug)]
pub struct EventStream<T> {
pub(crate) rx_event: mpsc::Receiver<T>,
}
impl<T> EventStream<T> {
pub fn from_receiver(rx_event: mpsc::Receiver<T>) -> Self {
Self { rx_event }
}
}
impl<T> Stream for EventStream<T> {
type Item = T;
fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<Self::Item>> {
self.rx_event.poll_recv(cx)
}
}
pub type ResponseStream = EventStream<Result<ResponseEvent>>;

View File

@@ -534,3 +534,204 @@ fn generate_index_ts(out_dir: &Path) -> Result<PathBuf> {
.with_context(|| format!("Failed to write {}", index_path.display()))?;
Ok(index_path)
}
#[cfg(test)]
mod tests {
use super::*;
use anyhow::Result;
use std::collections::BTreeSet;
use std::fs;
use std::path::PathBuf;
use uuid::Uuid;
#[test]
fn generated_ts_has_no_optional_nullable_fields() -> Result<()> {
let output_dir = std::env::temp_dir().join(format!("codex_ts_types_{}", Uuid::now_v7()));
fs::create_dir(&output_dir)?;
struct TempDirGuard(PathBuf);
impl Drop for TempDirGuard {
fn drop(&mut self) {
let _ = fs::remove_dir_all(&self.0);
}
}
let _guard = TempDirGuard(output_dir.clone());
generate_ts(&output_dir, None)?;
let mut undefined_offenders = Vec::new();
let mut optional_nullable_offenders = BTreeSet::new();
let mut stack = vec![output_dir];
while let Some(dir) = stack.pop() {
for entry in fs::read_dir(&dir)? {
let entry = entry?;
let path = entry.path();
if path.is_dir() {
stack.push(path);
continue;
}
if matches!(path.extension().and_then(|ext| ext.to_str()), Some("ts")) {
let contents = fs::read_to_string(&path)?;
if contents.contains("| undefined") {
undefined_offenders.push(path.clone());
}
const SKIP_PREFIXES: &[&str] = &[
"const ",
"let ",
"var ",
"export const ",
"export let ",
"export var ",
];
let mut search_start = 0;
while let Some(idx) = contents[search_start..].find("| null") {
let abs_idx = search_start + idx;
// Find the property-colon for this field by scanning forward
// from the start of the segment and ignoring nested braces,
// brackets, and parens. This avoids colons inside nested
// type literals like `{ [k in string]?: string }`.
let line_start_idx =
contents[..abs_idx].rfind('\n').map(|i| i + 1).unwrap_or(0);
let mut segment_start_idx = line_start_idx;
if let Some(rel_idx) = contents[line_start_idx..abs_idx].rfind(',') {
segment_start_idx = segment_start_idx.max(line_start_idx + rel_idx + 1);
}
if let Some(rel_idx) = contents[line_start_idx..abs_idx].rfind('{') {
segment_start_idx = segment_start_idx.max(line_start_idx + rel_idx + 1);
}
if let Some(rel_idx) = contents[line_start_idx..abs_idx].rfind('}') {
segment_start_idx = segment_start_idx.max(line_start_idx + rel_idx + 1);
}
// Scan forward for the colon that separates the field name from its type.
let mut level_brace = 0_i32;
let mut level_brack = 0_i32;
let mut level_paren = 0_i32;
let mut in_single = false;
let mut in_double = false;
let mut escape = false;
let mut prop_colon_idx = None;
for (i, ch) in contents[segment_start_idx..abs_idx].char_indices() {
let idx_abs = segment_start_idx + i;
if escape {
escape = false;
continue;
}
match ch {
'\\' => {
// Only treat as escape when inside a string.
if in_single || in_double {
escape = true;
}
}
'\'' => {
if !in_double {
in_single = !in_single;
}
}
'"' => {
if !in_single {
in_double = !in_double;
}
}
'{' if !in_single && !in_double => level_brace += 1,
'}' if !in_single && !in_double => level_brace -= 1,
'[' if !in_single && !in_double => level_brack += 1,
']' if !in_single && !in_double => level_brack -= 1,
'(' if !in_single && !in_double => level_paren += 1,
')' if !in_single && !in_double => level_paren -= 1,
':' if !in_single
&& !in_double
&& level_brace == 0
&& level_brack == 0
&& level_paren == 0 =>
{
prop_colon_idx = Some(idx_abs);
break;
}
_ => {}
}
}
let Some(colon_idx) = prop_colon_idx else {
search_start = abs_idx + 5;
continue;
};
let mut field_prefix = contents[segment_start_idx..colon_idx].trim();
if field_prefix.is_empty() {
search_start = abs_idx + 5;
continue;
}
if let Some(comment_idx) = field_prefix.rfind("*/") {
field_prefix = field_prefix[comment_idx + 2..].trim_start();
}
if field_prefix.is_empty() {
search_start = abs_idx + 5;
continue;
}
if SKIP_PREFIXES
.iter()
.any(|prefix| field_prefix.starts_with(prefix))
{
search_start = abs_idx + 5;
continue;
}
if field_prefix.contains('(') {
search_start = abs_idx + 5;
continue;
}
// If the last non-whitespace before ':' is '?', then this is an
// optional field with a nullable type (i.e., "?: T | null"),
// which we explicitly disallow.
if field_prefix.chars().rev().find(|c| !c.is_whitespace()) == Some('?') {
let line_number =
contents[..abs_idx].chars().filter(|c| *c == '\n').count() + 1;
let offending_line_end = contents[line_start_idx..]
.find('\n')
.map(|i| line_start_idx + i)
.unwrap_or(contents.len());
let offending_snippet =
contents[line_start_idx..offending_line_end].trim();
optional_nullable_offenders.insert(format!(
"{}:{}: {offending_snippet}",
path.display(),
line_number
));
}
search_start = abs_idx + 5;
}
}
}
}
assert!(
undefined_offenders.is_empty(),
"Generated TypeScript still includes unions with `undefined` in {undefined_offenders:?}"
);
// If this assertion fails, it means a field was generated as
// "?: T | null" — i.e., both optional (undefined) and nullable (null).
// We only want either "?: T" or ": T | null".
assert!(
optional_nullable_offenders.is_empty(),
"Generated TypeScript has optional fields with nullable types (disallowed '?: T | null'), add #[ts(optional)] to fix:\n{optional_nullable_offenders:?}"
);
Ok(())
}
}

View File

@@ -34,6 +34,7 @@ pub struct JSONRPCRequest {
pub id: RequestId,
pub method: String,
#[serde(default, skip_serializing_if = "Option::is_none")]
#[ts(optional)]
pub params: Option<serde_json::Value>,
}
@@ -42,6 +43,7 @@ pub struct JSONRPCRequest {
pub struct JSONRPCNotification {
pub method: String,
#[serde(default, skip_serializing_if = "Option::is_none")]
#[ts(optional)]
pub params: Option<serde_json::Value>,
}
@@ -63,6 +65,7 @@ pub struct JSONRPCError {
pub struct JSONRPCErrorError {
pub code: i64,
#[serde(default, skip_serializing_if = "Option::is_none")]
#[ts(optional)]
pub data: Option<serde_json::Value>,
pub message: String,
}

View File

@@ -269,7 +269,6 @@ pub struct InitializeParams {
#[serde(rename_all = "camelCase")]
pub struct ClientInfo {
pub name: String,
#[serde(skip_serializing_if = "Option::is_none")]
pub title: Option<String>,
pub version: String,
}
@@ -284,42 +283,41 @@ pub struct InitializeResponse {
#[serde(rename_all = "camelCase")]
pub struct NewConversationParams {
/// Optional override for the model name (e.g. "o3", "o4-mini").
#[serde(skip_serializing_if = "Option::is_none")]
pub model: Option<String>,
/// Override the model provider to use for this session.
#[serde(skip_serializing_if = "Option::is_none")]
pub model_provider: Option<String>,
/// Configuration profile from config.toml to specify default options.
#[serde(skip_serializing_if = "Option::is_none")]
pub profile: Option<String>,
/// Working directory for the session. If relative, it is resolved against
/// the server process's current working directory.
#[serde(skip_serializing_if = "Option::is_none")]
pub cwd: Option<String>,
/// Approval policy for shell commands generated by the model:
/// `untrusted`, `on-failure`, `on-request`, `never`.
#[serde(skip_serializing_if = "Option::is_none")]
pub approval_policy: Option<AskForApproval>,
/// Sandbox mode: `read-only`, `workspace-write`, or `danger-full-access`.
#[serde(skip_serializing_if = "Option::is_none")]
pub sandbox: Option<SandboxMode>,
/// Individual config settings that will override what is in
/// CODEX_HOME/config.toml.
#[serde(skip_serializing_if = "Option::is_none")]
pub config: Option<HashMap<String, serde_json::Value>>,
/// The set of instructions to use instead of the default ones.
#[serde(skip_serializing_if = "Option::is_none")]
pub base_instructions: Option<String>,
/// Whether to include the apply patch tool in the conversation.
/// Developer instructions that will be sent as a `developer` role message.
#[serde(skip_serializing_if = "Option::is_none")]
pub developer_instructions: Option<String>,
/// Prompt used during conversation compaction.
#[serde(skip_serializing_if = "Option::is_none")]
pub compact_prompt: Option<String>,
/// Whether to include the apply patch tool in the conversation.
pub include_apply_patch_tool: Option<bool>,
}
@@ -329,7 +327,6 @@ pub struct NewConversationResponse {
pub conversation_id: ConversationId,
pub model: String,
/// Note this could be ignored by the model.
#[serde(skip_serializing_if = "Option::is_none")]
pub reasoning_effort: Option<ReasoningEffort>,
pub rollout_path: PathBuf,
}
@@ -339,7 +336,6 @@ pub struct NewConversationResponse {
pub struct ResumeConversationResponse {
pub conversation_id: ConversationId,
pub model: String,
#[serde(skip_serializing_if = "Option::is_none")]
pub initial_messages: Option<Vec<EventMsg>>,
pub rollout_path: PathBuf,
}
@@ -371,16 +367,13 @@ pub struct GetConversationSummaryResponse {
#[serde(rename_all = "camelCase")]
pub struct ListConversationsParams {
/// Optional page size; defaults to a reasonable server-side value.
#[serde(skip_serializing_if = "Option::is_none")]
pub page_size: Option<usize>,
/// Opaque pagination cursor returned by a previous call.
#[serde(skip_serializing_if = "Option::is_none")]
pub cursor: Option<String>,
/// Optional model provider filter (matches against session metadata).
/// - None => filter by the server's default model provider
/// - Some([]) => no filtering, include all providers
/// - Some([...]) => only include sessions with one of the specified providers
#[serde(skip_serializing_if = "Option::is_none")]
pub model_providers: Option<Vec<String>>,
}
@@ -391,7 +384,6 @@ pub struct ConversationSummary {
pub path: PathBuf,
pub preview: String,
/// RFC3339 timestamp string for the session start, if available.
#[serde(skip_serializing_if = "Option::is_none")]
pub timestamp: Option<String>,
/// Model provider recorded for the session (resolved when absent in metadata).
pub model_provider: String,
@@ -403,7 +395,6 @@ pub struct ListConversationsResponse {
pub items: Vec<ConversationSummary>,
/// Opaque cursor to pass to the next call to continue after the last item.
/// if None, there are no more items to return.
#[serde(skip_serializing_if = "Option::is_none")]
pub next_cursor: Option<String>,
}
@@ -411,10 +402,8 @@ pub struct ListConversationsResponse {
#[serde(rename_all = "camelCase")]
pub struct ListModelsParams {
/// Optional page size; defaults to a reasonable server-side value.
#[serde(skip_serializing_if = "Option::is_none")]
pub page_size: Option<usize>,
/// Opaque pagination cursor returned by a previous call.
#[serde(skip_serializing_if = "Option::is_none")]
pub cursor: Option<String>,
}
@@ -444,7 +433,6 @@ pub struct ListModelsResponse {
pub items: Vec<Model>,
/// Opaque cursor to pass to the next call to continue after the last item.
/// if None, there are no more items to return.
#[serde(skip_serializing_if = "Option::is_none")]
pub next_cursor: Option<String>,
}
@@ -452,9 +440,7 @@ pub struct ListModelsResponse {
#[serde(rename_all = "camelCase")]
pub struct UploadFeedbackParams {
pub classification: String,
#[serde(skip_serializing_if = "Option::is_none")]
pub reason: Option<String>,
#[serde(skip_serializing_if = "Option::is_none")]
pub conversation_id: Option<ConversationId>,
pub include_logs: bool,
}
@@ -501,16 +487,12 @@ pub struct LogoutAccountResponse {}
#[serde(rename_all = "camelCase")]
pub struct ResumeConversationParams {
/// Absolute path to the rollout JSONL file, when explicitly resuming a known rollout.
#[serde(skip_serializing_if = "Option::is_none")]
pub path: Option<PathBuf>,
/// If the rollout path is not known, it can be discovered via the conversation id at the cost of extra latency.
#[serde(skip_serializing_if = "Option::is_none")]
pub conversation_id: Option<ConversationId>,
/// if the rollout path or conversation id is not known, it can be resumed from given history
#[serde(skip_serializing_if = "Option::is_none")]
pub history: Option<Vec<ResponseItem>>,
/// Optional overrides to apply when spawning the resumed session.
#[serde(skip_serializing_if = "Option::is_none")]
pub overrides: Option<NewConversationParams>,
}
@@ -592,10 +574,8 @@ pub struct LogoutChatGptResponse {}
#[serde(rename_all = "camelCase")]
pub struct GetAuthStatusParams {
/// If true, include the current auth token (if available) in the response.
#[serde(skip_serializing_if = "Option::is_none")]
pub include_token: Option<bool>,
/// If true, attempt to refresh the token before returning status.
#[serde(skip_serializing_if = "Option::is_none")]
pub refresh_token: Option<bool>,
}
@@ -608,10 +588,8 @@ pub struct ExecOneOffCommandParams {
/// If not specified, a sensible default is used server-side.
pub timeout_ms: Option<u64>,
/// Optional working directory for the process. Defaults to server config cwd.
#[serde(skip_serializing_if = "Option::is_none")]
pub cwd: Option<PathBuf>,
/// Optional explicit sandbox policy overriding the server default.
#[serde(skip_serializing_if = "Option::is_none")]
pub sandbox_policy: Option<SandboxPolicy>,
}
@@ -632,15 +610,12 @@ pub struct GetAccountRateLimitsResponse {
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
pub struct GetAuthStatusResponse {
#[serde(skip_serializing_if = "Option::is_none")]
pub auth_method: Option<AuthMode>,
#[serde(skip_serializing_if = "Option::is_none")]
pub auth_token: Option<String>,
// Indicates that auth method must be valid to use the server.
// This can be false if using a custom provider that is configured
// with requires_openai_auth == false.
#[serde(skip_serializing_if = "Option::is_none")]
pub requires_openai_auth: Option<bool>,
}
@@ -670,11 +645,9 @@ pub struct GetUserSavedConfigResponse {
#[serde(rename_all = "camelCase")]
pub struct SetDefaultModelParams {
/// If set to None, this means `model` should be cleared in config.toml.
#[serde(skip_serializing_if = "Option::is_none")]
pub model: Option<String>,
/// If set to None, this means `model_reasoning_effort` should be cleared
/// in config.toml.
#[serde(skip_serializing_if = "Option::is_none")]
pub reasoning_effort: Option<ReasoningEffort>,
}
@@ -689,36 +662,24 @@ pub struct SetDefaultModelResponse {}
#[serde(rename_all = "camelCase")]
pub struct UserSavedConfig {
/// Approvals
#[serde(skip_serializing_if = "Option::is_none")]
pub approval_policy: Option<AskForApproval>,
#[serde(skip_serializing_if = "Option::is_none")]
pub sandbox_mode: Option<SandboxMode>,
#[serde(skip_serializing_if = "Option::is_none")]
pub sandbox_settings: Option<SandboxSettings>,
#[serde(skip_serializing_if = "Option::is_none")]
pub forced_chatgpt_workspace_id: Option<String>,
#[serde(skip_serializing_if = "Option::is_none")]
pub forced_login_method: Option<ForcedLoginMethod>,
/// Model-specific configuration
#[serde(skip_serializing_if = "Option::is_none")]
pub model: Option<String>,
#[serde(skip_serializing_if = "Option::is_none")]
pub model_reasoning_effort: Option<ReasoningEffort>,
#[serde(skip_serializing_if = "Option::is_none")]
pub model_reasoning_summary: Option<ReasoningSummary>,
#[serde(skip_serializing_if = "Option::is_none")]
pub model_verbosity: Option<Verbosity>,
/// Tools
#[serde(skip_serializing_if = "Option::is_none")]
pub tools: Option<Tools>,
/// Profiles
#[serde(skip_serializing_if = "Option::is_none")]
pub profile: Option<String>,
#[serde(default)]
pub profiles: HashMap<String, Profile>,
}
@@ -736,27 +697,23 @@ pub struct Profile {
pub model_verbosity: Option<Verbosity>,
pub chatgpt_base_url: Option<String>,
}
/// MCP representation of a [`codex_core::config::ToolsToml`].
#[derive(Deserialize, Debug, Clone, PartialEq, Serialize, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
pub struct Tools {
#[serde(skip_serializing_if = "Option::is_none")]
pub web_search: Option<bool>,
#[serde(skip_serializing_if = "Option::is_none")]
pub view_image: Option<bool>,
}
/// MCP representation of a [`codex_core::config_types::SandboxWorkspaceWrite`].
/// MCP representation of a [`codex_core::config::types::SandboxWorkspaceWrite`].
#[derive(Deserialize, Debug, Clone, PartialEq, Serialize, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
pub struct SandboxSettings {
#[serde(default)]
pub writable_roots: Vec<PathBuf>,
#[serde(skip_serializing_if = "Option::is_none")]
pub network_access: Option<bool>,
#[serde(skip_serializing_if = "Option::is_none")]
pub exclude_tmpdir_env_var: Option<bool>,
#[serde(skip_serializing_if = "Option::is_none")]
pub exclude_slash_tmp: Option<bool>,
}
@@ -776,7 +733,6 @@ pub struct SendUserTurnParams {
pub approval_policy: AskForApproval,
pub sandbox_policy: SandboxPolicy,
pub model: String,
#[serde(skip_serializing_if = "Option::is_none")]
pub effort: Option<ReasoningEffort>,
pub summary: ReasoningSummary,
}
@@ -919,11 +875,9 @@ pub struct ApplyPatchApprovalParams {
pub call_id: String,
pub file_changes: HashMap<PathBuf, FileChange>,
/// Optional explanatory reason (e.g. request for extra write access).
#[serde(skip_serializing_if = "Option::is_none")]
pub reason: Option<String>,
/// When set, the agent is asking the user to allow writes under this root
/// for the remainder of the session (unclear if this is honored today).
#[serde(skip_serializing_if = "Option::is_none")]
pub grant_root: Option<PathBuf>,
}
@@ -936,9 +890,7 @@ pub struct ExecCommandApprovalParams {
pub call_id: String,
pub command: Vec<String>,
pub cwd: PathBuf,
#[serde(skip_serializing_if = "Option::is_none")]
pub reason: Option<String>,
#[serde(skip_serializing_if = "Option::is_none")]
pub risk: Option<SandboxCommandAssessment>,
pub parsed_cmd: Vec<ParsedCommand>,
}
@@ -960,7 +912,6 @@ pub struct FuzzyFileSearchParams {
pub query: String,
pub roots: Vec<String>,
// if provided, will cancel any previous request that used the same value
#[serde(skip_serializing_if = "Option::is_none")]
pub cancellation_token: Option<String>,
}
@@ -971,7 +922,6 @@ pub struct FuzzyFileSearchResult {
pub path: String,
pub file_name: String,
pub score: u32,
#[serde(skip_serializing_if = "Option::is_none")]
pub indices: Option<Vec<u32>>,
}
@@ -986,7 +936,6 @@ pub struct LoginChatGptCompleteNotification {
#[schemars(with = "String")]
pub login_id: Uuid,
pub success: bool,
#[serde(skip_serializing_if = "Option::is_none")]
pub error: Option<String>,
}
@@ -1000,7 +949,6 @@ pub struct SessionConfiguredNotification {
pub model: String,
/// The effort the model is putting into reasoning about the user's request.
#[serde(skip_serializing_if = "Option::is_none")]
pub reasoning_effort: Option<ReasoningEffort>,
/// Identifier of the history log file (inode on Unix, 0 otherwise).
@@ -1012,7 +960,6 @@ pub struct SessionConfiguredNotification {
/// Optional initial messages (as events) for resumed sessions.
/// When present, UIs can use these to seed the history.
#[serde(skip_serializing_if = "Option::is_none")]
pub initial_messages: Option<Vec<EventMsg>>,
pub rollout_path: PathBuf,
@@ -1022,7 +969,6 @@ pub struct SessionConfiguredNotification {
#[serde(rename_all = "camelCase")]
pub struct AuthStatusChangeNotification {
/// Current authentication method; omitted if signed out.
#[serde(skip_serializing_if = "Option::is_none")]
pub auth_method: Option<AuthMode>,
}
@@ -1095,6 +1041,8 @@ mod tests {
sandbox: None,
config: None,
base_instructions: None,
developer_instructions: None,
compact_prompt: None,
include_apply_patch_tool: None,
},
};
@@ -1104,7 +1052,14 @@ mod tests {
"id": 42,
"params": {
"model": "gpt-5-codex",
"approvalPolicy": "on-request"
"modelProvider": null,
"profile": null,
"cwd": null,
"approvalPolicy": "on-request",
"sandbox": null,
"config": null,
"baseInstructions": null,
"includeApplyPatchTool": null
}
}),
serde_json::to_value(&request)?,
@@ -1177,6 +1132,7 @@ mod tests {
"command": ["echo", "hello"],
"cwd": "/tmp",
"reason": "because tests",
"risk": null,
"parsedCmd": [
{
"type": "unknown",
@@ -1321,7 +1277,10 @@ mod tests {
json!({
"method": "model/list",
"id": 6,
"params": {}
"params": {
"pageSize": null,
"cursor": null
}
}),
serde_json::to_value(&request)?,
);

View File

@@ -73,10 +73,8 @@ use codex_core::auth::login_with_api_key;
use codex_core::config::Config;
use codex_core::config::ConfigOverrides;
use codex_core::config::ConfigToml;
use codex_core::config::load_config_as_toml;
use codex_core::config_edit::CONFIG_KEY_EFFORT;
use codex_core::config_edit::CONFIG_KEY_MODEL;
use codex_core::config_edit::persist_overrides_and_clear_if_none;
use codex_core::config::edit::ConfigEditsBuilder;
use codex_core::config_loader::load_config_as_toml;
use codex_core::default_client::get_codex_user_agent;
use codex_core::exec::ExecParams;
use codex_core::exec_env::create_env;
@@ -689,19 +687,12 @@ impl CodexMessageProcessor {
model,
reasoning_effort,
} = params;
let effort_str = reasoning_effort.map(|effort| effort.to_string());
let overrides: [(&[&str], Option<&str>); 2] = [
(&[CONFIG_KEY_MODEL], model.as_deref()),
(&[CONFIG_KEY_EFFORT], effort_str.as_deref()),
];
match persist_overrides_and_clear_if_none(
&self.config.codex_home,
self.config.active_profile.as_deref(),
&overrides,
)
.await
match ConfigEditsBuilder::new(&self.config.codex_home)
.with_profile(self.config.active_profile.as_deref())
.set_model(model.as_deref(), reasoning_effort)
.apply()
.await
{
Ok(()) => {
let response = SetDefaultModelResponse {};
@@ -710,7 +701,7 @@ impl CodexMessageProcessor {
Err(err) => {
let error = JSONRPCErrorError {
code: INTERNAL_ERROR_CODE,
message: format!("failed to persist overrides: {err}"),
message: format!("failed to persist model selection: {err}"),
data: None,
};
self.outgoing.send_error(request_id, error).await;
@@ -1769,6 +1760,8 @@ async fn derive_config_from_params(
sandbox: sandbox_mode,
config: cli_overrides,
base_instructions,
developer_instructions,
compact_prompt,
include_apply_patch_tool,
} = params;
let overrides = ConfigOverrides {
@@ -1781,6 +1774,8 @@ async fn derive_config_from_params(
model_provider,
codex_linux_sandbox_exe,
base_instructions,
developer_instructions,
compact_prompt,
include_apply_patch_tool,
include_view_image_tool: None,
show_raw_agent_reasoning: None,

View File

@@ -166,6 +166,7 @@ mod tests {
"params": {
"loginId": Uuid::nil(),
"success": true,
"error": null,
},
}),
serde_json::to_value(jsonrpc_notification)

View File

@@ -16,6 +16,7 @@ use codex_app_server_protocol::SendUserMessageResponse;
use codex_protocol::ConversationId;
use codex_protocol::models::ContentItem;
use codex_protocol::models::ResponseItem;
use codex_protocol::protocol::RawResponseItemEvent;
use pretty_assertions::assert_eq;
use std::path::Path;
use tempfile::TempDir;
@@ -43,7 +44,9 @@ async fn test_send_message_success() -> Result<()> {
// Start a conversation using the new wire API.
let new_conv_id = mcp
.send_new_conversation_request(NewConversationParams::default())
.send_new_conversation_request(NewConversationParams {
..Default::default()
})
.await?;
let new_conv_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
@@ -142,7 +145,10 @@ async fn test_send_message_raw_notifications_opt_in() -> Result<()> {
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
let new_conv_id = mcp
.send_new_conversation_request(NewConversationParams::default())
.send_new_conversation_request(NewConversationParams {
developer_instructions: Some("Use the test harness tools.".to_string()),
..Default::default()
})
.await?;
let new_conv_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
@@ -176,6 +182,9 @@ async fn test_send_message_raw_notifications_opt_in() -> Result<()> {
})
.await?;
let developer = read_raw_response_item(&mut mcp, conversation_id).await;
assert_developer_message(&developer, "Use the test harness tools.");
let instructions = read_raw_response_item(&mut mcp, conversation_id).await;
assert_instructions_message(&instructions);
@@ -294,7 +303,9 @@ async fn read_raw_response_item(
.cloned()
.expect("raw response item should include msg payload");
serde_json::from_value(msg_value).expect("deserialize raw response item")
let event: RawResponseItemEvent =
serde_json::from_value(msg_value).expect("deserialize raw response item");
event.item
}
fn assert_instructions_message(item: &ResponseItem) {
@@ -313,6 +324,21 @@ fn assert_instructions_message(item: &ResponseItem) {
}
}
fn assert_developer_message(item: &ResponseItem, expected_text: &str) {
match item {
ResponseItem::Message { role, content, .. } => {
assert_eq!(role, "developer");
let texts = content_texts(content);
assert_eq!(
texts,
vec![expected_text],
"expected developer instructions message, got {texts:?}"
);
}
other => panic!("expected developer instructions message, got {other:?}"),
}
}
fn assert_environment_message(item: &ResponseItem) {
match item {
ResponseItem::Message { role, content, .. } => {

View File

@@ -9,11 +9,11 @@ use codex_common::CliConfigOverrides;
use codex_common::format_env_display::format_env_display;
use codex_core::config::Config;
use codex_core::config::ConfigOverrides;
use codex_core::config::edit::ConfigEditsBuilder;
use codex_core::config::find_codex_home;
use codex_core::config::load_global_mcp_servers;
use codex_core::config::write_global_mcp_servers;
use codex_core::config_types::McpServerConfig;
use codex_core::config_types::McpServerTransportConfig;
use codex_core::config::types::McpServerConfig;
use codex_core::config::types::McpServerTransportConfig;
use codex_core::features::Feature;
use codex_core::mcp::auth::compute_auth_statuses;
use codex_core::protocol::McpAuthStatus;
@@ -263,7 +263,10 @@ async fn run_add(config_overrides: &CliConfigOverrides, add_args: AddArgs) -> Re
servers.insert(name.clone(), new_entry);
write_global_mcp_servers(&codex_home, &servers)
ConfigEditsBuilder::new(&codex_home)
.replace_mcp_servers(&servers)
.apply()
.await
.with_context(|| format!("failed to write MCP servers to {}", codex_home.display()))?;
println!("Added global MCP server '{name}'.");
@@ -321,7 +324,10 @@ async fn run_remove(config_overrides: &CliConfigOverrides, remove_args: RemoveAr
let removed = servers.remove(&name).is_some();
if removed {
write_global_mcp_servers(&codex_home, &servers)
ConfigEditsBuilder::new(&codex_home)
.replace_mcp_servers(&servers)
.apply()
.await
.with_context(|| format!("failed to write MCP servers to {}", codex_home.display()))?;
}
@@ -341,9 +347,7 @@ async fn run_login(config_overrides: &CliConfigOverrides, login_args: LoginArgs)
.context("failed to load configuration")?;
if !config.features.enabled(Feature::RmcpClient) {
bail!(
"OAuth login is only supported when experimental_use_rmcp_client is true in config.toml."
);
bail!("OAuth login is only supported when [feature].rmcp_client is true in config.toml.");
}
let LoginArgs { name, scopes } = login_args;

View File

@@ -2,7 +2,7 @@ use std::path::Path;
use anyhow::Result;
use codex_core::config::load_global_mcp_servers;
use codex_core::config_types::McpServerTransportConfig;
use codex_core::config::types::McpServerTransportConfig;
use predicates::str::contains;
use pretty_assertions::assert_eq;
use tempfile::TempDir;

View File

@@ -1,9 +1,9 @@
use std::path::Path;
use anyhow::Result;
use codex_core::config::edit::ConfigEditsBuilder;
use codex_core::config::load_global_mcp_servers;
use codex_core::config::write_global_mcp_servers;
use codex_core::config_types::McpServerTransportConfig;
use codex_core::config::types::McpServerTransportConfig;
use predicates::prelude::PredicateBooleanExt;
use predicates::str::contains;
use pretty_assertions::assert_eq;
@@ -59,7 +59,9 @@ async fn list_and_get_render_expected_output() -> Result<()> {
}
other => panic!("unexpected transport: {other:?}"),
}
write_global_mcp_servers(codex_home.path(), &servers)?;
ConfigEditsBuilder::new(codex_home.path())
.replace_mcp_servers(&servers)
.apply_blocking()?;
let mut list_cmd = codex_command(codex_home.path())?;
let list_output = list_cmd.args(["mcp", "list"]).output()?;
@@ -149,7 +151,9 @@ async fn get_disabled_server_shows_single_line() -> Result<()> {
.get_mut("docs")
.expect("docs server should exist after add");
docs.enabled = false;
write_global_mcp_servers(codex_home.path(), &servers)?;
ConfigEditsBuilder::new(codex_home.path())
.replace_mcp_servers(&servers)
.apply_blocking()?;
let mut get_cmd = codex_command(codex_home.path())?;
let get_output = get_cmd.args(["mcp", "get", "docs"]).output()?;

View File

@@ -16,3 +16,6 @@ path = "src/lib.rs"
serde = { version = "1", features = ["derive"] }
serde_json = "1"
serde_with = "3"
[package.metadata.cargo-shear]
ignored = ["serde_with"]

View File

@@ -8,6 +8,7 @@ workspace = true
[dependencies]
clap = { workspace = true, features = ["derive", "wrap_help"], optional = true }
codex-api-client = { workspace = true }
codex-core = { workspace = true }
codex-protocol = { workspace = true }
codex-app-server-protocol = { workspace = true }

View File

@@ -1,4 +1,4 @@
use codex_core::WireApi;
use codex_api_client::WireApi;
use codex_core::config::Config;
use crate::sandbox_summary::summarize_sandbox_policy;

View File

@@ -22,6 +22,7 @@ chrono = { workspace = true, features = ["serde"] }
codex-app-server-protocol = { workspace = true }
codex-apply-patch = { workspace = true }
codex-async-utils = { workspace = true }
codex-api-client = { workspace = true }
codex-file-search = { workspace = true }
codex-git = { workspace = true }
codex-keyring-store = { workspace = true }

View File

@@ -61,7 +61,13 @@ pub(crate) async fn apply_patch(
// that similar patches can be auto-approved in the future during
// this session.
let rx_approve = sess
.request_patch_approval(turn_context, call_id.to_owned(), &action, None, None)
.request_patch_approval(
turn_context,
call_id.to_owned(),
convert_apply_patch_to_protocol(&action),
None,
None,
)
.await;
match rx_approve.await.unwrap_or_default() {
ReviewDecision::Approved | ReviewDecision::ApprovedForSession => {

View File

@@ -22,7 +22,6 @@ use crate::auth::storage::AuthStorageBackend;
use crate::auth::storage::create_auth_storage;
use crate::config::Config;
use crate::default_client::CodexHttpClient;
use crate::token_data::PlanType;
use crate::token_data::TokenData;
use crate::token_data::parse_id_token;
use crate::util::try_parse_error_message;
@@ -153,11 +152,6 @@ impl CodexAuth {
self.get_current_token_data().and_then(|t| t.id_token.email)
}
pub(crate) fn get_plan_type(&self) -> Option<PlanType> {
self.get_current_token_data()
.and_then(|t| t.id_token.chatgpt_plan_type)
}
fn get_current_auth_json(&self) -> Option<AuthDotJson> {
#[expect(clippy::unwrap_used)]
self.auth_dot_json.lock().unwrap().clone()

View File

@@ -1,946 +0,0 @@
use std::time::Duration;
use crate::ModelProviderInfo;
use crate::client_common::Prompt;
use crate::client_common::ResponseEvent;
use crate::client_common::ResponseStream;
use crate::default_client::CodexHttpClient;
use crate::error::CodexErr;
use crate::error::ConnectionFailedError;
use crate::error::ResponseStreamFailed;
use crate::error::Result;
use crate::error::RetryLimitReachedError;
use crate::error::UnexpectedResponseError;
use crate::model_family::ModelFamily;
use crate::tools::spec::create_tools_json_for_chat_completions_api;
use crate::util::backoff;
use bytes::Bytes;
use codex_otel::otel_event_manager::OtelEventManager;
use codex_protocol::models::ContentItem;
use codex_protocol::models::FunctionCallOutputContentItem;
use codex_protocol::models::ReasoningItemContent;
use codex_protocol::models::ResponseItem;
use eventsource_stream::Eventsource;
use futures::Stream;
use futures::StreamExt;
use futures::TryStreamExt;
use reqwest::StatusCode;
use serde_json::json;
use std::pin::Pin;
use std::task::Context;
use std::task::Poll;
use tokio::sync::mpsc;
use tokio::time::timeout;
use tracing::debug;
use tracing::trace;
/// Implementation for the classic Chat Completions API.
pub(crate) async fn stream_chat_completions(
prompt: &Prompt,
model_family: &ModelFamily,
client: &CodexHttpClient,
provider: &ModelProviderInfo,
otel_event_manager: &OtelEventManager,
) -> Result<ResponseStream> {
if prompt.output_schema.is_some() {
return Err(CodexErr::UnsupportedOperation(
"output_schema is not supported for Chat Completions API".to_string(),
));
}
// Build messages array
let mut messages = Vec::<serde_json::Value>::new();
let full_instructions = prompt.get_full_instructions(model_family);
messages.push(json!({"role": "system", "content": full_instructions}));
let input = prompt.get_formatted_input();
// Pre-scan: map Reasoning blocks to the adjacent assistant anchor after the last user.
// - If the last emitted message is a user message, drop all reasoning.
// - Otherwise, for each Reasoning item after the last user message, attach it
// to the immediate previous assistant message (stop turns) or the immediate
// next assistant anchor (tool-call turns: function/local shell call, or assistant message).
let mut reasoning_by_anchor_index: std::collections::HashMap<usize, String> =
std::collections::HashMap::new();
// Determine the last role that would be emitted to Chat Completions.
let mut last_emitted_role: Option<&str> = None;
for item in &input {
match item {
ResponseItem::Message { role, .. } => last_emitted_role = Some(role.as_str()),
ResponseItem::FunctionCall { .. } | ResponseItem::LocalShellCall { .. } => {
last_emitted_role = Some("assistant")
}
ResponseItem::FunctionCallOutput { .. } => last_emitted_role = Some("tool"),
ResponseItem::Reasoning { .. } | ResponseItem::Other => {}
ResponseItem::CustomToolCall { .. } => {}
ResponseItem::CustomToolCallOutput { .. } => {}
ResponseItem::WebSearchCall { .. } => {}
ResponseItem::GhostSnapshot { .. } => {}
}
}
// Find the last user message index in the input.
let mut last_user_index: Option<usize> = None;
for (idx, item) in input.iter().enumerate() {
if let ResponseItem::Message { role, .. } = item
&& role == "user"
{
last_user_index = Some(idx);
}
}
// Attach reasoning only if the conversation does not end with a user message.
if !matches!(last_emitted_role, Some("user")) {
for (idx, item) in input.iter().enumerate() {
// Only consider reasoning that appears after the last user message.
if let Some(u_idx) = last_user_index
&& idx <= u_idx
{
continue;
}
if let ResponseItem::Reasoning {
content: Some(items),
..
} = item
{
let mut text = String::new();
for entry in items {
match entry {
ReasoningItemContent::ReasoningText { text: segment }
| ReasoningItemContent::Text { text: segment } => text.push_str(segment),
}
}
if text.trim().is_empty() {
continue;
}
// Prefer immediate previous assistant message (stop turns)
let mut attached = false;
if idx > 0
&& let ResponseItem::Message { role, .. } = &input[idx - 1]
&& role == "assistant"
{
reasoning_by_anchor_index
.entry(idx - 1)
.and_modify(|v| v.push_str(&text))
.or_insert(text.clone());
attached = true;
}
// Otherwise, attach to immediate next assistant anchor (tool-calls or assistant message)
if !attached && idx + 1 < input.len() {
match &input[idx + 1] {
ResponseItem::FunctionCall { .. } | ResponseItem::LocalShellCall { .. } => {
reasoning_by_anchor_index
.entry(idx + 1)
.and_modify(|v| v.push_str(&text))
.or_insert(text.clone());
}
ResponseItem::Message { role, .. } if role == "assistant" => {
reasoning_by_anchor_index
.entry(idx + 1)
.and_modify(|v| v.push_str(&text))
.or_insert(text.clone());
}
_ => {}
}
}
}
}
}
// Track last assistant text we emitted to avoid duplicate assistant messages
// in the outbound Chat Completions payload (can happen if a final
// aggregated assistant message was recorded alongside an earlier partial).
let mut last_assistant_text: Option<String> = None;
for (idx, item) in input.iter().enumerate() {
match item {
ResponseItem::Message { role, content, .. } => {
// Build content either as a plain string (typical for assistant text)
// or as an array of content items when images are present (user/tool multimodal).
let mut text = String::new();
let mut items: Vec<serde_json::Value> = Vec::new();
let mut saw_image = false;
for c in content {
match c {
ContentItem::InputText { text: t }
| ContentItem::OutputText { text: t } => {
text.push_str(t);
items.push(json!({"type":"text","text": t}));
}
ContentItem::InputImage { image_url } => {
saw_image = true;
items.push(json!({"type":"image_url","image_url": {"url": image_url}}));
}
}
}
// Skip exact-duplicate assistant messages.
if role == "assistant" {
if let Some(prev) = &last_assistant_text
&& prev == &text
{
continue;
}
last_assistant_text = Some(text.clone());
}
// For assistant messages, always send a plain string for compatibility.
// For user messages, if an image is present, send an array of content items.
let content_value = if role == "assistant" {
json!(text)
} else if saw_image {
json!(items)
} else {
json!(text)
};
let mut msg = json!({"role": role, "content": content_value});
if role == "assistant"
&& let Some(reasoning) = reasoning_by_anchor_index.get(&idx)
&& let Some(obj) = msg.as_object_mut()
{
obj.insert("reasoning".to_string(), json!(reasoning));
}
messages.push(msg);
}
ResponseItem::FunctionCall {
name,
arguments,
call_id,
..
} => {
let mut msg = json!({
"role": "assistant",
"content": null,
"tool_calls": [{
"id": call_id,
"type": "function",
"function": {
"name": name,
"arguments": arguments,
}
}]
});
if let Some(reasoning) = reasoning_by_anchor_index.get(&idx)
&& let Some(obj) = msg.as_object_mut()
{
obj.insert("reasoning".to_string(), json!(reasoning));
}
messages.push(msg);
}
ResponseItem::LocalShellCall {
id,
call_id: _,
status,
action,
} => {
// Confirm with API team.
let mut msg = json!({
"role": "assistant",
"content": null,
"tool_calls": [{
"id": id.clone().unwrap_or_else(|| "".to_string()),
"type": "local_shell_call",
"status": status,
"action": action,
}]
});
if let Some(reasoning) = reasoning_by_anchor_index.get(&idx)
&& let Some(obj) = msg.as_object_mut()
{
obj.insert("reasoning".to_string(), json!(reasoning));
}
messages.push(msg);
}
ResponseItem::FunctionCallOutput { call_id, output } => {
// Prefer structured content items when available (e.g., images)
// otherwise fall back to the legacy plain-string content.
let content_value = if let Some(items) = &output.content_items {
let mapped: Vec<serde_json::Value> = items
.iter()
.map(|it| match it {
FunctionCallOutputContentItem::InputText { text } => {
json!({"type":"text","text": text})
}
FunctionCallOutputContentItem::InputImage { image_url } => {
json!({"type":"image_url","image_url": {"url": image_url}})
}
})
.collect();
json!(mapped)
} else {
json!(output.content)
};
messages.push(json!({
"role": "tool",
"tool_call_id": call_id,
"content": content_value,
}));
}
ResponseItem::CustomToolCall {
id,
call_id: _,
name,
input,
status: _,
} => {
messages.push(json!({
"role": "assistant",
"content": null,
"tool_calls": [{
"id": id,
"type": "custom",
"custom": {
"name": name,
"input": input,
}
}]
}));
}
ResponseItem::CustomToolCallOutput { call_id, output } => {
messages.push(json!({
"role": "tool",
"tool_call_id": call_id,
"content": output,
}));
}
ResponseItem::GhostSnapshot { .. } => {
// Ghost snapshots annotate history but are not sent to the model.
continue;
}
ResponseItem::Reasoning { .. }
| ResponseItem::WebSearchCall { .. }
| ResponseItem::Other => {
// Omit these items from the conversation history.
continue;
}
}
}
let tools_json = create_tools_json_for_chat_completions_api(&prompt.tools)?;
let payload = json!({
"model": model_family.slug,
"messages": messages,
"stream": true,
"tools": tools_json,
});
debug!(
"POST to {}: {}",
provider.get_full_url(&None),
serde_json::to_string_pretty(&payload).unwrap_or_default()
);
let mut attempt = 0;
let max_retries = provider.request_max_retries();
loop {
attempt += 1;
let req_builder = provider.create_request_builder(client, &None).await?;
let res = otel_event_manager
.log_request(attempt, || {
req_builder
.header(reqwest::header::ACCEPT, "text/event-stream")
.json(&payload)
.send()
})
.await;
match res {
Ok(resp) if resp.status().is_success() => {
let (tx_event, rx_event) = mpsc::channel::<Result<ResponseEvent>>(1600);
let stream = resp.bytes_stream().map_err(|e| {
CodexErr::ResponseStreamFailed(ResponseStreamFailed {
source: e,
request_id: None,
})
});
tokio::spawn(process_chat_sse(
stream,
tx_event,
provider.stream_idle_timeout(),
otel_event_manager.clone(),
));
return Ok(ResponseStream { rx_event });
}
Ok(res) => {
let status = res.status();
if !(status == StatusCode::TOO_MANY_REQUESTS || status.is_server_error()) {
let body = (res.text().await).unwrap_or_default();
return Err(CodexErr::UnexpectedStatus(UnexpectedResponseError {
status,
body,
request_id: None,
}));
}
if attempt > max_retries {
return Err(CodexErr::RetryLimit(RetryLimitReachedError {
status,
request_id: None,
}));
}
let retry_after_secs = res
.headers()
.get(reqwest::header::RETRY_AFTER)
.and_then(|v| v.to_str().ok())
.and_then(|s| s.parse::<u64>().ok());
let delay = retry_after_secs
.map(|s| Duration::from_millis(s * 1_000))
.unwrap_or_else(|| backoff(attempt));
tokio::time::sleep(delay).await;
}
Err(e) => {
if attempt > max_retries {
return Err(CodexErr::ConnectionFailed(ConnectionFailedError {
source: e,
}));
}
let delay = backoff(attempt);
tokio::time::sleep(delay).await;
}
}
}
}
/// Lightweight SSE processor for the Chat Completions streaming format. The
/// output is mapped onto Codex's internal [`ResponseEvent`] so that the rest
/// of the pipeline can stay agnostic of the underlying wire format.
async fn process_chat_sse<S>(
stream: S,
tx_event: mpsc::Sender<Result<ResponseEvent>>,
idle_timeout: Duration,
otel_event_manager: OtelEventManager,
) where
S: Stream<Item = Result<Bytes>> + Unpin,
{
let mut stream = stream.eventsource();
// State to accumulate a function call across streaming chunks.
// OpenAI may split the `arguments` string over multiple `delta` events
// until the chunk whose `finish_reason` is `tool_calls` is emitted. We
// keep collecting the pieces here and forward a single
// `ResponseItem::FunctionCall` once the call is complete.
#[derive(Default)]
struct FunctionCallState {
name: Option<String>,
arguments: String,
call_id: Option<String>,
active: bool,
}
let mut fn_call_state = FunctionCallState::default();
let mut assistant_text = String::new();
let mut reasoning_text = String::new();
loop {
let start = std::time::Instant::now();
let response = timeout(idle_timeout, stream.next()).await;
let duration = start.elapsed();
otel_event_manager.log_sse_event(&response, duration);
let sse = match response {
Ok(Some(Ok(ev))) => ev,
Ok(Some(Err(e))) => {
let _ = tx_event
.send(Err(CodexErr::Stream(e.to_string(), None)))
.await;
return;
}
Ok(None) => {
// Stream closed gracefully emit Completed with dummy id.
let _ = tx_event
.send(Ok(ResponseEvent::Completed {
response_id: String::new(),
token_usage: None,
}))
.await;
return;
}
Err(_) => {
let _ = tx_event
.send(Err(CodexErr::Stream(
"idle timeout waiting for SSE".into(),
None,
)))
.await;
return;
}
};
// OpenAI Chat streaming sends a literal string "[DONE]" when finished.
if sse.data.trim() == "[DONE]" {
// Emit any finalized items before closing so downstream consumers receive
// terminal events for both assistant content and raw reasoning.
if !assistant_text.is_empty() {
let item = ResponseItem::Message {
role: "assistant".to_string(),
content: vec![ContentItem::OutputText {
text: std::mem::take(&mut assistant_text),
}],
id: None,
};
let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
}
if !reasoning_text.is_empty() {
let item = ResponseItem::Reasoning {
id: String::new(),
summary: Vec::new(),
content: Some(vec![ReasoningItemContent::ReasoningText {
text: std::mem::take(&mut reasoning_text),
}]),
encrypted_content: None,
};
let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
}
let _ = tx_event
.send(Ok(ResponseEvent::Completed {
response_id: String::new(),
token_usage: None,
}))
.await;
return;
}
// Parse JSON chunk
let chunk: serde_json::Value = match serde_json::from_str(&sse.data) {
Ok(v) => v,
Err(_) => continue,
};
trace!("chat_completions received SSE chunk: {chunk:?}");
let choice_opt = chunk.get("choices").and_then(|c| c.get(0));
if let Some(choice) = choice_opt {
// Handle assistant content tokens as streaming deltas.
if let Some(content) = choice
.get("delta")
.and_then(|d| d.get("content"))
.and_then(|c| c.as_str())
&& !content.is_empty()
{
assistant_text.push_str(content);
let _ = tx_event
.send(Ok(ResponseEvent::OutputTextDelta(content.to_string())))
.await;
}
// Forward any reasoning/thinking deltas if present.
// Some providers stream `reasoning` as a plain string while others
// nest the text under an object (e.g. `{ "reasoning": { "text": "…" } }`).
if let Some(reasoning_val) = choice.get("delta").and_then(|d| d.get("reasoning")) {
let mut maybe_text = reasoning_val
.as_str()
.map(str::to_string)
.filter(|s| !s.is_empty());
if maybe_text.is_none() && reasoning_val.is_object() {
if let Some(s) = reasoning_val
.get("text")
.and_then(|t| t.as_str())
.filter(|s| !s.is_empty())
{
maybe_text = Some(s.to_string());
} else if let Some(s) = reasoning_val
.get("content")
.and_then(|t| t.as_str())
.filter(|s| !s.is_empty())
{
maybe_text = Some(s.to_string());
}
}
if let Some(reasoning) = maybe_text {
// Accumulate so we can emit a terminal Reasoning item at the end.
reasoning_text.push_str(&reasoning);
let _ = tx_event
.send(Ok(ResponseEvent::ReasoningContentDelta(reasoning)))
.await;
}
}
// Some providers only include reasoning on the final message object.
if let Some(message_reasoning) = choice.get("message").and_then(|m| m.get("reasoning"))
{
// Accept either a plain string or an object with { text | content }
if let Some(s) = message_reasoning.as_str() {
if !s.is_empty() {
reasoning_text.push_str(s);
let _ = tx_event
.send(Ok(ResponseEvent::ReasoningContentDelta(s.to_string())))
.await;
}
} else if let Some(obj) = message_reasoning.as_object()
&& let Some(s) = obj
.get("text")
.and_then(|v| v.as_str())
.or_else(|| obj.get("content").and_then(|v| v.as_str()))
&& !s.is_empty()
{
reasoning_text.push_str(s);
let _ = tx_event
.send(Ok(ResponseEvent::ReasoningContentDelta(s.to_string())))
.await;
}
}
// Handle streaming function / tool calls.
if let Some(tool_calls) = choice
.get("delta")
.and_then(|d| d.get("tool_calls"))
.and_then(|tc| tc.as_array())
&& let Some(tool_call) = tool_calls.first()
{
// Mark that we have an active function call in progress.
fn_call_state.active = true;
// Extract call_id if present.
if let Some(id) = tool_call.get("id").and_then(|v| v.as_str()) {
fn_call_state.call_id.get_or_insert_with(|| id.to_string());
}
// Extract function details if present.
if let Some(function) = tool_call.get("function") {
if let Some(name) = function.get("name").and_then(|n| n.as_str()) {
fn_call_state.name.get_or_insert_with(|| name.to_string());
}
if let Some(args_fragment) = function.get("arguments").and_then(|a| a.as_str())
{
fn_call_state.arguments.push_str(args_fragment);
}
}
}
// Emit end-of-turn when finish_reason signals completion.
if let Some(finish_reason) = choice.get("finish_reason").and_then(|v| v.as_str()) {
match finish_reason {
"tool_calls" if fn_call_state.active => {
// First, flush the terminal raw reasoning so UIs can finalize
// the reasoning stream before any exec/tool events begin.
if !reasoning_text.is_empty() {
let item = ResponseItem::Reasoning {
id: String::new(),
summary: Vec::new(),
content: Some(vec![ReasoningItemContent::ReasoningText {
text: std::mem::take(&mut reasoning_text),
}]),
encrypted_content: None,
};
let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
}
// Then emit the FunctionCall response item.
let item = ResponseItem::FunctionCall {
id: None,
name: fn_call_state.name.clone().unwrap_or_else(|| "".to_string()),
arguments: fn_call_state.arguments.clone(),
call_id: fn_call_state.call_id.clone().unwrap_or_else(String::new),
};
let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
}
"stop" => {
// Regular turn without tool-call. Emit the final assistant message
// as a single OutputItemDone so non-delta consumers see the result.
if !assistant_text.is_empty() {
let item = ResponseItem::Message {
role: "assistant".to_string(),
content: vec![ContentItem::OutputText {
text: std::mem::take(&mut assistant_text),
}],
id: None,
};
let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
}
// Also emit a terminal Reasoning item so UIs can finalize raw reasoning.
if !reasoning_text.is_empty() {
let item = ResponseItem::Reasoning {
id: String::new(),
summary: Vec::new(),
content: Some(vec![ReasoningItemContent::ReasoningText {
text: std::mem::take(&mut reasoning_text),
}]),
encrypted_content: None,
};
let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
}
}
_ => {}
}
// Emit Completed regardless of reason so the agent can advance.
let _ = tx_event
.send(Ok(ResponseEvent::Completed {
response_id: String::new(),
token_usage: None,
}))
.await;
// Prepare for potential next turn (should not happen in same stream).
// fn_call_state = FunctionCallState::default();
return; // End processing for this SSE stream.
}
}
}
}
/// Optional client-side aggregation helper
///
/// Stream adapter that merges the incremental `OutputItemDone` chunks coming from
/// [`process_chat_sse`] into a *running* assistant message, **suppressing the
/// per-token deltas**. The stream stays silent while the model is thinking
/// and only emits two events per turn:
///
/// 1. `ResponseEvent::OutputItemDone` with the *complete* assistant message
/// (fully concatenated).
/// 2. The original `ResponseEvent::Completed` right after it.
///
/// This mirrors the behaviour the TypeScript CLI exposes to its higher layers.
///
/// The adapter is intentionally *lossless*: callers who do **not** opt in via
/// [`AggregateStreamExt::aggregate()`] keep receiving the original unmodified
/// events.
#[derive(Copy, Clone, Eq, PartialEq)]
enum AggregateMode {
AggregatedOnly,
Streaming,
}
pub(crate) struct AggregatedChatStream<S> {
inner: S,
cumulative: String,
cumulative_reasoning: String,
pending: std::collections::VecDeque<ResponseEvent>,
mode: AggregateMode,
}
impl<S> Stream for AggregatedChatStream<S>
where
S: Stream<Item = Result<ResponseEvent>> + Unpin,
{
type Item = Result<ResponseEvent>;
fn poll_next(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<Self::Item>> {
let this = self.get_mut();
// First, flush any buffered events from the previous call.
if let Some(ev) = this.pending.pop_front() {
return Poll::Ready(Some(Ok(ev)));
}
loop {
match Pin::new(&mut this.inner).poll_next(cx) {
Poll::Pending => return Poll::Pending,
Poll::Ready(None) => return Poll::Ready(None),
Poll::Ready(Some(Err(e))) => return Poll::Ready(Some(Err(e))),
Poll::Ready(Some(Ok(ResponseEvent::OutputItemDone(item)))) => {
// If this is an incremental assistant message chunk, accumulate but
// do NOT emit yet. Forward any other item (e.g. FunctionCall) right
// away so downstream consumers see it.
let is_assistant_message = matches!(
&item,
codex_protocol::models::ResponseItem::Message { role, .. } if role == "assistant"
);
if is_assistant_message {
match this.mode {
AggregateMode::AggregatedOnly => {
// Only use the final assistant message if we have not
// seen any deltas; otherwise, deltas already built the
// cumulative text and this would duplicate it.
if this.cumulative.is_empty()
&& let codex_protocol::models::ResponseItem::Message {
content,
..
} = &item
&& let Some(text) = content.iter().find_map(|c| match c {
codex_protocol::models::ContentItem::OutputText {
text,
} => Some(text),
_ => None,
})
{
this.cumulative.push_str(text);
}
// Swallow assistant message here; emit on Completed.
continue;
}
AggregateMode::Streaming => {
// In streaming mode, if we have not seen any deltas, forward
// the final assistant message directly. If deltas were seen,
// suppress the final message to avoid duplication.
if this.cumulative.is_empty() {
return Poll::Ready(Some(Ok(ResponseEvent::OutputItemDone(
item,
))));
} else {
continue;
}
}
}
}
// Not an assistant message forward immediately.
return Poll::Ready(Some(Ok(ResponseEvent::OutputItemDone(item))));
}
Poll::Ready(Some(Ok(ResponseEvent::RateLimits(snapshot)))) => {
return Poll::Ready(Some(Ok(ResponseEvent::RateLimits(snapshot))));
}
Poll::Ready(Some(Ok(ResponseEvent::Completed {
response_id,
token_usage,
}))) => {
// Build any aggregated items in the correct order: Reasoning first, then Message.
let mut emitted_any = false;
if !this.cumulative_reasoning.is_empty()
&& matches!(this.mode, AggregateMode::AggregatedOnly)
{
let aggregated_reasoning =
codex_protocol::models::ResponseItem::Reasoning {
id: String::new(),
summary: Vec::new(),
content: Some(vec![
codex_protocol::models::ReasoningItemContent::ReasoningText {
text: std::mem::take(&mut this.cumulative_reasoning),
},
]),
encrypted_content: None,
};
this.pending
.push_back(ResponseEvent::OutputItemDone(aggregated_reasoning));
emitted_any = true;
}
// Always emit the final aggregated assistant message when any
// content deltas have been observed. In AggregatedOnly mode this
// is the sole assistant output; in Streaming mode this finalizes
// the streamed deltas into a terminal OutputItemDone so callers
// can persist/render the message once per turn.
if !this.cumulative.is_empty() {
let aggregated_message = codex_protocol::models::ResponseItem::Message {
id: None,
role: "assistant".to_string(),
content: vec![codex_protocol::models::ContentItem::OutputText {
text: std::mem::take(&mut this.cumulative),
}],
};
this.pending
.push_back(ResponseEvent::OutputItemDone(aggregated_message));
emitted_any = true;
}
// Always emit Completed last when anything was aggregated.
if emitted_any {
this.pending.push_back(ResponseEvent::Completed {
response_id: response_id.clone(),
token_usage: token_usage.clone(),
});
// Return the first pending event now.
if let Some(ev) = this.pending.pop_front() {
return Poll::Ready(Some(Ok(ev)));
}
}
// Nothing aggregated forward Completed directly.
return Poll::Ready(Some(Ok(ResponseEvent::Completed {
response_id,
token_usage,
})));
}
Poll::Ready(Some(Ok(ResponseEvent::Created))) => {
// These events are exclusive to the Responses API and
// will never appear in a Chat Completions stream.
continue;
}
Poll::Ready(Some(Ok(ResponseEvent::OutputTextDelta(delta)))) => {
// Always accumulate deltas so we can emit a final OutputItemDone at Completed.
this.cumulative.push_str(&delta);
if matches!(this.mode, AggregateMode::Streaming) {
// In streaming mode, also forward the delta immediately.
return Poll::Ready(Some(Ok(ResponseEvent::OutputTextDelta(delta))));
} else {
continue;
}
}
Poll::Ready(Some(Ok(ResponseEvent::ReasoningContentDelta(delta)))) => {
// Always accumulate reasoning deltas so we can emit a final Reasoning item at Completed.
this.cumulative_reasoning.push_str(&delta);
if matches!(this.mode, AggregateMode::Streaming) {
// In streaming mode, also forward the delta immediately.
return Poll::Ready(Some(Ok(ResponseEvent::ReasoningContentDelta(delta))));
} else {
continue;
}
}
Poll::Ready(Some(Ok(ResponseEvent::ReasoningSummaryDelta(_)))) => {
continue;
}
Poll::Ready(Some(Ok(ResponseEvent::ReasoningSummaryPartAdded))) => {
continue;
}
Poll::Ready(Some(Ok(ResponseEvent::WebSearchCallBegin { call_id }))) => {
return Poll::Ready(Some(Ok(ResponseEvent::WebSearchCallBegin { call_id })));
}
}
}
}
}
/// Extension trait that activates aggregation on any stream of [`ResponseEvent`].
pub(crate) trait AggregateStreamExt: Stream<Item = Result<ResponseEvent>> + Sized {
/// Returns a new stream that emits **only** the final assistant message
/// per turn instead of every incremental delta. The produced
/// `ResponseEvent` sequence for a typical text turn looks like:
///
/// ```ignore
/// OutputItemDone(<full message>)
/// Completed
/// ```
///
/// No other `OutputItemDone` events will be seen by the caller.
///
/// Usage:
///
/// ```ignore
/// let agg_stream = client.stream(&prompt).await?.aggregate();
/// while let Some(event) = agg_stream.next().await {
/// // event now contains cumulative text
/// }
/// ```
fn aggregate(self) -> AggregatedChatStream<Self> {
AggregatedChatStream::new(self, AggregateMode::AggregatedOnly)
}
}
impl<T> AggregateStreamExt for T where T: Stream<Item = Result<ResponseEvent>> + Sized {}
impl<S> AggregatedChatStream<S> {
fn new(inner: S, mode: AggregateMode) -> Self {
AggregatedChatStream {
inner,
cumulative: String::new(),
cumulative_reasoning: String::new(),
pending: std::collections::VecDeque::new(),
mode,
}
}
pub(crate) fn streaming_mode(inner: S) -> Self {
Self::new(inner, AggregateMode::Streaming)
}
}

File diff suppressed because it is too large Load Diff

View File

@@ -1,345 +1,45 @@
use crate::client_common::tools::ToolSpec;
use std::borrow::Cow;
use std::ops::Deref;
use crate::error::Result;
use crate::model_family::ModelFamily;
use crate::protocol::RateLimitSnapshot;
use crate::protocol::TokenUsage;
use codex_api_client::EventStream;
pub use codex_api_client::Prompt;
pub use codex_api_client::Reasoning;
pub use codex_api_client::TextControls;
pub use codex_api_client::TextFormat;
pub use codex_api_client::TextFormatType;
use codex_apply_patch::APPLY_PATCH_TOOL_INSTRUCTIONS;
use codex_protocol::config_types::ReasoningEffort as ReasoningEffortConfig;
use codex_protocol::config_types::ReasoningSummary as ReasoningSummaryConfig;
use codex_protocol::config_types::Verbosity as VerbosityConfig;
use codex_protocol::models::ResponseItem;
use futures::Stream;
use serde::Deserialize;
use serde::Serialize;
use serde_json::Value;
use std::borrow::Cow;
use std::collections::HashSet;
use std::ops::Deref;
use std::pin::Pin;
use std::task::Context;
use std::task::Poll;
use tokio::sync::mpsc;
use crate::model_family::ModelFamily;
/// Review thread system prompt. Edit `core/src/review_prompt.md` to customize.
pub const REVIEW_PROMPT: &str = include_str!("../review_prompt.md");
/// API request payload for a single model turn
#[derive(Default, Debug, Clone)]
pub struct Prompt {
/// Conversation context input items.
pub input: Vec<ResponseItem>,
pub const REVIEW_EXIT_SUCCESS_TMPL: &str = include_str!("../templates/review/exit_success.xml");
pub const REVIEW_EXIT_INTERRUPTED_TMPL: &str =
include_str!("../templates/review/exit_interrupted.xml");
/// Tools available to the model, including additional tools sourced from
/// external MCP servers.
pub(crate) tools: Vec<ToolSpec>,
/// Whether parallel tool calls are permitted for this prompt.
pub(crate) parallel_tool_calls: bool,
/// Optional override for the built-in BASE_INSTRUCTIONS.
pub base_instructions_override: Option<String>,
/// Optional the output schema for the model's response.
pub output_schema: Option<Value>,
}
impl Prompt {
pub(crate) fn get_full_instructions<'a>(&'a self, model: &'a ModelFamily) -> Cow<'a, str> {
let base = self
.base_instructions_override
.as_deref()
.unwrap_or(model.base_instructions.deref());
// When there are no custom instructions, add apply_patch_tool_instructions if:
// - the model needs special instructions (4.1)
// AND
// - there is no apply_patch tool present
let is_apply_patch_tool_present = self.tools.iter().any(|tool| match tool {
ToolSpec::Function(f) => f.name == "apply_patch",
ToolSpec::Freeform(f) => f.name == "apply_patch",
_ => false,
});
if self.base_instructions_override.is_none()
&& model.needs_special_apply_patch_instructions
&& !is_apply_patch_tool_present
{
Cow::Owned(format!("{base}\n{APPLY_PATCH_TOOL_INSTRUCTIONS}"))
} else {
Cow::Borrowed(base)
}
}
pub(crate) fn get_formatted_input(&self) -> Vec<ResponseItem> {
let mut input = self.input.clone();
// when using the *Freeform* apply_patch tool specifically, tool outputs
// should be structured text, not json. Do NOT reserialize when using
// the Function tool - note that this differs from the check above for
// instructions. We declare the result as a named variable for clarity.
let is_freeform_apply_patch_tool_present = self.tools.iter().any(|tool| match tool {
ToolSpec::Freeform(f) => f.name == "apply_patch",
_ => false,
});
if is_freeform_apply_patch_tool_present {
reserialize_shell_outputs(&mut input);
}
input
pub fn compute_full_instructions<'a>(
base_override: Option<&'a str>,
model: &'a ModelFamily,
is_apply_patch_present: bool,
) -> Cow<'a, str> {
let base = base_override.unwrap_or(model.base_instructions.deref());
if base_override.is_none()
&& model.needs_special_apply_patch_instructions
&& !is_apply_patch_present
{
Cow::Owned(format!("{base}\n{APPLY_PATCH_TOOL_INSTRUCTIONS}"))
} else {
Cow::Borrowed(base)
}
}
fn reserialize_shell_outputs(items: &mut [ResponseItem]) {
let mut shell_call_ids: HashSet<String> = HashSet::new();
items.iter_mut().for_each(|item| match item {
ResponseItem::LocalShellCall { call_id, id, .. } => {
if let Some(identifier) = call_id.clone().or_else(|| id.clone()) {
shell_call_ids.insert(identifier);
}
}
ResponseItem::CustomToolCall {
id: _,
status: _,
call_id,
name,
input: _,
} => {
if name == "apply_patch" {
shell_call_ids.insert(call_id.clone());
}
}
ResponseItem::CustomToolCallOutput { call_id, output } => {
if shell_call_ids.remove(call_id)
&& let Some(structured) = parse_structured_shell_output(output)
{
*output = structured
}
}
ResponseItem::FunctionCall { name, call_id, .. }
if is_shell_tool_name(name) || name == "apply_patch" =>
{
shell_call_ids.insert(call_id.clone());
}
ResponseItem::FunctionCallOutput { call_id, output } => {
if shell_call_ids.remove(call_id)
&& let Some(structured) = parse_structured_shell_output(&output.content)
{
output.content = structured
}
}
_ => {}
})
}
fn is_shell_tool_name(name: &str) -> bool {
matches!(name, "shell" | "container.exec")
}
#[derive(Deserialize)]
struct ExecOutputJson {
output: String,
metadata: ExecOutputMetadataJson,
}
#[derive(Deserialize)]
struct ExecOutputMetadataJson {
exit_code: i32,
duration_seconds: f32,
}
fn parse_structured_shell_output(raw: &str) -> Option<String> {
let parsed: ExecOutputJson = serde_json::from_str(raw).ok()?;
Some(build_structured_output(&parsed))
}
fn build_structured_output(parsed: &ExecOutputJson) -> String {
let mut sections = Vec::new();
sections.push(format!("Exit code: {}", parsed.metadata.exit_code));
sections.push(format!(
"Wall time: {} seconds",
parsed.metadata.duration_seconds
));
let mut output = parsed.output.clone();
if let Some(total_lines) = extract_total_output_lines(&parsed.output) {
sections.push(format!("Total output lines: {total_lines}"));
if let Some(stripped) = strip_total_output_header(&output) {
output = stripped.to_string();
}
}
sections.push("Output:".to_string());
sections.push(output);
sections.join("\n")
}
fn extract_total_output_lines(output: &str) -> Option<u32> {
let marker_start = output.find("[... omitted ")?;
let marker = &output[marker_start..];
let (_, after_of) = marker.split_once(" of ")?;
let (total_segment, _) = after_of.split_once(' ')?;
total_segment.parse::<u32>().ok()
}
fn strip_total_output_header(output: &str) -> Option<&str> {
let after_prefix = output.strip_prefix("Total output lines: ")?;
let (_, remainder) = after_prefix.split_once('\n')?;
let remainder = remainder.strip_prefix('\n').unwrap_or(remainder);
Some(remainder)
}
#[derive(Debug)]
pub enum ResponseEvent {
Created,
OutputItemDone(ResponseItem),
Completed {
response_id: String,
token_usage: Option<TokenUsage>,
},
OutputTextDelta(String),
ReasoningSummaryDelta(String),
ReasoningContentDelta(String),
ReasoningSummaryPartAdded,
WebSearchCallBegin {
call_id: String,
},
RateLimits(RateLimitSnapshot),
}
#[derive(Debug, Serialize)]
pub(crate) struct Reasoning {
#[serde(skip_serializing_if = "Option::is_none")]
pub(crate) effort: Option<ReasoningEffortConfig>,
#[serde(skip_serializing_if = "Option::is_none")]
pub(crate) summary: Option<ReasoningSummaryConfig>,
}
#[derive(Debug, Serialize, Default, Clone)]
#[serde(rename_all = "snake_case")]
pub(crate) enum TextFormatType {
#[default]
JsonSchema,
}
#[derive(Debug, Serialize, Default, Clone)]
pub(crate) struct TextFormat {
pub(crate) r#type: TextFormatType,
pub(crate) strict: bool,
pub(crate) schema: Value,
pub(crate) name: String,
}
/// Controls under the `text` field in the Responses API for GPT-5.
#[derive(Debug, Serialize, Default, Clone)]
pub(crate) struct TextControls {
#[serde(skip_serializing_if = "Option::is_none")]
pub(crate) verbosity: Option<OpenAiVerbosity>,
#[serde(skip_serializing_if = "Option::is_none")]
pub(crate) format: Option<TextFormat>,
}
#[derive(Debug, Serialize, Default, Clone)]
#[serde(rename_all = "lowercase")]
pub(crate) enum OpenAiVerbosity {
Low,
#[default]
Medium,
High,
}
impl From<VerbosityConfig> for OpenAiVerbosity {
fn from(v: VerbosityConfig) -> Self {
match v {
VerbosityConfig::Low => OpenAiVerbosity::Low,
VerbosityConfig::Medium => OpenAiVerbosity::Medium,
VerbosityConfig::High => OpenAiVerbosity::High,
}
}
}
/// Request object that is serialized as JSON and POST'ed when using the
/// Responses API.
#[derive(Debug, Serialize)]
pub(crate) struct ResponsesApiRequest<'a> {
pub(crate) model: &'a str,
pub(crate) instructions: &'a str,
// TODO(mbolin): ResponseItem::Other should not be serialized. Currently,
// we code defensively to avoid this case, but perhaps we should use a
// separate enum for serialization.
pub(crate) input: &'a Vec<ResponseItem>,
pub(crate) tools: &'a [serde_json::Value],
pub(crate) tool_choice: &'static str,
pub(crate) parallel_tool_calls: bool,
pub(crate) reasoning: Option<Reasoning>,
pub(crate) store: bool,
pub(crate) stream: bool,
pub(crate) include: Vec<String>,
#[serde(skip_serializing_if = "Option::is_none")]
pub(crate) prompt_cache_key: Option<String>,
#[serde(skip_serializing_if = "Option::is_none")]
pub(crate) text: Option<TextControls>,
}
pub(crate) mod tools {
use crate::tools::spec::JsonSchema;
use serde::Deserialize;
use serde::Serialize;
/// When serialized as JSON, this produces a valid "Tool" in the OpenAI
/// Responses API.
#[derive(Debug, Clone, Serialize, PartialEq)]
#[serde(tag = "type")]
pub(crate) enum ToolSpec {
#[serde(rename = "function")]
Function(ResponsesApiTool),
#[serde(rename = "local_shell")]
LocalShell {},
// TODO: Understand why we get an error on web_search although the API docs say it's supported.
// https://platform.openai.com/docs/guides/tools-web-search?api-mode=responses#:~:text=%7B%20type%3A%20%22web_search%22%20%7D%2C
#[serde(rename = "web_search")]
WebSearch {},
#[serde(rename = "custom")]
Freeform(FreeformTool),
}
impl ToolSpec {
pub(crate) fn name(&self) -> &str {
match self {
ToolSpec::Function(tool) => tool.name.as_str(),
ToolSpec::LocalShell {} => "local_shell",
ToolSpec::WebSearch {} => "web_search",
ToolSpec::Freeform(tool) => tool.name.as_str(),
}
}
}
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct FreeformTool {
pub(crate) name: String,
pub(crate) description: String,
pub(crate) format: FreeformToolFormat,
}
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct FreeformToolFormat {
pub(crate) r#type: String,
pub(crate) syntax: String,
pub(crate) definition: String,
}
#[derive(Debug, Clone, Serialize, PartialEq)]
pub struct ResponsesApiTool {
pub(crate) name: String,
pub(crate) description: String,
/// TODO: Validation. When strict is set to true, the JSON schema,
/// `required` and `additional_properties` must be present. All fields in
/// `properties` must be present in `required`.
pub(crate) strict: bool,
pub(crate) parameters: JsonSchema,
}
}
pub(crate) fn create_reasoning_param_for_request(
pub fn create_reasoning_param_for_request(
model_family: &ModelFamily,
effort: Option<ReasoningEffortConfig>,
summary: ReasoningSummaryConfig,
@@ -354,7 +54,7 @@ pub(crate) fn create_reasoning_param_for_request(
})
}
pub(crate) fn create_text_param_for_request(
pub fn create_text_param_for_request(
verbosity: Option<VerbosityConfig>,
output_schema: &Option<Value>,
) -> Option<TextControls> {
@@ -363,7 +63,11 @@ pub(crate) fn create_text_param_for_request(
}
Some(TextControls {
verbosity: verbosity.map(std::convert::Into::into),
verbosity: verbosity.map(|v| match v {
VerbosityConfig::Low => "low".to_string(),
VerbosityConfig::Medium => "medium".to_string(),
VerbosityConfig::High => "high".to_string(),
}),
format: output_schema.as_ref().map(|schema| TextFormat {
r#type: TextFormatType::JsonSchema,
strict: true,
@@ -373,178 +77,54 @@ pub(crate) fn create_text_param_for_request(
})
}
pub struct ResponseStream {
pub(crate) rx_event: mpsc::Receiver<Result<ResponseEvent>>,
}
pub use codex_api_client::ResponseEvent;
impl Stream for ResponseStream {
type Item = Result<ResponseEvent>;
fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<Self::Item>> {
self.rx_event.poll_recv(cx)
}
}
pub type ResponseStream = EventStream<Result<ResponseEvent>>;
#[cfg(test)]
mod tests {
use crate::model_family::find_family_for_model;
use pretty_assertions::assert_eq;
use super::*;
struct InstructionsTestCase {
pub slug: &'static str,
pub expects_apply_patch_instructions: bool,
}
#[test]
fn get_full_instructions_no_user_content() {
let prompt = Prompt {
..Default::default()
};
let test_cases = vec![
InstructionsTestCase {
slug: "gpt-3.5",
expects_apply_patch_instructions: true,
},
InstructionsTestCase {
slug: "gpt-4.1",
expects_apply_patch_instructions: true,
},
InstructionsTestCase {
slug: "gpt-4o",
expects_apply_patch_instructions: true,
},
InstructionsTestCase {
slug: "gpt-5",
expects_apply_patch_instructions: true,
},
InstructionsTestCase {
slug: "codex-mini-latest",
expects_apply_patch_instructions: true,
},
InstructionsTestCase {
slug: "gpt-oss:120b",
expects_apply_patch_instructions: false,
},
InstructionsTestCase {
slug: "gpt-5-codex",
expects_apply_patch_instructions: false,
},
];
for test_case in test_cases {
let model_family = find_family_for_model(test_case.slug).expect("known model slug");
let expected = if test_case.expects_apply_patch_instructions {
format!(
"{}\n{}",
model_family.clone().base_instructions,
APPLY_PATCH_TOOL_INSTRUCTIONS
)
} else {
model_family.clone().base_instructions
};
let full = prompt.get_full_instructions(&model_family);
assert_eq!(full, expected);
}
}
use crate::model_family::find_family_for_model;
#[test]
fn serializes_text_verbosity_when_set() {
let input: Vec<ResponseItem> = vec![];
let tools: Vec<serde_json::Value> = vec![];
let req = ResponsesApiRequest {
model: "gpt-5",
instructions: "i",
input: &input,
tools: &tools,
tool_choice: "auto",
parallel_tool_calls: true,
reasoning: None,
store: false,
stream: true,
include: vec![],
prompt_cache_key: None,
text: Some(TextControls {
verbosity: Some(OpenAiVerbosity::Low),
format: None,
}),
};
fn compute_full_instructions_respects_apply_patch_flag() {
let model = find_family_for_model("gpt-4.1").expect("model");
let with_tool = compute_full_instructions(None, &model, true);
assert_eq!(with_tool.as_ref(), model.base_instructions.deref());
let v = serde_json::to_value(&req).expect("json");
assert_eq!(
v.get("text")
.and_then(|t| t.get("verbosity"))
.and_then(|s| s.as_str()),
Some("low")
let without_tool = compute_full_instructions(None, &model, false);
assert!(
without_tool
.as_ref()
.ends_with(APPLY_PATCH_TOOL_INSTRUCTIONS)
);
}
#[test]
fn serializes_text_schema_with_strict_format() {
let input: Vec<ResponseItem> = vec![];
let tools: Vec<serde_json::Value> = vec![];
fn create_text_controls_includes_verbosity() {
let controls = create_text_param_for_request(Some(VerbosityConfig::Low), &None)
.expect("text controls");
assert_eq!(controls.verbosity.as_deref(), Some("low"));
assert!(controls.format.is_none());
}
#[test]
fn create_text_controls_includes_schema() {
let schema = serde_json::json!({
"type": "object",
"properties": {
"answer": {"type": "string"}
},
"properties": {"answer": {"type": "string"}},
"required": ["answer"],
});
let text_controls =
let controls =
create_text_param_for_request(None, &Some(schema.clone())).expect("text controls");
let req = ResponsesApiRequest {
model: "gpt-5",
instructions: "i",
input: &input,
tools: &tools,
tool_choice: "auto",
parallel_tool_calls: true,
reasoning: None,
store: false,
stream: true,
include: vec![],
prompt_cache_key: None,
text: Some(text_controls),
};
let v = serde_json::to_value(&req).expect("json");
let text = v.get("text").expect("text field");
assert!(text.get("verbosity").is_none());
let format = text.get("format").expect("format field");
assert_eq!(
format.get("name"),
Some(&serde_json::Value::String("codex_output_schema".into()))
);
assert_eq!(
format.get("type"),
Some(&serde_json::Value::String("json_schema".into()))
);
assert_eq!(format.get("strict"), Some(&serde_json::Value::Bool(true)));
assert_eq!(format.get("schema"), Some(&schema));
let format = controls.format.expect("format");
assert_eq!(format.name, "codex_output_schema");
assert!(format.strict);
assert_eq!(format.schema, schema);
}
#[test]
fn omits_text_when_not_set() {
let input: Vec<ResponseItem> = vec![];
let tools: Vec<serde_json::Value> = vec![];
let req = ResponsesApiRequest {
model: "gpt-5",
instructions: "i",
input: &input,
tools: &tools,
tool_choice: "auto",
parallel_tool_calls: true,
reasoning: None,
store: false,
stream: true,
include: vec![],
prompt_cache_key: None,
text: None,
};
let v = serde_json::to_value(&req).expect("json");
assert!(v.get("text").is_none());
fn create_text_controls_none_when_no_options() {
assert!(create_text_param_for_request(None, &None).is_none());
}
}

File diff suppressed because it is too large Load Diff

View File

@@ -3,7 +3,7 @@ use std::sync::Arc;
use super::Session;
use super::TurnContext;
use super::get_last_assistant_message_from_turn;
use crate::Prompt;
use crate::client_common::Prompt;
use crate::client_common::ResponseEvent;
use crate::error::CodexErr;
use crate::error::Result as CodexResult;
@@ -13,7 +13,6 @@ use crate::protocol::ErrorEvent;
use crate::protocol::EventMsg;
use crate::protocol::TaskStartedEvent;
use crate::protocol::TurnContextItem;
use crate::state::TaskKind;
use crate::truncate::truncate_middle;
use crate::util::backoff;
use askama::Template;
@@ -40,9 +39,8 @@ pub(crate) async fn run_inline_auto_compact_task(
sess: Arc<Session>,
turn_context: Arc<TurnContext>,
) {
let input = vec![UserInput::Text {
text: SUMMARIZATION_PROMPT.to_string(),
}];
let prompt = turn_context.compact_prompt().to_string();
let input = vec![UserInput::Text { text: prompt }];
run_compact_task_inner(sess, turn_context, input).await;
}
@@ -86,11 +84,9 @@ async fn run_compact_task_inner(
loop {
let turn_input = history.get_history_for_prompt();
let prompt = Prompt {
input: turn_input.clone(),
..Default::default()
};
let attempt_result = drain_to_completed(&sess, turn_context.as_ref(), &prompt).await;
let turn_input_len = turn_input.len();
let (prompt, _) = crate::state::build_prompt_from_items(turn_input, None);
let attempt_result = drain_to_completed(&sess, turn_context.as_ref(), prompt).await;
match attempt_result {
Ok(()) => {
@@ -109,7 +105,7 @@ async fn run_compact_task_inner(
return;
}
Err(e @ CodexErr::ContextWindowExceeded) => {
if turn_input.len() > 1 {
if turn_input_len > 1 {
// Trim from the beginning to preserve cache (prefix-based) and keep recent messages intact.
error!(
"Context window exceeded while compacting; removing oldest history item. Error: {e}"
@@ -253,13 +249,9 @@ fn build_compacted_history_with_limit(
async fn drain_to_completed(
sess: &Session,
turn_context: &TurnContext,
prompt: &Prompt,
prompt: Prompt,
) -> CodexResult<()> {
let mut stream = turn_context
.client
.clone()
.stream_with_task_kind(prompt, TaskKind::Compact)
.await?;
let mut stream = turn_context.client.clone().stream(&prompt).await?;
loop {
let maybe_event = stream.next().await;
let Some(event) = maybe_event else {

View File

@@ -0,0 +1,295 @@
use std::sync::Arc;
use std::sync::atomic::AtomicU64;
use async_channel::Receiver;
use async_channel::Sender;
use codex_async_utils::OrCancelExt;
use codex_protocol::protocol::ApplyPatchApprovalRequestEvent;
use codex_protocol::protocol::Event;
use codex_protocol::protocol::EventMsg;
use codex_protocol::protocol::ExecApprovalRequestEvent;
use codex_protocol::protocol::Op;
use codex_protocol::protocol::SessionSource;
use codex_protocol::protocol::SubAgentSource;
use codex_protocol::protocol::Submission;
use codex_protocol::user_input::UserInput;
use tokio_util::sync::CancellationToken;
use crate::AuthManager;
use crate::codex::Codex;
use crate::codex::CodexSpawnOk;
use crate::codex::SUBMISSION_CHANNEL_CAPACITY;
use crate::codex::Session;
use crate::codex::TurnContext;
use crate::config::Config;
use crate::error::CodexErr;
use codex_protocol::protocol::InitialHistory;
/// Start an interactive sub-Codex conversation and return IO channels.
///
/// The returned `events_rx` yields non-approval events emitted by the sub-agent.
/// Approval requests are handled via `parent_session` and are not surfaced.
/// The returned `ops_tx` allows the caller to submit additional `Op`s to the sub-agent.
pub(crate) async fn run_codex_conversation_interactive(
config: Config,
auth_manager: Arc<AuthManager>,
parent_session: Arc<Session>,
parent_ctx: Arc<TurnContext>,
cancel_token: CancellationToken,
initial_history: Option<InitialHistory>,
) -> Result<Codex, CodexErr> {
let (tx_sub, rx_sub) = async_channel::bounded(SUBMISSION_CHANNEL_CAPACITY);
let (tx_ops, rx_ops) = async_channel::bounded(SUBMISSION_CHANNEL_CAPACITY);
let CodexSpawnOk { codex, .. } = Codex::spawn(
config,
auth_manager,
initial_history.unwrap_or(InitialHistory::New),
SessionSource::SubAgent(SubAgentSource::Review),
)
.await?;
let codex = Arc::new(codex);
// Use a child token so parent cancel cascades but we can scope it to this task
let cancel_token_events = cancel_token.child_token();
let cancel_token_ops = cancel_token.child_token();
// Forward events from the sub-agent to the consumer, filtering approvals and
// routing them to the parent session for decisions.
let parent_session_clone = Arc::clone(&parent_session);
let parent_ctx_clone = Arc::clone(&parent_ctx);
let codex_for_events = Arc::clone(&codex);
tokio::spawn(async move {
let _ = forward_events(
codex_for_events,
tx_sub,
parent_session_clone,
parent_ctx_clone,
cancel_token_events.clone(),
)
.or_cancel(&cancel_token_events)
.await;
});
// Forward ops from the caller to the sub-agent.
let codex_for_ops = Arc::clone(&codex);
tokio::spawn(async move {
forward_ops(codex_for_ops, rx_ops, cancel_token_ops).await;
});
Ok(Codex {
next_id: AtomicU64::new(0),
tx_sub: tx_ops,
rx_event: rx_sub,
})
}
/// Convenience wrapper for one-time use with an initial prompt.
///
/// Internally calls the interactive variant, then immediately submits the provided input.
pub(crate) async fn run_codex_conversation_one_shot(
config: Config,
auth_manager: Arc<AuthManager>,
input: Vec<UserInput>,
parent_session: Arc<Session>,
parent_ctx: Arc<TurnContext>,
cancel_token: CancellationToken,
initial_history: Option<InitialHistory>,
) -> Result<Codex, CodexErr> {
// Use a child token so we can stop the delegate after completion without
// requiring the caller to cancel the parent token.
let child_cancel = cancel_token.child_token();
let io = run_codex_conversation_interactive(
config,
auth_manager,
parent_session,
parent_ctx,
child_cancel.clone(),
initial_history,
)
.await?;
// Send the initial input to kick off the one-shot turn.
io.submit(Op::UserInput { items: input }).await?;
// Bridge events so we can observe completion and shut down automatically.
let (tx_bridge, rx_bridge) = async_channel::bounded(SUBMISSION_CHANNEL_CAPACITY);
let ops_tx = io.tx_sub.clone();
let io_for_bridge = io;
tokio::spawn(async move {
while let Ok(event) = io_for_bridge.next_event().await {
let should_shutdown = matches!(
event.msg,
EventMsg::TaskComplete(_) | EventMsg::TurnAborted(_)
);
let _ = tx_bridge.send(event).await;
if should_shutdown {
let _ = ops_tx
.send(Submission {
id: "shutdown".to_string(),
op: Op::Shutdown {},
})
.await;
child_cancel.cancel();
break;
}
}
});
// For one-shot usage, return a closed `tx_sub` so callers cannot submit
// additional ops after the initial request. Create a channel and drop the
// receiver to close it immediately.
let (tx_closed, rx_closed) = async_channel::bounded(SUBMISSION_CHANNEL_CAPACITY);
drop(rx_closed);
Ok(Codex {
next_id: AtomicU64::new(0),
rx_event: rx_bridge,
tx_sub: tx_closed,
})
}
async fn forward_events(
codex: Arc<Codex>,
tx_sub: Sender<Event>,
parent_session: Arc<Session>,
parent_ctx: Arc<TurnContext>,
cancel_token: CancellationToken,
) {
while let Ok(event) = codex.next_event().await {
match event {
Event {
id: _,
msg: EventMsg::SessionConfigured(_),
} => continue,
Event {
id,
msg: EventMsg::ExecApprovalRequest(event),
} => {
// Initiate approval via parent session; do not surface to consumer.
handle_exec_approval(
&codex,
id,
&parent_session,
&parent_ctx,
event,
&cancel_token,
)
.await;
}
Event {
id,
msg: EventMsg::ApplyPatchApprovalRequest(event),
} => {
handle_patch_approval(
&codex,
id,
&parent_session,
&parent_ctx,
event,
&cancel_token,
)
.await;
}
other => {
let _ = tx_sub.send(other).await;
}
}
}
}
/// Forward ops from a caller to a sub-agent, respecting cancellation.
async fn forward_ops(
codex: Arc<Codex>,
rx_ops: Receiver<Submission>,
cancel_token_ops: CancellationToken,
) {
loop {
let op: Op = match rx_ops.recv().or_cancel(&cancel_token_ops).await {
Ok(Ok(Submission { id: _, op })) => op,
Ok(Err(_)) | Err(_) => break,
};
let _ = codex.submit(op).await;
}
}
/// Handle an ExecApprovalRequest by consulting the parent session and replying.
async fn handle_exec_approval(
codex: &Codex,
id: String,
parent_session: &Session,
parent_ctx: &TurnContext,
event: ExecApprovalRequestEvent,
cancel_token: &CancellationToken,
) {
// Race approval with cancellation and timeout to avoid hangs.
let approval_fut = parent_session.request_command_approval(
parent_ctx,
parent_ctx.sub_id.clone(),
event.command,
event.cwd,
event.reason,
event.risk,
);
let decision = await_approval_with_cancel(
approval_fut,
parent_session,
&parent_ctx.sub_id,
cancel_token,
)
.await;
let _ = codex.submit(Op::ExecApproval { id, decision }).await;
}
/// Handle an ApplyPatchApprovalRequest by consulting the parent session and replying.
async fn handle_patch_approval(
codex: &Codex,
id: String,
parent_session: &Session,
parent_ctx: &TurnContext,
event: ApplyPatchApprovalRequestEvent,
cancel_token: &CancellationToken,
) {
let decision_rx = parent_session
.request_patch_approval(
parent_ctx,
parent_ctx.sub_id.clone(),
event.changes,
event.reason,
event.grant_root,
)
.await;
let decision = await_approval_with_cancel(
async move { decision_rx.await.unwrap_or_default() },
parent_session,
&parent_ctx.sub_id,
cancel_token,
)
.await;
let _ = codex.submit(Op::PatchApproval { id, decision }).await;
}
/// Await an approval decision, aborting on cancellation.
async fn await_approval_with_cancel<F>(
fut: F,
parent_session: &Session,
sub_id: &str,
cancel_token: &CancellationToken,
) -> codex_protocol::protocol::ReviewDecision
where
F: core::future::Future<Output = codex_protocol::protocol::ReviewDecision>,
{
tokio::select! {
biased;
_ = cancel_token.cancelled() => {
parent_session
.notify_approval(sub_id, codex_protocol::protocol::ReviewDecision::Abort)
.await;
codex_protocol::protocol::ReviewDecision::Abort
}
decision = fut => {
decision
}
}
}

View File

@@ -0,0 +1,954 @@
use crate::config::CONFIG_TOML_FILE;
use crate::config::types::McpServerConfig;
use crate::config::types::Notice;
use anyhow::Context;
use codex_protocol::config_types::ReasoningEffort;
use std::collections::BTreeMap;
use std::path::Path;
use std::path::PathBuf;
use tempfile::NamedTempFile;
use tokio::task;
use toml_edit::DocumentMut;
use toml_edit::Item as TomlItem;
use toml_edit::Table as TomlTable;
use toml_edit::value;
/// Discrete config mutations supported by the persistence engine.
#[derive(Clone, Debug)]
pub enum ConfigEdit {
/// Update the active (or default) model selection and optional reasoning effort.
SetModel {
model: Option<String>,
effort: Option<ReasoningEffort>,
},
/// Toggle the acknowledgement flag under `[notice]`.
SetNoticeHideFullAccessWarning(bool),
/// Toggle the Windows onboarding acknowledgement flag.
SetWindowsWslSetupAcknowledged(bool),
/// Replace the entire `[mcp_servers]` table.
ReplaceMcpServers(BTreeMap<String, McpServerConfig>),
/// Set trust_level = "trusted" under `[projects."<path>"]`,
/// migrating inline tables to explicit tables.
SetProjectTrusted(PathBuf),
/// Set the value stored at the exact dotted path.
SetPath {
segments: Vec<String>,
value: TomlItem,
},
/// Remove the value stored at the exact dotted path.
ClearPath { segments: Vec<String> },
}
// TODO(jif) move to a dedicated file
mod document_helpers {
use crate::config::types::McpServerConfig;
use crate::config::types::McpServerTransportConfig;
use toml_edit::Array as TomlArray;
use toml_edit::InlineTable;
use toml_edit::Item as TomlItem;
use toml_edit::Table as TomlTable;
use toml_edit::value;
pub(super) fn ensure_table_for_write(item: &mut TomlItem) -> Option<&mut TomlTable> {
match item {
TomlItem::Table(table) => Some(table),
TomlItem::Value(value) => {
if let Some(inline) = value.as_inline_table() {
*item = TomlItem::Table(table_from_inline(inline));
item.as_table_mut()
} else {
*item = TomlItem::Table(new_implicit_table());
item.as_table_mut()
}
}
TomlItem::None => {
*item = TomlItem::Table(new_implicit_table());
item.as_table_mut()
}
_ => None,
}
}
pub(super) fn ensure_table_for_read(item: &mut TomlItem) -> Option<&mut TomlTable> {
match item {
TomlItem::Table(table) => Some(table),
TomlItem::Value(value) => {
let inline = value.as_inline_table()?;
*item = TomlItem::Table(table_from_inline(inline));
item.as_table_mut()
}
_ => None,
}
}
pub(super) fn serialize_mcp_server(config: &McpServerConfig) -> TomlItem {
let mut entry = TomlTable::new();
entry.set_implicit(false);
match &config.transport {
McpServerTransportConfig::Stdio {
command,
args,
env,
env_vars,
cwd,
} => {
entry["command"] = value(command.clone());
if !args.is_empty() {
entry["args"] = array_from_iter(args.iter().cloned());
}
if let Some(env) = env
&& !env.is_empty()
{
entry["env"] = table_from_pairs(env.iter());
}
if !env_vars.is_empty() {
entry["env_vars"] = array_from_iter(env_vars.iter().cloned());
}
if let Some(cwd) = cwd {
entry["cwd"] = value(cwd.to_string_lossy().to_string());
}
}
McpServerTransportConfig::StreamableHttp {
url,
bearer_token_env_var,
http_headers,
env_http_headers,
} => {
entry["url"] = value(url.clone());
if let Some(env_var) = bearer_token_env_var {
entry["bearer_token_env_var"] = value(env_var.clone());
}
if let Some(headers) = http_headers
&& !headers.is_empty()
{
entry["http_headers"] = table_from_pairs(headers.iter());
}
if let Some(headers) = env_http_headers
&& !headers.is_empty()
{
entry["env_http_headers"] = table_from_pairs(headers.iter());
}
}
}
if !config.enabled {
entry["enabled"] = value(false);
}
if let Some(timeout) = config.startup_timeout_sec {
entry["startup_timeout_sec"] = value(timeout.as_secs_f64());
}
if let Some(timeout) = config.tool_timeout_sec {
entry["tool_timeout_sec"] = value(timeout.as_secs_f64());
}
if let Some(enabled_tools) = &config.enabled_tools
&& !enabled_tools.is_empty()
{
entry["enabled_tools"] = array_from_iter(enabled_tools.iter().cloned());
}
if let Some(disabled_tools) = &config.disabled_tools
&& !disabled_tools.is_empty()
{
entry["disabled_tools"] = array_from_iter(disabled_tools.iter().cloned());
}
TomlItem::Table(entry)
}
fn table_from_inline(inline: &InlineTable) -> TomlTable {
let mut table = new_implicit_table();
for (key, value) in inline.iter() {
let mut value = value.clone();
let decor = value.decor_mut();
decor.set_suffix("");
table.insert(key, TomlItem::Value(value));
}
table
}
pub(super) fn new_implicit_table() -> TomlTable {
let mut table = TomlTable::new();
table.set_implicit(true);
table
}
fn array_from_iter<I>(iter: I) -> TomlItem
where
I: Iterator<Item = String>,
{
let mut array = TomlArray::new();
for value in iter {
array.push(value);
}
TomlItem::Value(array.into())
}
fn table_from_pairs<'a, I>(pairs: I) -> TomlItem
where
I: IntoIterator<Item = (&'a String, &'a String)>,
{
let mut entries: Vec<_> = pairs.into_iter().collect();
entries.sort_by(|(a, _), (b, _)| a.cmp(b));
let mut table = TomlTable::new();
table.set_implicit(false);
for (key, val) in entries {
table.insert(key, value(val.clone()));
}
TomlItem::Table(table)
}
}
struct ConfigDocument {
doc: DocumentMut,
profile: Option<String>,
}
#[derive(Copy, Clone)]
enum Scope {
Global,
Profile,
}
#[derive(Copy, Clone)]
enum TraversalMode {
Create,
Existing,
}
impl ConfigDocument {
fn new(doc: DocumentMut, profile: Option<String>) -> Self {
Self { doc, profile }
}
fn apply(&mut self, edit: &ConfigEdit) -> anyhow::Result<bool> {
match edit {
ConfigEdit::SetModel { model, effort } => Ok({
let mut mutated = false;
mutated |= self.write_profile_value(
&["model"],
model.as_ref().map(|model_value| value(model_value.clone())),
);
mutated |= self.write_profile_value(
&["model_reasoning_effort"],
effort.map(|effort| value(effort.to_string())),
);
mutated
}),
ConfigEdit::SetNoticeHideFullAccessWarning(acknowledged) => Ok(self.write_value(
Scope::Global,
&[Notice::TABLE_KEY, "hide_full_access_warning"],
value(*acknowledged),
)),
ConfigEdit::SetWindowsWslSetupAcknowledged(acknowledged) => Ok(self.write_value(
Scope::Global,
&["windows_wsl_setup_acknowledged"],
value(*acknowledged),
)),
ConfigEdit::ReplaceMcpServers(servers) => Ok(self.replace_mcp_servers(servers)),
ConfigEdit::SetPath { segments, value } => Ok(self.insert(segments, value.clone())),
ConfigEdit::ClearPath { segments } => Ok(self.clear_owned(segments)),
ConfigEdit::SetProjectTrusted(project_path) => {
// Delegate to the existing, tested logic in config.rs to
// ensure tables are explicit and migration is preserved.
crate::config::set_project_trusted_inner(&mut self.doc, project_path.as_path())?;
Ok(true)
}
}
}
fn write_profile_value(&mut self, segments: &[&str], value: Option<TomlItem>) -> bool {
match value {
Some(item) => self.write_value(Scope::Profile, segments, item),
None => self.clear(Scope::Profile, segments),
}
}
fn write_value(&mut self, scope: Scope, segments: &[&str], value: TomlItem) -> bool {
let resolved = self.scoped_segments(scope, segments);
self.insert(&resolved, value)
}
fn clear(&mut self, scope: Scope, segments: &[&str]) -> bool {
let resolved = self.scoped_segments(scope, segments);
self.remove(&resolved)
}
fn clear_owned(&mut self, segments: &[String]) -> bool {
self.remove(segments)
}
fn replace_mcp_servers(&mut self, servers: &BTreeMap<String, McpServerConfig>) -> bool {
if servers.is_empty() {
return self.clear(Scope::Global, &["mcp_servers"]);
}
let mut table = TomlTable::new();
table.set_implicit(true);
for (name, config) in servers {
table.insert(name, document_helpers::serialize_mcp_server(config));
}
let item = TomlItem::Table(table);
self.write_value(Scope::Global, &["mcp_servers"], item)
}
fn scoped_segments(&self, scope: Scope, segments: &[&str]) -> Vec<String> {
let resolved: Vec<String> = segments
.iter()
.map(|segment| (*segment).to_string())
.collect();
if matches!(scope, Scope::Profile)
&& resolved.first().is_none_or(|segment| segment != "profiles")
&& let Some(profile) = self.profile.as_deref()
{
let mut scoped = Vec::with_capacity(resolved.len() + 2);
scoped.push("profiles".to_string());
scoped.push(profile.to_string());
scoped.extend(resolved);
return scoped;
}
resolved
}
fn insert(&mut self, segments: &[String], value: TomlItem) -> bool {
let Some((last, parents)) = segments.split_last() else {
return false;
};
let Some(parent) = self.descend(parents, TraversalMode::Create) else {
return false;
};
parent[last] = value;
true
}
fn remove(&mut self, segments: &[String]) -> bool {
let Some((last, parents)) = segments.split_last() else {
return false;
};
let Some(parent) = self.descend(parents, TraversalMode::Existing) else {
return false;
};
parent.remove(last).is_some()
}
fn descend(&mut self, segments: &[String], mode: TraversalMode) -> Option<&mut TomlTable> {
let mut current = self.doc.as_table_mut();
for segment in segments {
match mode {
TraversalMode::Create => {
if !current.contains_key(segment.as_str()) {
current.insert(
segment.as_str(),
TomlItem::Table(document_helpers::new_implicit_table()),
);
}
let item = current.get_mut(segment.as_str())?;
current = document_helpers::ensure_table_for_write(item)?;
}
TraversalMode::Existing => {
let item = current.get_mut(segment.as_str())?;
current = document_helpers::ensure_table_for_read(item)?;
}
}
}
Some(current)
}
}
/// Persist edits using a blocking strategy.
pub fn apply_blocking(
codex_home: &Path,
profile: Option<&str>,
edits: &[ConfigEdit],
) -> anyhow::Result<()> {
if edits.is_empty() {
return Ok(());
}
let config_path = codex_home.join(CONFIG_TOML_FILE);
let serialized = match std::fs::read_to_string(&config_path) {
Ok(contents) => contents,
Err(err) if err.kind() == std::io::ErrorKind::NotFound => String::new(),
Err(err) => return Err(err.into()),
};
let doc = if serialized.is_empty() {
DocumentMut::new()
} else {
serialized.parse::<DocumentMut>()?
};
let profile = profile.map(ToOwned::to_owned).or_else(|| {
doc.get("profile")
.and_then(|item| item.as_str())
.map(ToOwned::to_owned)
});
let mut document = ConfigDocument::new(doc, profile);
let mut mutated = false;
for edit in edits {
mutated |= document.apply(edit)?;
}
if !mutated {
return Ok(());
}
std::fs::create_dir_all(codex_home).with_context(|| {
format!(
"failed to create Codex home directory at {}",
codex_home.display()
)
})?;
let tmp = NamedTempFile::new_in(codex_home)?;
std::fs::write(tmp.path(), document.doc.to_string()).with_context(|| {
format!(
"failed to write temporary config file at {}",
tmp.path().display()
)
})?;
tmp.persist(config_path)?;
Ok(())
}
/// Persist edits asynchronously by offloading the blocking writer.
pub async fn apply(
codex_home: &Path,
profile: Option<&str>,
edits: Vec<ConfigEdit>,
) -> anyhow::Result<()> {
let codex_home = codex_home.to_path_buf();
let profile = profile.map(ToOwned::to_owned);
task::spawn_blocking(move || apply_blocking(&codex_home, profile.as_deref(), &edits))
.await
.context("config persistence task panicked")?
}
/// Fluent builder to batch config edits and apply them atomically.
#[derive(Default)]
pub struct ConfigEditsBuilder {
codex_home: PathBuf,
profile: Option<String>,
edits: Vec<ConfigEdit>,
}
impl ConfigEditsBuilder {
pub fn new(codex_home: &Path) -> Self {
Self {
codex_home: codex_home.to_path_buf(),
profile: None,
edits: Vec::new(),
}
}
pub fn with_profile(mut self, profile: Option<&str>) -> Self {
self.profile = profile.map(ToOwned::to_owned);
self
}
pub fn set_model(mut self, model: Option<&str>, effort: Option<ReasoningEffort>) -> Self {
self.edits.push(ConfigEdit::SetModel {
model: model.map(ToOwned::to_owned),
effort,
});
self
}
pub fn set_hide_full_access_warning(mut self, acknowledged: bool) -> Self {
self.edits
.push(ConfigEdit::SetNoticeHideFullAccessWarning(acknowledged));
self
}
pub fn set_windows_wsl_setup_acknowledged(mut self, acknowledged: bool) -> Self {
self.edits
.push(ConfigEdit::SetWindowsWslSetupAcknowledged(acknowledged));
self
}
pub fn replace_mcp_servers(mut self, servers: &BTreeMap<String, McpServerConfig>) -> Self {
self.edits
.push(ConfigEdit::ReplaceMcpServers(servers.clone()));
self
}
pub fn set_project_trusted<P: Into<PathBuf>>(mut self, project_path: P) -> Self {
self.edits
.push(ConfigEdit::SetProjectTrusted(project_path.into()));
self
}
/// Apply edits on a blocking thread.
pub fn apply_blocking(self) -> anyhow::Result<()> {
apply_blocking(&self.codex_home, self.profile.as_deref(), &self.edits)
}
/// Apply edits asynchronously via a blocking offload.
pub async fn apply(self) -> anyhow::Result<()> {
task::spawn_blocking(move || {
apply_blocking(&self.codex_home, self.profile.as_deref(), &self.edits)
})
.await
.context("config persistence task panicked")?
}
}
#[cfg(test)]
mod tests {
use super::*;
use crate::config::types::McpServerTransportConfig;
use codex_protocol::config_types::ReasoningEffort;
use pretty_assertions::assert_eq;
use tempfile::tempdir;
use tokio::runtime::Builder;
use toml::Value as TomlValue;
#[test]
fn blocking_set_model_top_level() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
apply_blocking(
codex_home,
None,
&[ConfigEdit::SetModel {
model: Some("gpt-5-codex".to_string()),
effort: Some(ReasoningEffort::High),
}],
)
.expect("persist");
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = r#"model = "gpt-5-codex"
model_reasoning_effort = "high"
"#;
assert_eq!(contents, expected);
}
#[test]
fn blocking_set_model_preserves_inline_table_contents() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
// Seed with inline tables for profiles to simulate common user config.
std::fs::write(
codex_home.join(CONFIG_TOML_FILE),
r#"profile = "fast"
profiles = { fast = { model = "gpt-4o", sandbox_mode = "strict" } }
"#,
)
.expect("seed");
apply_blocking(
codex_home,
None,
&[ConfigEdit::SetModel {
model: Some("o4-mini".to_string()),
effort: None,
}],
)
.expect("persist");
let raw = std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let value: TomlValue = toml::from_str(&raw).expect("parse config");
// Ensure sandbox_mode is preserved under profiles.fast and model updated.
let profiles_tbl = value
.get("profiles")
.and_then(|v| v.as_table())
.expect("profiles table");
let fast_tbl = profiles_tbl
.get("fast")
.and_then(|v| v.as_table())
.expect("fast table");
assert_eq!(
fast_tbl.get("sandbox_mode").and_then(|v| v.as_str()),
Some("strict")
);
assert_eq!(
fast_tbl.get("model").and_then(|v| v.as_str()),
Some("o4-mini")
);
}
#[test]
fn blocking_clear_model_removes_inline_table_entry() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
std::fs::write(
codex_home.join(CONFIG_TOML_FILE),
r#"profile = "fast"
profiles = { fast = { model = "gpt-4o", sandbox_mode = "strict" } }
"#,
)
.expect("seed");
apply_blocking(
codex_home,
None,
&[ConfigEdit::SetModel {
model: None,
effort: Some(ReasoningEffort::High),
}],
)
.expect("persist");
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = r#"profile = "fast"
[profiles.fast]
sandbox_mode = "strict"
model_reasoning_effort = "high"
"#;
assert_eq!(contents, expected);
}
#[test]
fn blocking_set_model_scopes_to_active_profile() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
std::fs::write(
codex_home.join(CONFIG_TOML_FILE),
r#"profile = "team"
[profiles.team]
model_reasoning_effort = "low"
"#,
)
.expect("seed");
apply_blocking(
codex_home,
None,
&[ConfigEdit::SetModel {
model: Some("o5-preview".to_string()),
effort: Some(ReasoningEffort::Minimal),
}],
)
.expect("persist");
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = r#"profile = "team"
[profiles.team]
model_reasoning_effort = "minimal"
model = "o5-preview"
"#;
assert_eq!(contents, expected);
}
#[test]
fn blocking_set_model_with_explicit_profile() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
std::fs::write(
codex_home.join(CONFIG_TOML_FILE),
r#"[profiles."team a"]
model = "gpt-5-codex"
"#,
)
.expect("seed");
apply_blocking(
codex_home,
Some("team a"),
&[ConfigEdit::SetModel {
model: Some("o4-mini".to_string()),
effort: None,
}],
)
.expect("persist");
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = r#"[profiles."team a"]
model = "o4-mini"
"#;
assert_eq!(contents, expected);
}
#[test]
fn blocking_set_hide_full_access_warning_preserves_table() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
std::fs::write(
codex_home.join(CONFIG_TOML_FILE),
r#"# Global comment
[notice]
# keep me
existing = "value"
"#,
)
.expect("seed");
apply_blocking(
codex_home,
None,
&[ConfigEdit::SetNoticeHideFullAccessWarning(true)],
)
.expect("persist");
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = r#"# Global comment
[notice]
# keep me
existing = "value"
hide_full_access_warning = true
"#;
assert_eq!(contents, expected);
}
#[test]
fn blocking_replace_mcp_servers_round_trips() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
let mut servers = BTreeMap::new();
servers.insert(
"stdio".to_string(),
McpServerConfig {
transport: McpServerTransportConfig::Stdio {
command: "cmd".to_string(),
args: vec!["--flag".to_string()],
env: Some(
[
("B".to_string(), "2".to_string()),
("A".to_string(), "1".to_string()),
]
.into_iter()
.collect(),
),
env_vars: vec!["FOO".to_string()],
cwd: None,
},
enabled: true,
startup_timeout_sec: None,
tool_timeout_sec: None,
enabled_tools: Some(vec!["one".to_string(), "two".to_string()]),
disabled_tools: None,
},
);
servers.insert(
"http".to_string(),
McpServerConfig {
transport: McpServerTransportConfig::StreamableHttp {
url: "https://example.com".to_string(),
bearer_token_env_var: Some("TOKEN".to_string()),
http_headers: Some(
[("Z-Header".to_string(), "z".to_string())]
.into_iter()
.collect(),
),
env_http_headers: None,
},
enabled: false,
startup_timeout_sec: Some(std::time::Duration::from_secs(5)),
tool_timeout_sec: None,
enabled_tools: None,
disabled_tools: Some(vec!["forbidden".to_string()]),
},
);
apply_blocking(
codex_home,
None,
&[ConfigEdit::ReplaceMcpServers(servers.clone())],
)
.expect("persist");
let raw = std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = "\
[mcp_servers.http]
url = \"https://example.com\"
bearer_token_env_var = \"TOKEN\"
enabled = false
startup_timeout_sec = 5.0
disabled_tools = [\"forbidden\"]
[mcp_servers.http.http_headers]
Z-Header = \"z\"
[mcp_servers.stdio]
command = \"cmd\"
args = [\"--flag\"]
env_vars = [\"FOO\"]
enabled_tools = [\"one\", \"two\"]
[mcp_servers.stdio.env]
A = \"1\"
B = \"2\"
";
assert_eq!(raw, expected);
}
#[test]
fn blocking_clear_path_noop_when_missing() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
apply_blocking(
codex_home,
None,
&[ConfigEdit::ClearPath {
segments: vec!["missing".to_string()],
}],
)
.expect("apply");
assert!(
!codex_home.join(CONFIG_TOML_FILE).exists(),
"config.toml should not be created on noop"
);
}
#[test]
fn blocking_set_path_updates_notifications() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
let item = value(false);
apply_blocking(
codex_home,
None,
&[ConfigEdit::SetPath {
segments: vec!["tui".to_string(), "notifications".to_string()],
value: item,
}],
)
.expect("apply");
let raw = std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let config: TomlValue = toml::from_str(&raw).expect("parse config");
let notifications = config
.get("tui")
.and_then(|item| item.as_table())
.and_then(|tbl| tbl.get("notifications"))
.and_then(toml::Value::as_bool);
assert_eq!(notifications, Some(false));
}
#[tokio::test]
async fn async_builder_set_model_persists() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path().to_path_buf();
ConfigEditsBuilder::new(&codex_home)
.set_model(Some("gpt-5-codex"), Some(ReasoningEffort::High))
.apply()
.await
.expect("persist");
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = r#"model = "gpt-5-codex"
model_reasoning_effort = "high"
"#;
assert_eq!(contents, expected);
}
#[test]
fn blocking_builder_set_model_round_trips_back_and_forth() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
let initial_expected = r#"model = "o4-mini"
model_reasoning_effort = "low"
"#;
ConfigEditsBuilder::new(codex_home)
.set_model(Some("o4-mini"), Some(ReasoningEffort::Low))
.apply_blocking()
.expect("persist initial");
let mut contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
assert_eq!(contents, initial_expected);
let updated_expected = r#"model = "gpt-5-codex"
model_reasoning_effort = "high"
"#;
ConfigEditsBuilder::new(codex_home)
.set_model(Some("gpt-5-codex"), Some(ReasoningEffort::High))
.apply_blocking()
.expect("persist update");
contents = std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
assert_eq!(contents, updated_expected);
ConfigEditsBuilder::new(codex_home)
.set_model(Some("o4-mini"), Some(ReasoningEffort::Low))
.apply_blocking()
.expect("persist revert");
contents = std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
assert_eq!(contents, initial_expected);
}
#[test]
fn blocking_set_asynchronous_helpers_available() {
let rt = Builder::new_current_thread()
.enable_all()
.build()
.expect("runtime");
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path().to_path_buf();
rt.block_on(async {
ConfigEditsBuilder::new(&codex_home)
.set_hide_full_access_warning(true)
.apply()
.await
.expect("persist");
});
let raw = std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let notice = toml::from_str::<TomlValue>(&raw)
.expect("parse config")
.get("notice")
.and_then(|item| item.as_table())
.and_then(|tbl| tbl.get("hide_full_access_warning"))
.and_then(toml::Value::as_bool);
assert_eq!(notice, Some(true));
}
#[test]
fn replace_mcp_servers_blocking_clears_table_when_empty() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
std::fs::write(
codex_home.join(CONFIG_TOML_FILE),
"[mcp_servers]\nfoo = { command = \"cmd\" }\n",
)
.expect("seed");
apply_blocking(
codex_home,
None,
&[ConfigEdit::ReplaceMcpServers(BTreeMap::new())],
)
.expect("persist");
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
assert!(!contents.contains("mcp_servers"));
}
}

View File

@@ -22,6 +22,7 @@ pub struct ConfigProfile {
pub model_verbosity: Option<Verbosity>,
pub chatgpt_base_url: Option<String>,
pub experimental_instructions_file: Option<PathBuf>,
pub experimental_compact_prompt_file: Option<PathBuf>,
pub include_apply_patch_tool: Option<bool>,
pub include_view_image_tool: Option<bool>,
pub experimental_use_unified_exec_tool: Option<bool>,

View File

@@ -361,7 +361,7 @@ pub struct Notice {
}
impl Notice {
/// used by set_hide_full_access_warning until we refactor config updates
/// referenced by config_edit helpers when writing notice flags
pub(crate) const TABLE_KEY: &'static str = "notice";
}

View File

@@ -1,748 +0,0 @@
use crate::config::CONFIG_TOML_FILE;
use anyhow::Result;
use std::path::Path;
use tempfile::NamedTempFile;
use toml_edit::DocumentMut;
pub const CONFIG_KEY_MODEL: &str = "model";
pub const CONFIG_KEY_EFFORT: &str = "model_reasoning_effort";
#[derive(Copy, Clone)]
enum NoneBehavior {
Skip,
Remove,
}
/// Persist overrides into `config.toml` using explicit key segments per
/// override. This avoids ambiguity with keys that contain dots or spaces.
pub async fn persist_overrides(
codex_home: &Path,
profile: Option<&str>,
overrides: &[(&[&str], &str)],
) -> Result<()> {
let with_options: Vec<(&[&str], Option<&str>)> = overrides
.iter()
.map(|(segments, value)| (*segments, Some(*value)))
.collect();
persist_overrides_with_behavior(codex_home, profile, &with_options, NoneBehavior::Skip).await
}
/// Persist overrides where values may be optional. Any entries with `None`
/// values are skipped. If all values are `None`, this becomes a no-op and
/// returns `Ok(())` without touching the file.
pub async fn persist_non_null_overrides(
codex_home: &Path,
profile: Option<&str>,
overrides: &[(&[&str], Option<&str>)],
) -> Result<()> {
persist_overrides_with_behavior(codex_home, profile, overrides, NoneBehavior::Skip).await
}
/// Persist overrides where `None` values clear any existing values from the
/// configuration file.
pub async fn persist_overrides_and_clear_if_none(
codex_home: &Path,
profile: Option<&str>,
overrides: &[(&[&str], Option<&str>)],
) -> Result<()> {
persist_overrides_with_behavior(codex_home, profile, overrides, NoneBehavior::Remove).await
}
/// Apply a single override onto a `toml_edit` document while preserving
/// existing formatting/comments.
/// The key is expressed as explicit segments to correctly handle keys that
/// contain dots or spaces.
fn apply_toml_edit_override_segments(
doc: &mut DocumentMut,
segments: &[&str],
value: toml_edit::Item,
) {
use toml_edit::Item;
if segments.is_empty() {
return;
}
let mut current = doc.as_table_mut();
for seg in &segments[..segments.len() - 1] {
if !current.contains_key(seg) {
current[*seg] = Item::Table(toml_edit::Table::new());
if let Some(t) = current[*seg].as_table_mut() {
t.set_implicit(true);
}
}
let maybe_item = current.get_mut(seg);
let Some(item) = maybe_item else { return };
if !item.is_table() {
*item = Item::Table(toml_edit::Table::new());
if let Some(t) = item.as_table_mut() {
t.set_implicit(true);
}
}
let Some(tbl) = item.as_table_mut() else {
return;
};
current = tbl;
}
let last = segments[segments.len() - 1];
current[last] = value;
}
async fn persist_overrides_with_behavior(
codex_home: &Path,
profile: Option<&str>,
overrides: &[(&[&str], Option<&str>)],
none_behavior: NoneBehavior,
) -> Result<()> {
if overrides.is_empty() {
return Ok(());
}
let should_skip = match none_behavior {
NoneBehavior::Skip => overrides.iter().all(|(_, value)| value.is_none()),
NoneBehavior::Remove => false,
};
if should_skip {
return Ok(());
}
let config_path = codex_home.join(CONFIG_TOML_FILE);
let read_result = tokio::fs::read_to_string(&config_path).await;
let mut doc = match read_result {
Ok(contents) => contents.parse::<DocumentMut>()?,
Err(e) if e.kind() == std::io::ErrorKind::NotFound => {
if overrides
.iter()
.all(|(_, value)| value.is_none() && matches!(none_behavior, NoneBehavior::Remove))
{
return Ok(());
}
tokio::fs::create_dir_all(codex_home).await?;
DocumentMut::new()
}
Err(e) => return Err(e.into()),
};
let effective_profile = if let Some(p) = profile {
Some(p.to_owned())
} else {
doc.get("profile")
.and_then(|i| i.as_str())
.map(str::to_string)
};
let mut mutated = false;
for (segments, value) in overrides.iter().copied() {
let mut seg_buf: Vec<&str> = Vec::new();
let segments_to_apply: &[&str];
if let Some(ref name) = effective_profile {
if segments.first().copied() == Some("profiles") {
segments_to_apply = segments;
} else {
seg_buf.reserve(2 + segments.len());
seg_buf.push("profiles");
seg_buf.push(name.as_str());
seg_buf.extend_from_slice(segments);
segments_to_apply = seg_buf.as_slice();
}
} else {
segments_to_apply = segments;
}
match value {
Some(v) => {
let item_value = toml_edit::value(v);
apply_toml_edit_override_segments(&mut doc, segments_to_apply, item_value);
mutated = true;
}
None => {
if matches!(none_behavior, NoneBehavior::Remove)
&& remove_toml_edit_segments(&mut doc, segments_to_apply)
{
mutated = true;
}
}
}
}
if !mutated {
return Ok(());
}
let tmp_file = NamedTempFile::new_in(codex_home)?;
tokio::fs::write(tmp_file.path(), doc.to_string()).await?;
tmp_file.persist(config_path)?;
Ok(())
}
fn remove_toml_edit_segments(doc: &mut DocumentMut, segments: &[&str]) -> bool {
use toml_edit::Item;
if segments.is_empty() {
return false;
}
let mut current = doc.as_table_mut();
for seg in &segments[..segments.len() - 1] {
let Some(item) = current.get_mut(seg) else {
return false;
};
match item {
Item::Table(table) => {
current = table;
}
_ => {
return false;
}
}
}
current.remove(segments[segments.len() - 1]).is_some()
}
#[cfg(test)]
mod tests {
use super::*;
use pretty_assertions::assert_eq;
use tempfile::tempdir;
/// Verifies model and effort are written at top-level when no profile is set.
#[tokio::test]
async fn set_default_model_and_effort_top_level_when_no_profile() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
persist_overrides(
codex_home,
None,
&[
(&[CONFIG_KEY_MODEL], "gpt-5-codex"),
(&[CONFIG_KEY_EFFORT], "high"),
],
)
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = r#"model = "gpt-5-codex"
model_reasoning_effort = "high"
"#;
assert_eq!(contents, expected);
}
/// Verifies values are written under the active profile when `profile` is set.
#[tokio::test]
async fn set_defaults_update_profile_when_profile_set() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
// Seed config with a profile selection but without profiles table
let seed = "profile = \"o3\"\n";
tokio::fs::write(codex_home.join(CONFIG_TOML_FILE), seed)
.await
.expect("seed write");
persist_overrides(
codex_home,
None,
&[
(&[CONFIG_KEY_MODEL], "o3"),
(&[CONFIG_KEY_EFFORT], "minimal"),
],
)
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = r#"profile = "o3"
[profiles.o3]
model = "o3"
model_reasoning_effort = "minimal"
"#;
assert_eq!(contents, expected);
}
/// Verifies profile names with dots/spaces are preserved via explicit segments.
#[tokio::test]
async fn set_defaults_update_profile_with_dot_and_space() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
// Seed config with a profile name that contains a dot and a space
let seed = "profile = \"my.team name\"\n";
tokio::fs::write(codex_home.join(CONFIG_TOML_FILE), seed)
.await
.expect("seed write");
persist_overrides(
codex_home,
None,
&[
(&[CONFIG_KEY_MODEL], "o3"),
(&[CONFIG_KEY_EFFORT], "minimal"),
],
)
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = r#"profile = "my.team name"
[profiles."my.team name"]
model = "o3"
model_reasoning_effort = "minimal"
"#;
assert_eq!(contents, expected);
}
/// Verifies explicit profile override writes under that profile even without active profile.
#[tokio::test]
async fn set_defaults_update_when_profile_override_supplied() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
// No profile key in config.toml
tokio::fs::write(codex_home.join(CONFIG_TOML_FILE), "")
.await
.expect("seed write");
// Persist with an explicit profile override
persist_overrides(
codex_home,
Some("o3"),
&[(&[CONFIG_KEY_MODEL], "o3"), (&[CONFIG_KEY_EFFORT], "high")],
)
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = r#"[profiles.o3]
model = "o3"
model_reasoning_effort = "high"
"#;
assert_eq!(contents, expected);
}
/// Verifies nested tables are created as needed when applying overrides.
#[tokio::test]
async fn persist_overrides_creates_nested_tables() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
persist_overrides(
codex_home,
None,
&[
(&["a", "b", "c"], "v"),
(&["x"], "y"),
(&["profiles", "p1", CONFIG_KEY_MODEL], "gpt-5-codex"),
],
)
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = r#"x = "y"
[a.b]
c = "v"
[profiles.p1]
model = "gpt-5-codex"
"#;
assert_eq!(contents, expected);
}
/// Verifies a scalar key becomes a table when nested keys are written.
#[tokio::test]
async fn persist_overrides_replaces_scalar_with_table() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
let seed = "foo = \"bar\"\n";
tokio::fs::write(codex_home.join(CONFIG_TOML_FILE), seed)
.await
.expect("seed write");
persist_overrides(codex_home, None, &[(&["foo", "bar", "baz"], "ok")])
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = r#"[foo.bar]
baz = "ok"
"#;
assert_eq!(contents, expected);
}
/// Verifies comments and spacing are preserved when writing under active profile.
#[tokio::test]
async fn set_defaults_preserve_comments() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
// Seed a config with comments and spacing we expect to preserve
let seed = r#"# Global comment
# Another line
profile = "o3"
# Profile settings
[profiles.o3]
# keep me
existing = "keep"
"#;
tokio::fs::write(codex_home.join(CONFIG_TOML_FILE), seed)
.await
.expect("seed write");
// Apply defaults; since profile is set, it should write under [profiles.o3]
persist_overrides(
codex_home,
None,
&[(&[CONFIG_KEY_MODEL], "o3"), (&[CONFIG_KEY_EFFORT], "high")],
)
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = r#"# Global comment
# Another line
profile = "o3"
# Profile settings
[profiles.o3]
# keep me
existing = "keep"
model = "o3"
model_reasoning_effort = "high"
"#;
assert_eq!(contents, expected);
}
/// Verifies comments and spacing are preserved when writing at top level.
#[tokio::test]
async fn set_defaults_preserve_global_comments() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
// Seed a config WITHOUT a profile, containing comments and spacing
let seed = r#"# Top-level comments
# should be preserved
existing = "keep"
"#;
tokio::fs::write(codex_home.join(CONFIG_TOML_FILE), seed)
.await
.expect("seed write");
// Since there is no profile, the defaults should be written at top-level
persist_overrides(
codex_home,
None,
&[
(&[CONFIG_KEY_MODEL], "gpt-5-codex"),
(&[CONFIG_KEY_EFFORT], "minimal"),
],
)
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = r#"# Top-level comments
# should be preserved
existing = "keep"
model = "gpt-5-codex"
model_reasoning_effort = "minimal"
"#;
assert_eq!(contents, expected);
}
/// Verifies errors on invalid TOML propagate and file is not clobbered.
#[tokio::test]
async fn persist_overrides_errors_on_parse_failure() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
// Write an intentionally invalid TOML file
let invalid = "invalid = [unclosed";
tokio::fs::write(codex_home.join(CONFIG_TOML_FILE), invalid)
.await
.expect("seed write");
// Attempting to persist should return an error and must not clobber the file.
let res = persist_overrides(codex_home, None, &[(&["x"], "y")]).await;
assert!(res.is_err(), "expected parse error to propagate");
// File should be unchanged
let contents = read_config(codex_home).await;
assert_eq!(contents, invalid);
}
/// Verifies changing model only preserves existing effort at top-level.
#[tokio::test]
async fn changing_only_model_preserves_existing_effort_top_level() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
// Seed with an effort value only
let seed = "model_reasoning_effort = \"minimal\"\n";
tokio::fs::write(codex_home.join(CONFIG_TOML_FILE), seed)
.await
.expect("seed write");
// Change only the model
persist_overrides(codex_home, None, &[(&[CONFIG_KEY_MODEL], "o3")])
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = r#"model_reasoning_effort = "minimal"
model = "o3"
"#;
assert_eq!(contents, expected);
}
/// Verifies changing effort only preserves existing model at top-level.
#[tokio::test]
async fn changing_only_effort_preserves_existing_model_top_level() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
// Seed with a model value only
let seed = "model = \"gpt-5-codex\"\n";
tokio::fs::write(codex_home.join(CONFIG_TOML_FILE), seed)
.await
.expect("seed write");
// Change only the effort
persist_overrides(codex_home, None, &[(&[CONFIG_KEY_EFFORT], "high")])
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = r#"model = "gpt-5-codex"
model_reasoning_effort = "high"
"#;
assert_eq!(contents, expected);
}
/// Verifies changing model only preserves existing effort in active profile.
#[tokio::test]
async fn changing_only_model_preserves_effort_in_active_profile() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
// Seed with an active profile and an existing effort under that profile
let seed = r#"profile = "p1"
[profiles.p1]
model_reasoning_effort = "low"
"#;
tokio::fs::write(codex_home.join(CONFIG_TOML_FILE), seed)
.await
.expect("seed write");
persist_overrides(codex_home, None, &[(&[CONFIG_KEY_MODEL], "o4-mini")])
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = r#"profile = "p1"
[profiles.p1]
model_reasoning_effort = "low"
model = "o4-mini"
"#;
assert_eq!(contents, expected);
}
/// Verifies changing effort only preserves existing model in a profile override.
#[tokio::test]
async fn changing_only_effort_preserves_model_in_profile_override() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
// No active profile key; we'll target an explicit override
let seed = r#"[profiles.team]
model = "gpt-5-codex"
"#;
tokio::fs::write(codex_home.join(CONFIG_TOML_FILE), seed)
.await
.expect("seed write");
persist_overrides(
codex_home,
Some("team"),
&[(&[CONFIG_KEY_EFFORT], "minimal")],
)
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = r#"[profiles.team]
model = "gpt-5-codex"
model_reasoning_effort = "minimal"
"#;
assert_eq!(contents, expected);
}
/// Verifies `persist_non_null_overrides` skips `None` entries and writes only present values at top-level.
#[tokio::test]
async fn persist_non_null_skips_none_top_level() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
persist_non_null_overrides(
codex_home,
None,
&[
(&[CONFIG_KEY_MODEL], Some("gpt-5-codex")),
(&[CONFIG_KEY_EFFORT], None),
],
)
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = "model = \"gpt-5-codex\"\n";
assert_eq!(contents, expected);
}
/// Verifies no-op behavior when all provided overrides are `None` (no file created/modified).
#[tokio::test]
async fn persist_non_null_noop_when_all_none() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
persist_non_null_overrides(
codex_home,
None,
&[(&["a"], None), (&["profiles", "p", "x"], None)],
)
.await
.expect("persist");
// Should not create config.toml on a pure no-op
assert!(!codex_home.join(CONFIG_TOML_FILE).exists());
}
/// Verifies entries are written under the specified profile and `None` entries are skipped.
#[tokio::test]
async fn persist_non_null_respects_profile_override() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
persist_non_null_overrides(
codex_home,
Some("team"),
&[
(&[CONFIG_KEY_MODEL], Some("o3")),
(&[CONFIG_KEY_EFFORT], None),
],
)
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = r#"[profiles.team]
model = "o3"
"#;
assert_eq!(contents, expected);
}
#[tokio::test]
async fn persist_clear_none_removes_top_level_value() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
let seed = r#"model = "gpt-5-codex"
model_reasoning_effort = "medium"
"#;
tokio::fs::write(codex_home.join(CONFIG_TOML_FILE), seed)
.await
.expect("seed write");
persist_overrides_and_clear_if_none(
codex_home,
None,
&[
(&[CONFIG_KEY_MODEL], None),
(&[CONFIG_KEY_EFFORT], Some("high")),
],
)
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = "model_reasoning_effort = \"high\"\n";
assert_eq!(contents, expected);
}
#[tokio::test]
async fn persist_clear_none_respects_active_profile() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
let seed = r#"profile = "team"
[profiles.team]
model = "gpt-4"
model_reasoning_effort = "minimal"
"#;
tokio::fs::write(codex_home.join(CONFIG_TOML_FILE), seed)
.await
.expect("seed write");
persist_overrides_and_clear_if_none(
codex_home,
None,
&[
(&[CONFIG_KEY_MODEL], None),
(&[CONFIG_KEY_EFFORT], Some("high")),
],
)
.await
.expect("persist");
let contents = read_config(codex_home).await;
let expected = r#"profile = "team"
[profiles.team]
model_reasoning_effort = "high"
"#;
assert_eq!(contents, expected);
}
#[tokio::test]
async fn persist_clear_none_noop_when_file_missing() {
let tmpdir = tempdir().expect("tmp");
let codex_home = tmpdir.path();
persist_overrides_and_clear_if_none(codex_home, None, &[(&[CONFIG_KEY_MODEL], None)])
.await
.expect("persist");
assert!(!codex_home.join(CONFIG_TOML_FILE).exists());
}
// Test helper moved to bottom per review guidance.
async fn read_config(codex_home: &Path) -> String {
let p = codex_home.join(CONFIG_TOML_FILE);
tokio::fs::read_to_string(p).await.unwrap_or_default()
}
}

View File

@@ -1,12 +1,14 @@
use codex_protocol::models::FunctionCallOutputContentItem;
use codex_protocol::models::FunctionCallOutputPayload;
use codex_protocol::models::ResponseItem;
use crate::util::error_or_panic;
use codex_protocol::protocol::TokenUsage;
use codex_protocol::protocol::TokenUsageInfo;
use codex_utils_string::take_bytes_at_char_boundary;
use codex_utils_string::take_last_bytes_at_char_boundary;
use std::collections::HashSet;
use std::ops::Deref;
use tracing::error;
// Model-formatting limits: clients get full streams; only content sent to the model is truncated.
pub(crate) const MODEL_FORMAT_MAX_BYTES: usize = 10 * 1024; // 10 KiB
@@ -21,6 +23,13 @@ pub(crate) struct ConversationHistory {
/// The oldest items are at the beginning of the vector.
items: Vec<ResponseItem>,
token_info: Option<TokenUsageInfo>,
responses_api_chain: Option<ResponsesApiChainState>,
}
#[derive(Debug, Clone, Default)]
pub(crate) struct ResponsesApiChainState {
pub last_response_id: Option<String>,
pub last_message_id: Option<String>,
}
impl ConversationHistory {
@@ -28,6 +37,7 @@ impl ConversationHistory {
Self {
items: Vec::new(),
token_info: TokenUsageInfo::new_or_append(&None, &None, None),
responses_api_chain: None,
}
}
@@ -70,6 +80,10 @@ impl ConversationHistory {
// Returns the history prepared for sending to the model.
// With extra response items filtered out and GhostCommits removed.
pub(crate) fn get_history_for_prompt(&mut self) -> Vec<ResponseItem> {
self.build_prompt_history()
}
fn build_prompt_history(&mut self) -> Vec<ResponseItem> {
let mut history = self.get_history();
Self::remove_ghost_snapshots(&mut history);
Self::remove_reasoning_before_last_turn(&mut history);
@@ -90,6 +104,7 @@ impl ConversationHistory {
pub(crate) fn replace(&mut self, items: Vec<ResponseItem>) {
self.items = items;
self.reset_responses_api_chain();
}
pub(crate) fn update_token_info(
@@ -428,6 +443,18 @@ impl ConversationHistory {
| ResponseItem::Other => item.clone(),
}
}
pub(crate) fn responses_api_chain(&self) -> Option<ResponsesApiChainState> {
self.responses_api_chain.clone()
}
pub(crate) fn reset_responses_api_chain(&mut self) {
self.responses_api_chain = None;
}
pub(crate) fn set_responses_api_chain(&mut self, chain: ResponsesApiChainState) {
self.responses_api_chain = Some(chain);
}
}
pub(crate) fn format_output_for_model_body(content: &str) -> String {
@@ -501,15 +528,6 @@ fn truncate_formatted_exec_output(content: &str, total_lines: usize) -> String {
result
}
#[inline]
fn error_or_panic(message: String) {
if cfg!(debug_assertions) || env!("CARGO_PKG_VERSION").contains("alpha") {
panic!("{message}");
} else {
error!("{message}");
}
}
/// Anything that is not a system message or "reasoning" message is considered
/// an API message.
fn is_api_message(message: &ResponseItem) -> bool {
@@ -527,6 +545,102 @@ fn is_api_message(message: &ResponseItem) -> bool {
}
}
fn reserialize_shell_outputs(items: &mut [ResponseItem]) {
let mut shell_call_ids: HashSet<String> = HashSet::new();
items.iter_mut().for_each(|item| match item {
ResponseItem::LocalShellCall { call_id, id, .. } => {
if let Some(identifier) = call_id.clone().or_else(|| id.clone()) {
shell_call_ids.insert(identifier);
}
}
ResponseItem::CustomToolCall { call_id, name, .. } => {
if name == "apply_patch" {
shell_call_ids.insert(call_id.clone());
}
}
ResponseItem::CustomToolCallOutput { call_id, output } => {
if shell_call_ids.remove(call_id)
&& let Some(structured) = parse_structured_shell_output(output)
{
*output = structured;
}
}
ResponseItem::FunctionCall { name, call_id, .. }
if name == "shell" || name == "container.exec" || name == "apply_patch" =>
{
shell_call_ids.insert(call_id.clone());
}
ResponseItem::FunctionCallOutput { call_id, output } => {
if shell_call_ids.remove(call_id)
&& let Some(structured) = parse_structured_shell_output(&output.content)
{
output.content = structured;
}
}
_ => {}
});
}
#[derive(serde::Deserialize)]
struct ExecOutputJson {
output: String,
metadata: ExecOutputMetadataJson,
}
#[derive(serde::Deserialize)]
struct ExecOutputMetadataJson {
exit_code: i32,
duration_seconds: f32,
}
fn parse_structured_shell_output(raw: &str) -> Option<String> {
let parsed: ExecOutputJson = serde_json::from_str(raw).ok()?;
Some(build_structured_output(&parsed))
}
fn build_structured_output(parsed: &ExecOutputJson) -> String {
let mut sections = Vec::new();
sections.push(format!("Exit code: {}", parsed.metadata.exit_code));
sections.push(format!(
"Wall time: {} seconds",
parsed.metadata.duration_seconds
));
let mut output = parsed.output.clone();
if let Some(total_lines) = extract_total_output_lines(&parsed.output) {
sections.push(format!("Total output lines: {total_lines}"));
if let Some(stripped) = strip_total_output_header(&output) {
output = stripped.to_string();
}
}
sections.push("Output:".to_string());
sections.push(output);
sections.join("\n")
}
fn extract_total_output_lines(output: &str) -> Option<u32> {
let marker_start = output.find("[... omitted ")?;
let marker = &output[marker_start..];
let (_, after_of) = marker.split_once(" of ")?;
let (total_segment, _) = after_of.split_once(' ')?;
total_segment.parse::<u32>().ok()
}
fn strip_total_output_header(output: &str) -> Option<&str> {
let after_prefix = output.strip_prefix("Total output lines: ")?;
let (_, remainder) = after_prefix.split_once('\n')?;
let remainder = remainder.strip_prefix('\n').unwrap_or(remainder);
Some(remainder)
}
pub(crate) fn format_prompt_items(items: &mut [ResponseItem], has_freeform_apply_patch: bool) {
if has_freeform_apply_patch {
reserialize_shell_outputs(items);
}
}
#[cfg(test)]
mod tests {
use super::*;

View File

@@ -73,7 +73,7 @@ impl ConversationManager {
config,
auth_manager,
InitialHistory::New,
self.session_source,
self.session_source.clone(),
)
.await?;
self.finalize_spawn(codex, conversation_id).await
@@ -145,7 +145,13 @@ impl ConversationManager {
let CodexSpawnOk {
codex,
conversation_id,
} = Codex::spawn(config, auth_manager, initial_history, self.session_source).await?;
} = Codex::spawn(
config,
auth_manager,
initial_history,
self.session_source.clone(),
)
.await?;
self.finalize_spawn(codex, conversation_id).await
}
@@ -179,7 +185,7 @@ impl ConversationManager {
let CodexSpawnOk {
codex,
conversation_id,
} = Codex::spawn(config, auth_manager, history, self.session_source).await?;
} = Codex::spawn(config, auth_manager, history, self.session_source.clone()).await?;
self.finalize_spawn(codex, conversation_id).await
}

View File

@@ -41,6 +41,14 @@ impl CodexHttpClient {
Self { inner }
}
pub fn inner(&self) -> &reqwest::Client {
&self.inner
}
pub fn clone_inner(&self) -> reqwest::Client {
self.inner.clone()
}
pub fn get<U>(&self, url: U) -> CodexRequestBuilder
where
U: IntoUrl,

View File

@@ -253,7 +253,7 @@ impl std::fmt::Display for UsageLimitReachedError {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
let message = match self.plan_type.as_ref() {
Some(PlanType::Known(KnownPlan::Plus)) => format!(
"You've hit your usage limit. Upgrade to Pro (https://openai.com/chatgpt/pricing){}",
"You've hit your usage limit. Upgrade to Pro (https://openai.com/chatgpt/pricing), visit chatgpt.com/codex/settings/usage to purchase more credits{}",
retry_suffix_after_or(self.resets_at.as_ref())
),
Some(PlanType::Known(KnownPlan::Team)) | Some(PlanType::Known(KnownPlan::Business)) => {
@@ -266,8 +266,11 @@ impl std::fmt::Display for UsageLimitReachedError {
"You've hit your usage limit. Upgrade to Plus to continue using Codex (https://openai.com/chatgpt/pricing)."
.to_string()
}
Some(PlanType::Known(KnownPlan::Pro))
| Some(PlanType::Known(KnownPlan::Enterprise))
Some(PlanType::Known(KnownPlan::Pro)) => format!(
"You've hit your usage limit. Visit chatgpt.com/codex/settings/usage to purchase more credits{}",
retry_suffix_after_or(self.resets_at.as_ref())
),
Some(PlanType::Known(KnownPlan::Enterprise))
| Some(PlanType::Known(KnownPlan::Edu)) => format!(
"You've hit your usage limit.{}",
retry_suffix(self.resets_at.as_ref())
@@ -467,7 +470,7 @@ mod tests {
};
assert_eq!(
err.to_string(),
"You've hit your usage limit. Upgrade to Pro (https://openai.com/chatgpt/pricing) or try again later."
"You've hit your usage limit. Upgrade to Pro (https://openai.com/chatgpt/pricing), visit chatgpt.com/codex/settings/usage to purchase more credits or try again later."
);
}
@@ -597,7 +600,7 @@ mod tests {
#[test]
fn usage_limit_reached_error_formats_default_for_other_plans() {
let err = UsageLimitReachedError {
plan_type: Some(PlanType::Known(KnownPlan::Pro)),
plan_type: Some(PlanType::Known(KnownPlan::Enterprise)),
resets_at: None,
rate_limits: Some(rate_limit_snapshot()),
};
@@ -607,6 +610,23 @@ mod tests {
);
}
#[test]
fn usage_limit_reached_error_formats_pro_plan_with_reset() {
let base = Utc.with_ymd_and_hms(2024, 1, 1, 0, 0, 0).unwrap();
let resets_at = base + ChronoDuration::hours(1);
with_now_override(base, move || {
let err = UsageLimitReachedError {
plan_type: Some(PlanType::Known(KnownPlan::Pro)),
resets_at: Some(resets_at),
rate_limits: Some(rate_limit_snapshot()),
};
assert_eq!(
err.to_string(),
"You've hit your usage limit. Visit chatgpt.com/codex/settings/usage to purchase more credits or try again in 1 hour."
);
});
}
#[test]
fn usage_limit_reached_includes_minutes_when_available() {
let base = Utc.with_ymd_and_hms(2024, 1, 1, 0, 0, 0).unwrap();
@@ -636,7 +656,7 @@ mod tests {
};
assert_eq!(
err.to_string(),
"You've hit your usage limit. Upgrade to Pro (https://openai.com/chatgpt/pricing) or try again in 3 hours 32 minutes."
"You've hit your usage limit. Upgrade to Pro (https://openai.com/chatgpt/pricing), visit chatgpt.com/codex/settings/usage to purchase more credits or try again in 3 hours 32 minutes."
);
});
}

View File

@@ -11,6 +11,7 @@ use codex_protocol::models::ResponseItem;
use codex_protocol::models::WebSearchAction;
use codex_protocol::user_input::UserInput;
use tracing::warn;
use uuid::Uuid;
fn is_session_prefix(text: &str) -> bool {
let trimmed = text.trim_start();
@@ -46,7 +47,7 @@ fn parse_user_message(message: &[ContentItem]) -> Option<UserMessageItem> {
Some(UserMessageItem::new(&content))
}
fn parse_agent_message(message: &[ContentItem]) -> AgentMessageItem {
fn parse_agent_message(id: Option<&String>, message: &[ContentItem]) -> AgentMessageItem {
let mut content: Vec<AgentMessageContent> = Vec::new();
for content_item in message.iter() {
match content_item {
@@ -61,14 +62,18 @@ fn parse_agent_message(message: &[ContentItem]) -> AgentMessageItem {
}
}
}
AgentMessageItem::new(&content)
let id = id.cloned().unwrap_or_else(|| Uuid::new_v4().to_string());
AgentMessageItem { id, content }
}
pub fn parse_turn_item(item: &ResponseItem) -> Option<TurnItem> {
match item {
ResponseItem::Message { role, content, .. } => match role.as_str() {
ResponseItem::Message { role, content, id } => match role.as_str() {
"user" => parse_user_message(content).map(TurnItem::UserMessage),
"assistant" => Some(TurnItem::AgentMessage(parse_agent_message(content))),
"assistant" => Some(TurnItem::AgentMessage(parse_agent_message(
id.as_ref(),
content,
))),
"system" => None,
_ => None,
},

View File

@@ -1,6 +1,6 @@
use crate::config_types::EnvironmentVariablePattern;
use crate::config_types::ShellEnvironmentPolicy;
use crate::config_types::ShellEnvironmentPolicyInherit;
use crate::config::types::EnvironmentVariablePattern;
use crate::config::types::ShellEnvironmentPolicy;
use crate::config::types::ShellEnvironmentPolicyInherit;
use std::collections::HashMap;
use std::collections::HashSet;
@@ -71,7 +71,7 @@ where
#[cfg(test)]
mod tests {
use super::*;
use crate::config_types::ShellEnvironmentPolicyInherit;
use crate::config::types::ShellEnvironmentPolicyInherit;
use maplit::hashmap;
fn make_vars(pairs: &[(&str, &str)]) -> Vec<(String, String)> {

View File

@@ -6,7 +6,7 @@
//! container attached to `Config`.
use crate::config::ConfigToml;
use crate::config_profile::ConfigProfile;
use crate::config::profile::ConfigProfile;
use serde::Deserialize;
use std::collections::BTreeMap;
use std::collections::BTreeSet;
@@ -43,6 +43,8 @@ pub enum Feature {
SandboxCommandAssessment,
/// Create a ghost commit at each turn.
GhostCommit,
/// Enable chaining Responses API calls via previous response IDs.
ResponsesApiChaining,
}
impl Feature {
@@ -122,8 +124,9 @@ impl Features {
self.enabled.insert(f);
}
pub fn disable(&mut self, f: Feature) {
pub fn disable(&mut self, f: Feature) -> &mut Self {
self.enabled.remove(&f);
self
}
pub fn record_legacy_usage_force(&mut self, alias: &str, feature: Feature) {
@@ -294,4 +297,10 @@ pub const FEATURES: &[FeatureSpec] = &[
stage: Stage::Experimental,
default_enabled: false,
},
FeatureSpec {
id: Feature::ResponsesApiChaining,
key: "responses_api_chaining",
stage: Stage::Experimental,
default_enabled: false,
},
];

View File

@@ -8,22 +8,21 @@
mod apply_patch;
pub mod auth;
pub mod bash;
mod chat_completions;
mod client;
mod client_common;
pub mod codex;
mod codex_conversation;
pub use codex_conversation::CodexConversation;
mod codex_delegate;
mod command_safety;
pub mod config;
pub mod config_edit;
pub mod config_loader;
pub mod config_profile;
pub mod config_types;
mod conversation_history;
mod conversation_manager;
pub mod custom_prompts;
mod environment_context;
pub mod error;
mod event_mapping;
pub mod exec;
pub mod exec_env;
pub mod features;
@@ -34,22 +33,14 @@ pub mod mcp;
mod mcp_connection_manager;
mod mcp_tool_call;
mod message_history;
mod model_provider_info;
pub mod parse_command;
mod response_processing;
pub mod review_format;
pub mod sandboxing;
pub mod token_data;
mod truncate;
mod unified_exec;
mod user_instructions;
pub use model_provider_info::BUILT_IN_OSS_MODEL_PROVIDER_ID;
pub use model_provider_info::ModelProviderInfo;
pub use model_provider_info::WireApi;
pub use model_provider_info::built_in_model_providers;
pub use model_provider_info::create_oss_provider_with_base_url;
mod conversation_manager;
mod event_mapping;
pub mod review_format;
pub use codex_protocol::protocol::InitialHistory;
pub use conversation_manager::ConversationManager;
pub use conversation_manager::NewConversation;

View File

@@ -7,8 +7,8 @@ use codex_rmcp_client::determine_streamable_http_auth_status;
use futures::future::join_all;
use tracing::warn;
use crate::config_types::McpServerConfig;
use crate::config_types::McpServerTransportConfig;
use crate::config::types::McpServerConfig;
use crate::config::types::McpServerTransportConfig;
#[derive(Debug, Clone)]
pub struct McpAuthStatusEntry {

View File

@@ -37,8 +37,8 @@ use tokio::task::JoinSet;
use tracing::info;
use tracing::warn;
use crate::config_types::McpServerConfig;
use crate::config_types::McpServerTransportConfig;
use crate::config::types::McpServerConfig;
use crate::config::types::McpServerTransportConfig;
/// Delimiter used to separate the server name from the tool name in a fully
/// qualified tool name.

View File

@@ -28,7 +28,7 @@ use tokio::fs;
use tokio::io::AsyncReadExt;
use crate::config::Config;
use crate::config_types::HistoryPersistence;
use crate::config::types::HistoryPersistence;
use codex_protocol::ConversationId;
#[cfg(unix)]

View File

@@ -1,4 +1,4 @@
use crate::config_types::ReasoningSummaryFormat;
use crate::config::types::ReasoningSummaryFormat;
use crate::tools::handlers::apply_patch::ApplyPatchToolType;
/// The `instructions` field in the payload sent to a model should always start

View File

@@ -1,6 +1,6 @@
use crate::config::Config;
use crate::config_types::OtelExporterKind as Kind;
use crate::config_types::OtelHttpProtocol as Protocol;
use crate::config::types::OtelExporterKind as Kind;
use crate::config::types::OtelHttpProtocol as Protocol;
use crate::default_client::originator;
use codex_otel::config::OtelExporter;
use codex_otel::config::OtelHttpProtocol;

View File

@@ -1,6 +1,5 @@
use crate::codex::Session;
use crate::codex::TurnContext;
use crate::conversation_history::ConversationHistory;
use codex_protocol::models::FunctionCallOutputPayload;
use codex_protocol::models::ResponseInputItem;
use codex_protocol::models::ResponseItem;
@@ -11,8 +10,6 @@ use tracing::warn;
/// - `ResponseInputItem`s to send back to the model on the next turn.
pub(crate) async fn process_items(
processed_items: Vec<crate::codex::ProcessedResponseItem>,
is_review_mode: bool,
review_thread_history: &mut ConversationHistory,
sess: &Session,
turn_context: &TurnContext,
) -> (Vec<ResponseInputItem>, Vec<ResponseItem>) {
@@ -100,12 +97,8 @@ pub(crate) async fn process_items(
// Only attempt to take the lock if there is something to record.
if !items_to_record_in_conversation_history.is_empty() {
if is_review_mode {
review_thread_history.record_items(items_to_record_in_conversation_history.iter());
} else {
sess.record_conversation_items(turn_context, &items_to_record_in_conversation_history)
.await;
}
sess.record_conversation_items(turn_context, &items_to_record_in_conversation_history)
.await;
}
(responses, items_to_record_in_conversation_history)
}

View File

@@ -409,7 +409,7 @@ async fn read_head_and_tail(
match rollout_line.item {
RolloutItem::SessionMeta(session_meta_line) => {
summary.source = Some(session_meta_line.meta.source);
summary.source = Some(session_meta_line.meta.source.clone());
summary.model_provider = session_meta_line.meta.model_provider.clone();
summary.created_at = summary
.created_at

View File

@@ -77,6 +77,9 @@ pub(crate) fn should_persist_event_msg(ev: &EventMsg) -> bool {
| EventMsg::ViewImageToolCall(_)
| EventMsg::DeprecationNotice(_)
| EventMsg::ItemStarted(_)
| EventMsg::ItemCompleted(_) => false,
| EventMsg::ItemCompleted(_)
| EventMsg::AgentMessageContentDelta(_)
| EventMsg::ReasoningContentDelta(_)
| EventMsg::ReasoningRawContentDelta(_) => false,
}
}

View File

@@ -5,18 +5,19 @@ use std::time::Duration;
use std::time::Instant;
use crate::AuthManager;
use crate::ModelProviderInfo;
use crate::client::ModelClient;
use crate::client_common::Prompt;
use crate::client_common::ResponseEvent;
use crate::config::Config;
use crate::protocol::SandboxPolicy;
use askama::Template;
use codex_api_client::ModelProviderInfo;
use codex_api_client::Prompt;
use codex_otel::otel_event_manager::OtelEventManager;
use codex_protocol::ConversationId;
use codex_protocol::models::ContentItem;
use codex_protocol::models::ResponseItem;
use codex_protocol::protocol::SandboxCommandAssessment;
use codex_protocol::protocol::SessionSource;
use futures::StreamExt;
use serde_json::json;
use tokio::time::timeout;
@@ -53,6 +54,7 @@ pub(crate) async fn assess_command(
auth_manager: Arc<AuthManager>,
parent_otel: &OtelEventManager,
conversation_id: ConversationId,
session_source: SessionSource,
call_id: &str,
command: &[String],
sandbox_policy: &SandboxPolicy,
@@ -124,12 +126,10 @@ pub(crate) async fn assess_command(
role: "user".to_string(),
content: vec![ContentItem::InputText { text: user_prompt }],
}],
tools: Vec::new(),
parallel_tool_calls: false,
base_instructions_override: Some(system_prompt),
output_schema: Some(sandbox_assessment_schema()),
instructions: system_prompt,
..Default::default()
};
let child_otel =
parent_otel.with_model(config.model.as_str(), config.model_family.slug.as_str());
@@ -141,6 +141,7 @@ pub(crate) async fn assess_command(
config.model_reasoning_effort,
config.model_reasoning_summary,
conversation_id,
session_source,
);
let start = Instant::now();

View File

@@ -4,6 +4,8 @@ mod turn;
pub(crate) use service::SessionServices;
pub(crate) use session::SessionState;
pub(crate) use session::build_prompt_from_items;
pub(crate) use session::response_item_id;
pub(crate) use turn::ActiveTurn;
pub(crate) use turn::RunningTask;
pub(crate) use turn::TaskKind;

View File

@@ -2,26 +2,41 @@
use codex_protocol::models::ResponseItem;
use crate::client_common::Prompt;
use crate::client_common::compute_full_instructions;
use crate::codex::SessionConfiguration;
use crate::conversation_history::ConversationHistory;
use crate::conversation_history::ResponsesApiChainState;
use crate::conversation_history::format_prompt_items;
use crate::features::Feature;
use crate::model_family::ModelFamily;
use crate::protocol::RateLimitSnapshot;
use crate::protocol::TokenUsage;
use crate::protocol::TokenUsageInfo;
use crate::tools::spec::ToolsConfig;
use crate::tools::spec::ToolsConfigParams;
use crate::tools::spec::build_specs;
use crate::tools::spec::tools_metadata_for_prompt;
/// Persistent, session-scoped state previously stored directly on `Session`.
pub(crate) struct SessionState {
pub(crate) session_configuration: SessionConfiguration,
pub(crate) history: ConversationHistory,
pub(crate) latest_rate_limits: Option<RateLimitSnapshot>,
pub(crate) model_family: ModelFamily,
}
impl SessionState {
/// Create a new session state mirroring previous `State::default()` semantics.
pub(crate) fn new(session_configuration: SessionConfiguration) -> Self {
pub(crate) fn new(
session_configuration: SessionConfiguration,
model_family: ModelFamily,
) -> Self {
Self {
session_configuration,
history: ConversationHistory::new(),
latest_rate_limits: None,
model_family,
}
}
@@ -42,6 +57,16 @@ impl SessionState {
self.history.replace(items);
}
pub(crate) fn reset_responses_api_chain(&mut self) {
self.history.reset_responses_api_chain();
}
pub(crate) fn set_responses_api_chain(&mut self, chain: ResponsesApiChainState) {
if self.session_configuration.features.enabled(Feature::ResponsesApiChaining) {
self.history.set_responses_api_chain(chain);
}
}
// Token/rate limit helpers
pub(crate) fn update_token_info_from_usage(
&mut self,
@@ -68,4 +93,84 @@ impl SessionState {
pub(crate) fn set_token_usage_full(&mut self, context_window: i64) {
self.history.set_token_usage_full(context_window);
}
pub(crate) fn prompt_for_turn(&mut self) -> Prompt {
let tools_config = ToolsConfig::new(&ToolsConfigParams {
model_family: &self.model_family,
features: &self.session_configuration.features,
});
let (tool_specs, _registry) = build_specs(&tools_config, None).build();
let tool_specs = tool_specs.into_iter().map(|c| c.spec).collect::<Vec<_>>();
let prompt_items = self.history.get_history_for_prompt();
let chain_state = self.history.responses_api_chain();
let (mut prompt, reset_chain) = build_prompt_from_items(prompt_items, chain_state.as_ref());
if reset_chain {
self.reset_responses_api_chain();
}
// Populate prompt fields that depend only on session state.
let (tools_json, has_freeform_apply_patch) =
tools_metadata_for_prompt(&tool_specs).unwrap_or_default();
format_prompt_items(&mut prompt.input, has_freeform_apply_patch);
let apply_patch_present = tool_specs.iter().any(|spec| spec.name() == "apply_patch");
let base_override = self.session_configuration.base_instructions.as_deref();
let instructions =
compute_full_instructions(base_override, &self.model_family, apply_patch_present)
.into_owned();
prompt.instructions = instructions;
prompt.tools = tools_json;
prompt.parallel_tool_calls = self.model_family.supports_parallel_tool_calls;
prompt
}
}
pub(crate) fn response_item_id(item: &ResponseItem) -> Option<&str> {
match item {
ResponseItem::Message { id: Some(id), .. }
| ResponseItem::Reasoning { id, .. }
| ResponseItem::LocalShellCall { id: Some(id), .. }
| ResponseItem::FunctionCall { id: Some(id), .. }
| ResponseItem::CustomToolCall { id: Some(id), .. }
| ResponseItem::WebSearchCall { id: Some(id), .. } => Some(id.as_str()),
_ => None,
}
}
pub(crate) fn build_prompt_from_items(
prompt_items: Vec<ResponseItem>,
chain_state: Option<&ResponsesApiChainState>,
) -> (Prompt, bool) {
let mut prompt = Prompt {
..Prompt::default()
};
if let Some(state) = chain_state {
if let Some(last_message_id) = state.last_message_id.as_ref() {
if let Some(position) = prompt_items
.iter()
.position(|item| response_item_id(item) == Some(last_message_id.as_str()))
{
if let Some(previous_response_id) = state.last_response_id.clone() {
prompt.previous_response_id = Some(previous_response_id);
}
prompt.input = prompt_items.into_iter().skip(position + 1).collect();
return (prompt, false);
}
prompt.input = prompt_items;
return (prompt, true);
}
if let Some(previous_response_id) = state.last_response_id.clone() {
prompt.previous_response_id = Some(previous_response_id);
}
prompt.input = prompt_items;
return (prompt, false);
}
prompt.input = prompt_items;
(prompt, false)
}

View File

@@ -37,16 +37,6 @@ pub(crate) enum TaskKind {
Compact,
}
impl TaskKind {
pub(crate) fn header_value(self) -> &'static str {
match self {
TaskKind::Regular => "standard",
TaskKind::Review => "review",
TaskKind::Compact => "compact",
}
}
}
#[derive(Clone)]
pub(crate) struct RunningTask {
pub(crate) done: Arc<Notify>,
@@ -123,15 +113,3 @@ impl ActiveTurn {
ts.clear_pending();
}
}
#[cfg(test)]
mod tests {
use super::TaskKind;
#[test]
fn header_value_matches_expected_labels() {
assert_eq!(TaskKind::Regular.header_value(), "standard");
assert_eq!(TaskKind::Review.header_value(), "review");
assert_eq!(TaskKind::Compact.header_value(), "compact");
}
}

View File

@@ -16,6 +16,7 @@ use tokio_util::task::AbortOnDropHandle;
use tracing::trace;
use tracing::warn;
use crate::AuthManager;
use crate::codex::Session;
use crate::codex::TurnContext;
use crate::protocol::EventMsg;
@@ -50,6 +51,10 @@ impl SessionTaskContext {
pub(crate) fn clone_session(&self) -> Arc<Session> {
Arc::clone(&self.session)
}
pub(crate) fn auth_manager(&self) -> Arc<AuthManager> {
Arc::clone(&self.session.services.auth_manager)
}
}
/// Async task that drives a [`Session`] turn.
@@ -123,7 +128,7 @@ impl Session {
task_cancellation_token.child_token(),
)
.await;
session_ctx.clone_session().flush_rollout().await;
if !task_cancellation_token.is_cancelled() {
// Emit completion uniformly from spawn site so all tasks share the same lifecycle.
let sess = session_ctx.clone_session();

View File

@@ -28,6 +28,6 @@ impl SessionTask for RegularTask {
cancellation_token: CancellationToken,
) -> Option<String> {
let sess = session.clone_session();
run_task(sess, ctx, input, TaskKind::Regular, cancellation_token).await
run_task(sess, ctx, input, cancellation_token).await
}
}

View File

@@ -1,11 +1,22 @@
use std::sync::Arc;
use async_trait::async_trait;
use codex_protocol::items::TurnItem;
use codex_protocol::models::ContentItem;
use codex_protocol::models::ResponseItem;
use codex_protocol::protocol::AgentMessageContentDeltaEvent;
use codex_protocol::protocol::AgentMessageDeltaEvent;
use codex_protocol::protocol::Event;
use codex_protocol::protocol::EventMsg;
use codex_protocol::protocol::ExitedReviewModeEvent;
use codex_protocol::protocol::ItemCompletedEvent;
use codex_protocol::protocol::ReviewOutputEvent;
use tokio_util::sync::CancellationToken;
use crate::codex::Session;
use crate::codex::TurnContext;
use crate::codex::exit_review_mode;
use crate::codex::run_task;
use crate::codex_delegate::run_codex_conversation_one_shot;
use crate::review_format::format_review_findings_block;
use crate::state::TaskKind;
use codex_protocol::user_input::UserInput;
@@ -28,11 +39,172 @@ impl SessionTask for ReviewTask {
input: Vec<UserInput>,
cancellation_token: CancellationToken,
) -> Option<String> {
let sess = session.clone_session();
run_task(sess, ctx, input, TaskKind::Review, cancellation_token).await
// Start sub-codex conversation and get the receiver for events.
let output = match start_review_conversation(
session.clone(),
ctx.clone(),
input,
cancellation_token.clone(),
)
.await
{
Some(receiver) => process_review_events(session.clone(), ctx.clone(), receiver).await,
None => None,
};
if !cancellation_token.is_cancelled() {
exit_review_mode(session.clone_session(), output.clone(), ctx.clone()).await;
}
None
}
async fn abort(&self, session: Arc<SessionTaskContext>, ctx: Arc<TurnContext>) {
exit_review_mode(session.clone_session(), ctx, None).await;
exit_review_mode(session.clone_session(), None, ctx).await;
}
}
async fn start_review_conversation(
session: Arc<SessionTaskContext>,
ctx: Arc<TurnContext>,
input: Vec<UserInput>,
cancellation_token: CancellationToken,
) -> Option<async_channel::Receiver<Event>> {
let config = ctx.client.config();
let mut sub_agent_config = config.as_ref().clone();
// Run with only reviewer rubric — drop outer user_instructions
sub_agent_config.user_instructions = None;
// Avoid loading project docs; reviewer only needs findings
sub_agent_config.project_doc_max_bytes = 0;
// Carry over review-only feature restrictions so the delegate cannot
// re-enable blocked tools (web search, view image, streamable shell).
sub_agent_config
.features
.disable(crate::features::Feature::WebSearchRequest)
.disable(crate::features::Feature::ViewImageTool)
.disable(crate::features::Feature::StreamableShell);
// Set explicit review rubric for the sub-agent
sub_agent_config.base_instructions = Some(crate::REVIEW_PROMPT.to_string());
(run_codex_conversation_one_shot(
sub_agent_config,
session.auth_manager(),
input,
session.clone_session(),
ctx.clone(),
cancellation_token,
None,
)
.await)
.ok()
.map(|io| io.rx_event)
}
async fn process_review_events(
session: Arc<SessionTaskContext>,
ctx: Arc<TurnContext>,
receiver: async_channel::Receiver<Event>,
) -> Option<ReviewOutputEvent> {
let mut prev_agent_message: Option<Event> = None;
while let Ok(event) = receiver.recv().await {
match event.clone().msg {
EventMsg::AgentMessage(_) => {
if let Some(prev) = prev_agent_message.take() {
session
.clone_session()
.send_event(ctx.as_ref(), prev.msg)
.await;
}
prev_agent_message = Some(event);
}
// Suppress ItemCompleted only for assistant messages: forwarding it
// would trigger legacy AgentMessage via as_legacy_events(), which this
// review flow intentionally hides in favor of structured output.
EventMsg::ItemCompleted(ItemCompletedEvent {
item: TurnItem::AgentMessage(_),
..
})
| EventMsg::AgentMessageDelta(AgentMessageDeltaEvent { .. })
| EventMsg::AgentMessageContentDelta(AgentMessageContentDeltaEvent { .. }) => {}
EventMsg::TaskComplete(task_complete) => {
// Parse review output from the last agent message (if present).
let out = task_complete
.last_agent_message
.as_deref()
.map(parse_review_output_event);
return out;
}
EventMsg::TurnAborted(_) => {
// Cancellation or abort: consumer will finalize with None.
return None;
}
other => {
session
.clone_session()
.send_event(ctx.as_ref(), other)
.await;
}
}
}
// Channel closed without TaskComplete: treat as interrupted.
None
}
/// Parse a ReviewOutputEvent from a text blob returned by the reviewer model.
/// If the text is valid JSON matching ReviewOutputEvent, deserialize it.
/// Otherwise, attempt to extract the first JSON object substring and parse it.
/// If parsing still fails, return a structured fallback carrying the plain text
/// in `overall_explanation`.
fn parse_review_output_event(text: &str) -> ReviewOutputEvent {
if let Ok(ev) = serde_json::from_str::<ReviewOutputEvent>(text) {
return ev;
}
if let (Some(start), Some(end)) = (text.find('{'), text.rfind('}'))
&& start < end
&& let Some(slice) = text.get(start..=end)
&& let Ok(ev) = serde_json::from_str::<ReviewOutputEvent>(slice)
{
return ev;
}
ReviewOutputEvent {
overall_explanation: text.to_string(),
..Default::default()
}
}
/// Emits an ExitedReviewMode Event with optional ReviewOutput,
/// and records a developer message with the review output.
pub(crate) async fn exit_review_mode(
session: Arc<Session>,
review_output: Option<ReviewOutputEvent>,
ctx: Arc<TurnContext>,
) {
let user_message = if let Some(out) = review_output.clone() {
let mut findings_str = String::new();
let text = out.overall_explanation.trim();
if !text.is_empty() {
findings_str.push_str(text);
}
if !out.findings.is_empty() {
let block = format_review_findings_block(&out.findings, None);
findings_str.push_str(&format!("\n{block}"));
}
crate::client_common::REVIEW_EXIT_SUCCESS_TMPL.replace("{results}", &findings_str)
} else {
crate::client_common::REVIEW_EXIT_INTERRUPTED_TMPL.to_string()
};
session
.record_conversation_items(
&ctx,
&[ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText { text: user_message }],
}],
)
.await;
session
.send_event(
ctx.as_ref(),
EventMsg::ExitedReviewMode(ExitedReviewModeEvent { review_output }),
)
.await;
}

View File

@@ -274,31 +274,32 @@ impl ToolEmitter {
ctx: ToolEventCtx<'_>,
out: Result<ExecToolCallOutput, ToolError>,
) -> Result<String, FunctionCallError> {
let event;
let result = match out {
let (event, result) = match out {
Ok(output) => {
let content = super::format_exec_output_for_model(&output);
let exit_code = output.exit_code;
event = ToolEventStage::Success(output);
if exit_code == 0 {
let event = ToolEventStage::Success(output);
let result = if exit_code == 0 {
Ok(content)
} else {
Err(FunctionCallError::RespondToModel(content))
}
};
(event, result)
}
Err(ToolError::Codex(CodexErr::Sandbox(SandboxErr::Timeout { output })))
| Err(ToolError::Codex(CodexErr::Sandbox(SandboxErr::Denied { output }))) => {
let response = super::format_exec_output_for_model(&output);
event = ToolEventStage::Failure(ToolEventFailure::Output(*output));
Err(FunctionCallError::RespondToModel(response))
let event = ToolEventStage::Failure(ToolEventFailure::Output(*output));
let result = Err(FunctionCallError::RespondToModel(response));
(event, result)
}
Err(ToolError::Codex(err)) => {
let message = format!("execution error: {err:?}");
let response = message.clone();
event = ToolEventStage::Failure(ToolEventFailure::Message(message));
Err(FunctionCallError::RespondToModel(response))
let event = ToolEventStage::Failure(ToolEventFailure::Message(message.clone()));
let result = Err(FunctionCallError::RespondToModel(message));
(event, result)
}
Err(ToolError::Rejected(msg)) | Err(ToolError::SandboxDenied(msg)) => {
Err(ToolError::Rejected(msg)) => {
// Normalize common rejection messages for exec tools so tests and
// users see a clear, consistent phrase.
let normalized = if msg == "rejected by user" {
@@ -306,9 +307,9 @@ impl ToolEmitter {
} else {
msg
};
let response = &normalized;
event = ToolEventStage::Failure(ToolEventFailure::Message(normalized.clone()));
Err(FunctionCallError::RespondToModel(response.clone()))
let event = ToolEventStage::Failure(ToolEventFailure::Message(normalized.clone()));
let result = Err(FunctionCallError::RespondToModel(normalized));
(event, result)
}
};
self.emit(ctx, event).await;

View File

@@ -3,10 +3,6 @@ use std::collections::BTreeMap;
use crate::apply_patch;
use crate::apply_patch::InternalApplyPatchInvocation;
use crate::apply_patch::convert_apply_patch_to_protocol;
use crate::client_common::tools::FreeformTool;
use crate::client_common::tools::FreeformToolFormat;
use crate::client_common::tools::ResponsesApiTool;
use crate::client_common::tools::ToolSpec;
use crate::function_tool::FunctionCallError;
use crate::tools::context::ToolInvocation;
use crate::tools::context::ToolOutput;
@@ -20,7 +16,11 @@ use crate::tools::runtimes::apply_patch::ApplyPatchRequest;
use crate::tools::runtimes::apply_patch::ApplyPatchRuntime;
use crate::tools::sandboxing::ToolCtx;
use crate::tools::spec::ApplyPatchToolArgs;
use crate::tools::spec::FreeformTool;
use crate::tools::spec::FreeformToolFormat;
use crate::tools::spec::JsonSchema;
use crate::tools::spec::ResponsesApiTool;
use crate::tools::spec::ToolSpec;
use async_trait::async_trait;
use serde::Deserialize;
use serde::Serialize;

View File

@@ -1,5 +1,3 @@
use crate::client_common::tools::ResponsesApiTool;
use crate::client_common::tools::ToolSpec;
use crate::codex::Session;
use crate::codex::TurnContext;
use crate::function_tool::FunctionCallError;
@@ -9,6 +7,8 @@ use crate::tools::context::ToolPayload;
use crate::tools::registry::ToolHandler;
use crate::tools::registry::ToolKind;
use crate::tools::spec::JsonSchema;
use crate::tools::spec::ResponsesApiTool;
use crate::tools::spec::ToolSpec;
use async_trait::async_trait;
use codex_protocol::plan_tool::UpdatePlanArgs;
use codex_protocol::protocol::EventMsg;

View File

@@ -98,15 +98,16 @@ impl ToolOrchestrator {
}
Err(ToolError::Codex(CodexErr::Sandbox(SandboxErr::Denied { output }))) => {
if !tool.escalate_on_failure() {
return Err(ToolError::SandboxDenied(
"sandbox denied and no retry".to_string(),
));
return Err(ToolError::Codex(CodexErr::Sandbox(SandboxErr::Denied {
output,
})));
}
// Under `Never` or `OnRequest`, do not retry without sandbox; surface a concise message
// derived from the actual output (platform-agnostic).
// Under `Never` or `OnRequest`, do not retry without sandbox; surface a concise
// sandbox denial that preserves the original output.
if !tool.wants_no_sandbox_approval(approval_policy) {
let msg = build_never_denied_message_from_output(output.as_ref());
return Err(ToolError::SandboxDenied(msg));
return Err(ToolError::Codex(CodexErr::Sandbox(SandboxErr::Denied {
output,
})));
}
// Ask for approval before retrying without sandbox.
@@ -167,29 +168,6 @@ impl ToolOrchestrator {
}
}
fn build_never_denied_message_from_output(output: &ExecToolCallOutput) -> String {
let body = format!(
"{}\n{}\n{}",
output.stderr.text, output.stdout.text, output.aggregated_output.text
)
.to_lowercase();
let detail = if body.contains("permission denied") {
Some("Permission denied")
} else if body.contains("operation not permitted") {
Some("Operation not permitted")
} else if body.contains("read-only file system") {
Some("Read-only file system")
} else {
None
};
match detail {
Some(tag) => format!("failed in sandbox: {tag}"),
None => "failed in sandbox".to_string(),
}
}
fn build_denial_reason_from_output(_output: &ExecToolCallOutput) -> String {
// Keep approval reason terse and stable for UX/tests, but accept the
// output so we can evolve heuristics later without touching call sites.

View File

@@ -6,11 +6,11 @@ use async_trait::async_trait;
use codex_protocol::models::ResponseInputItem;
use tracing::warn;
use crate::client_common::tools::ToolSpec;
use crate::function_tool::FunctionCallError;
use crate::tools::context::ToolInvocation;
use crate::tools::context::ToolOutput;
use crate::tools::context::ToolPayload;
use crate::tools::spec::ToolSpec;
#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
pub enum ToolKind {

View File

@@ -1,7 +1,6 @@
use std::collections::HashMap;
use std::sync::Arc;
use crate::client_common::tools::ToolSpec;
use crate::codex::Session;
use crate::codex::TurnContext;
use crate::function_tool::FunctionCallError;
@@ -10,6 +9,7 @@ use crate::tools::context::ToolInvocation;
use crate::tools::context::ToolPayload;
use crate::tools::registry::ConfiguredToolSpec;
use crate::tools::registry::ToolRegistry;
use crate::tools::spec::ToolSpec;
use crate::tools::spec::ToolsConfig;
use crate::tools::spec::build_specs;
use codex_protocol::models::LocalShellAction;

View File

@@ -173,7 +173,6 @@ pub(crate) trait ProvidesSandboxRetryData {
#[derive(Debug)]
pub(crate) enum ToolError {
Rejected(String),
SandboxDenied(String),
Codex(CodexErr),
}

View File

@@ -1,5 +1,3 @@
use crate::client_common::tools::ResponsesApiTool;
use crate::client_common::tools::ToolSpec;
use crate::features::Feature;
use crate::features::Features;
use crate::model_family::ModelFamily;
@@ -22,6 +20,52 @@ pub enum ConfigShellToolType {
Streamable,
}
#[derive(Debug, Clone, Serialize, PartialEq)]
#[serde(tag = "type")]
pub(crate) enum ToolSpec {
#[serde(rename = "function")]
Function(ResponsesApiTool),
#[serde(rename = "local_shell")]
LocalShell {},
#[serde(rename = "web_search")]
WebSearch {},
#[serde(rename = "custom")]
Freeform(FreeformTool),
}
impl ToolSpec {
pub(crate) fn name(&self) -> &str {
match self {
ToolSpec::Function(tool) => tool.name.as_str(),
ToolSpec::LocalShell {} => "local_shell",
ToolSpec::WebSearch {} => "web_search",
ToolSpec::Freeform(tool) => tool.name.as_str(),
}
}
}
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct FreeformTool {
pub(crate) name: String,
pub(crate) description: String,
pub(crate) format: FreeformToolFormat,
}
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
pub struct FreeformToolFormat {
pub(crate) r#type: String,
pub(crate) syntax: String,
pub(crate) definition: String,
}
#[derive(Debug, Clone, Serialize, PartialEq)]
pub struct ResponsesApiTool {
pub(crate) name: String,
pub(crate) description: String,
pub(crate) strict: bool,
pub(crate) parameters: JsonSchema,
}
#[derive(Debug, Clone)]
pub(crate) struct ToolsConfig {
pub shell_type: ConfigShellToolType,
@@ -666,9 +710,6 @@ pub(crate) struct ApplyPatchToolArgs {
pub(crate) input: String,
}
/// Returns JSON values that are compatible with Function Calling in the
/// Responses API:
/// https://platform.openai.com/docs/guides/function-calling?api-mode=responses
pub fn create_tools_json_for_responses_api(
tools: &[ToolSpec],
) -> crate::error::Result<Vec<serde_json::Value>> {
@@ -681,35 +722,16 @@ pub fn create_tools_json_for_responses_api(
Ok(tools_json)
}
/// Returns JSON values that are compatible with Function Calling in the
/// Chat Completions API:
/// https://platform.openai.com/docs/guides/function-calling?api-mode=chat
pub(crate) fn create_tools_json_for_chat_completions_api(
tools: &[ToolSpec],
) -> crate::error::Result<Vec<serde_json::Value>> {
// We start with the JSON for the Responses API and than rewrite it to match
// the chat completions tool call format.
let responses_api_tools_json = create_tools_json_for_responses_api(tools)?;
let tools_json = responses_api_tools_json
.into_iter()
.filter_map(|mut tool| {
if tool.get("type") != Some(&serde_json::Value::String("function".to_string())) {
return None;
}
if let Some(map) = tool.as_object_mut() {
// Remove "type" field as it is not needed in chat completions.
map.remove("type");
Some(json!({
"type": "function",
"function": map,
}))
} else {
None
}
})
.collect::<Vec<serde_json::Value>>();
Ok(tools_json)
pub fn tools_metadata_for_prompt(
tools: &[ToolSpec],
) -> crate::error::Result<(Vec<serde_json::Value>, bool)> {
let tools_json = create_tools_json_for_responses_api(tools)?;
let has_freeform_apply_patch = tools.iter().any(|tool| match tool {
ToolSpec::Freeform(freeform) => freeform.name == "apply_patch",
_ => false,
});
Ok((tools_json, has_freeform_apply_patch))
}
pub(crate) fn mcp_tool_to_openai_tool(
@@ -1002,7 +1024,6 @@ pub(crate) fn build_specs(
#[cfg(test)]
mod tests {
use crate::client_common::tools::FreeformTool;
use crate::model_family::find_family_for_model;
use crate::tools::registry::ConfiguredToolSpec;
use mcp_types::ToolInputSchema;

View File

@@ -40,3 +40,31 @@ impl From<UserInstructions> for ResponseItem {
}
}
}
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
#[serde(rename = "developer_instructions", rename_all = "snake_case")]
pub(crate) struct DeveloperInstructions {
text: String,
}
impl DeveloperInstructions {
pub fn new<T: Into<String>>(text: T) -> Self {
Self { text: text.into() }
}
pub fn into_text(self) -> String {
self.text
}
}
impl From<DeveloperInstructions> for ResponseItem {
fn from(di: DeveloperInstructions) -> Self {
ResponseItem::Message {
id: None,
role: "developer".to_string(),
content: vec![ContentItem::InputText {
text: di.into_text(),
}],
}
}
}

View File

@@ -2,6 +2,7 @@ use std::time::Duration;
use rand::Rng;
use tracing::debug;
use tracing::error;
const INITIAL_DELAY_MS: u64 = 200;
const BACKOFF_FACTOR: f64 = 2.0;
@@ -13,6 +14,14 @@ pub(crate) fn backoff(attempt: u64) -> Duration {
Duration::from_millis((base as f64 * jitter) as u64)
}
pub(crate) fn error_or_panic(message: String) {
if cfg!(debug_assertions) || env!("CARGO_PKG_VERSION").contains("alpha") {
panic!("{message}");
} else {
error!("{message}");
}
}
pub(crate) fn try_parse_error_message(text: &str) -> String {
debug!("Parsing server error response: {}", text);
let json = serde_json::from_str::<serde_json::Value>(text).unwrap_or_default();

View File

@@ -0,0 +1,8 @@
<user_action>
<context>User initiated a review task, but was interrupted. If user asks about this, tell them to re-initiate a review with `/review` and wait for it to complete.</context>
<action>review</action>
<results>
None.
</results>
</user_action>

View File

@@ -0,0 +1,8 @@
<user_action>
<context>User initiated a review task. Here's the full review output from reviewer model. User may select one or more comments to resolve.</context>
<action>review</action>
<results>
{results}
</results>
</user_action>

View File

@@ -0,0 +1,8 @@
<user_action>
<context>User initiated a review task. Here's the full review output from reviewer model. User may select one or more comments to resolve.</context>
<action>review</action>
<results>
{findings}
</results>
</user_action>

View File

@@ -0,0 +1,8 @@
<user_action>
<context>User initiated a review task, but was interrupted. If user asks about this, tell them to re-initiate a review with `/review` and wait for it to complete.</context>
<action>review</action>
<results>
None.
</results>
</user_action>

View File

@@ -1,15 +1,15 @@
use std::sync::Arc;
use codex_api_client::ModelProviderInfo;
use codex_api_client::WireApi;
use codex_app_server_protocol::AuthMode;
use codex_core::ContentItem;
use codex_core::LocalShellAction;
use codex_core::LocalShellExecAction;
use codex_core::LocalShellStatus;
use codex_core::ModelClient;
use codex_core::ModelProviderInfo;
use codex_core::Prompt;
use codex_core::ResponseItem;
use codex_core::WireApi;
use codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR;
use codex_otel::otel_event_manager::OtelEventManager;
use codex_protocol::ConversationId;
@@ -94,12 +94,15 @@ async fn run_request(input: Vec<ResponseItem>) -> Value {
effort,
summary,
conversation_id,
codex_protocol::protocol::SessionSource::Exec,
);
let mut prompt = Prompt::default();
prompt.input = input;
let prompt = Prompt {
input,
..Prompt::default()
};
let mut stream = match client.stream(&prompt).await {
let mut stream = match client.stream_for_test(prompt).await {
Ok(s) => s,
Err(e) => panic!("stream chat failed: {e}"),
};

View File

@@ -2,14 +2,14 @@ use assert_matches::assert_matches;
use std::sync::Arc;
use tracing_test::traced_test;
use codex_api_client::ModelProviderInfo;
use codex_api_client::WireApi;
use codex_app_server_protocol::AuthMode;
use codex_core::ContentItem;
use codex_core::ModelClient;
use codex_core::ModelProviderInfo;
use codex_core::Prompt;
use codex_core::ResponseEvent;
use codex_core::ResponseItem;
use codex_core::WireApi;
use codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR;
use codex_otel::otel_event_manager::OtelEventManager;
use codex_protocol::ConversationId;
@@ -94,18 +94,21 @@ async fn run_stream_with_bytes(sse_body: &[u8]) -> Vec<ResponseEvent> {
effort,
summary,
conversation_id,
codex_protocol::protocol::SessionSource::Exec,
);
let mut prompt = Prompt::default();
prompt.input = vec![ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: "hello".to_string(),
let prompt = Prompt {
input: vec![ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: "hello".to_string(),
}],
}],
}];
..Prompt::default()
};
let mut stream = match client.stream(&prompt).await {
let mut stream = match client.stream_for_test(prompt).await {
Ok(s) => s,
Err(e) => panic!("stream chat failed: {e}"),
};
@@ -170,19 +173,24 @@ async fn streams_text_without_reasoning() {
);
let events = run_stream(sse).await;
assert_eq!(events.len(), 3, "unexpected events: {events:?}");
assert_eq!(events.len(), 4, "unexpected events: {events:?}");
match &events[0] {
ResponseEvent::OutputItemAdded(ResponseItem::Message { .. }) => {}
other => panic!("expected initial assistant item, got {other:?}"),
}
match &events[1] {
ResponseEvent::OutputTextDelta(text) => assert_eq!(text, "hi"),
other => panic!("expected text delta, got {other:?}"),
}
match &events[1] {
match &events[2] {
ResponseEvent::OutputItemDone(item) => assert_message(item, "hi"),
other => panic!("expected terminal message, got {other:?}"),
}
assert_matches!(events[2], ResponseEvent::Completed { .. });
assert_matches!(events[3], ResponseEvent::Completed { .. });
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
@@ -201,29 +209,39 @@ async fn streams_reasoning_from_string_delta() {
);
let events = run_stream(sse).await;
assert_eq!(events.len(), 5, "unexpected events: {events:?}");
assert_eq!(events.len(), 7, "unexpected events: {events:?}");
match &events[0] {
ResponseEvent::OutputItemAdded(ResponseItem::Reasoning { .. }) => {}
other => panic!("expected initial reasoning item, got {other:?}"),
}
match &events[1] {
ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "think1"),
other => panic!("expected reasoning delta, got {other:?}"),
}
match &events[1] {
match &events[2] {
ResponseEvent::OutputItemAdded(ResponseItem::Message { .. }) => {}
other => panic!("expected initial message item, got {other:?}"),
}
match &events[3] {
ResponseEvent::OutputTextDelta(text) => assert_eq!(text, "ok"),
other => panic!("expected text delta, got {other:?}"),
}
match &events[2] {
match &events[4] {
ResponseEvent::OutputItemDone(item) => assert_reasoning(item, "think1"),
other => panic!("expected reasoning item, got {other:?}"),
other => panic!("expected terminal reasoning, got {other:?}"),
}
match &events[3] {
match &events[5] {
ResponseEvent::OutputItemDone(item) => assert_message(item, "ok"),
other => panic!("expected message item, got {other:?}"),
other => panic!("expected terminal message, got {other:?}"),
}
assert_matches!(events[4], ResponseEvent::Completed { .. });
assert_matches!(events[6], ResponseEvent::Completed { .. });
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
@@ -243,34 +261,44 @@ async fn streams_reasoning_from_object_delta() {
);
let events = run_stream(sse).await;
assert_eq!(events.len(), 6, "unexpected events: {events:?}");
assert_eq!(events.len(), 8, "unexpected events: {events:?}");
match &events[0] {
ResponseEvent::OutputItemAdded(ResponseItem::Reasoning { .. }) => {}
other => panic!("expected initial reasoning item, got {other:?}"),
}
match &events[1] {
ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "partA"),
other => panic!("expected reasoning delta, got {other:?}"),
}
match &events[1] {
match &events[2] {
ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "partB"),
other => panic!("expected reasoning delta, got {other:?}"),
}
match &events[2] {
match &events[3] {
ResponseEvent::OutputItemAdded(ResponseItem::Message { .. }) => {}
other => panic!("expected initial message item, got {other:?}"),
}
match &events[4] {
ResponseEvent::OutputTextDelta(text) => assert_eq!(text, "answer"),
other => panic!("expected text delta, got {other:?}"),
}
match &events[3] {
match &events[5] {
ResponseEvent::OutputItemDone(item) => assert_reasoning(item, "partApartB"),
other => panic!("expected reasoning item, got {other:?}"),
other => panic!("expected terminal reasoning, got {other:?}"),
}
match &events[4] {
match &events[6] {
ResponseEvent::OutputItemDone(item) => assert_message(item, "answer"),
other => panic!("expected message item, got {other:?}"),
other => panic!("expected terminal message, got {other:?}"),
}
assert_matches!(events[5], ResponseEvent::Completed { .. });
assert_matches!(events[7], ResponseEvent::Completed { .. });
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
@@ -285,19 +313,24 @@ async fn streams_reasoning_from_final_message() {
let sse = "data: {\"choices\":[{\"message\":{\"reasoning\":\"final-cot\"},\"finish_reason\":\"stop\"}]}\n\n";
let events = run_stream(sse).await;
assert_eq!(events.len(), 3, "unexpected events: {events:?}");
assert_eq!(events.len(), 4, "unexpected events: {events:?}");
match &events[0] {
ResponseEvent::OutputItemAdded(ResponseItem::Reasoning { .. }) => {}
other => panic!("expected initial reasoning item, got {other:?}"),
}
match &events[1] {
ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "final-cot"),
other => panic!("expected reasoning delta, got {other:?}"),
}
match &events[1] {
match &events[2] {
ResponseEvent::OutputItemDone(item) => assert_reasoning(item, "final-cot"),
other => panic!("expected reasoning item, got {other:?}"),
}
assert_matches!(events[2], ResponseEvent::Completed { .. });
assert_matches!(events[3], ResponseEvent::Completed { .. });
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
@@ -315,19 +348,24 @@ async fn streams_reasoning_before_tool_call() {
);
let events = run_stream(sse).await;
assert_eq!(events.len(), 4, "unexpected events: {events:?}");
assert_eq!(events.len(), 5, "unexpected events: {events:?}");
match &events[0] {
ResponseEvent::OutputItemAdded(ResponseItem::Reasoning { .. }) => {}
other => panic!("expected initial reasoning item, got {other:?}"),
}
match &events[1] {
ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "pre-tool"),
other => panic!("expected reasoning delta, got {other:?}"),
}
match &events[1] {
match &events[2] {
ResponseEvent::OutputItemDone(item) => assert_reasoning(item, "pre-tool"),
other => panic!("expected reasoning item, got {other:?}"),
}
match &events[2] {
match &events[3] {
ResponseEvent::OutputItemDone(ResponseItem::FunctionCall {
name,
arguments,
@@ -341,7 +379,7 @@ async fn streams_reasoning_before_tool_call() {
other => panic!("expected function call, got {other:?}"),
}
assert_matches!(events[3], ResponseEvent::Completed { .. });
assert_matches!(events[4], ResponseEvent::Completed { .. });
}
#[tokio::test]

View File

@@ -9,6 +9,7 @@ path = "lib.rs"
[dependencies]
anyhow = { workspace = true }
assert_cmd = { workspace = true }
codex-api-client = { workspace = true }
codex-core = { workspace = true }
codex-protocol = { workspace = true }
notify = { workspace = true }

View File

@@ -217,6 +217,25 @@ pub fn ev_assistant_message(id: &str, text: &str) -> Value {
})
}
pub fn ev_message_item_added(id: &str, text: &str) -> Value {
serde_json::json!({
"type": "response.output_item.added",
"item": {
"type": "message",
"role": "assistant",
"id": id,
"content": [{"type": "output_text", "text": text}]
}
})
}
pub fn ev_output_text_delta(delta: &str) -> Value {
serde_json::json!({
"type": "response.output_text.delta",
"delta": delta,
})
}
pub fn ev_reasoning_item(id: &str, summary: &[&str], raw_content: &[&str]) -> Value {
let summary_entries: Vec<Value> = summary
.iter()
@@ -243,6 +262,36 @@ pub fn ev_reasoning_item(id: &str, summary: &[&str], raw_content: &[&str]) -> Va
event
}
pub fn ev_reasoning_item_added(id: &str, summary: &[&str]) -> Value {
let summary_entries: Vec<Value> = summary
.iter()
.map(|text| serde_json::json!({"type": "summary_text", "text": text}))
.collect();
serde_json::json!({
"type": "response.output_item.added",
"item": {
"type": "reasoning",
"id": id,
"summary": summary_entries,
}
})
}
pub fn ev_reasoning_summary_text_delta(delta: &str) -> Value {
serde_json::json!({
"type": "response.reasoning_summary_text.delta",
"delta": delta,
})
}
pub fn ev_reasoning_text_delta(delta: &str) -> Value {
serde_json::json!({
"type": "response.reasoning_text.delta",
"delta": delta,
})
}
pub fn ev_web_search_call_added(id: &str, status: &str, query: &str) -> Value {
serde_json::json!({
"type": "response.output_item.added",

View File

@@ -4,11 +4,11 @@ use std::path::PathBuf;
use std::sync::Arc;
use anyhow::Result;
use codex_api_client::ModelProviderInfo;
use codex_api_client::built_in_model_providers;
use codex_core::CodexAuth;
use codex_core::CodexConversation;
use codex_core::ConversationManager;
use codex_core::ModelProviderInfo;
use codex_core::built_in_model_providers;
use codex_core::config::Config;
use codex_core::features::Feature;
use codex_core::protocol::AskForApproval;
@@ -240,6 +240,30 @@ impl TestCodexHarness {
.expect("output string")
.to_string()
}
pub async fn custom_tool_call_output(&self, call_id: &str) -> String {
let bodies = self.request_bodies().await;
custom_tool_call_output(&bodies, call_id)
.get("output")
.and_then(Value::as_str)
.expect("output string")
.to_string()
}
}
fn custom_tool_call_output<'a>(bodies: &'a [Value], call_id: &str) -> &'a Value {
for body in bodies {
if let Some(items) = body.get("input").and_then(Value::as_array) {
for item in items {
if item.get("type").and_then(Value::as_str) == Some("custom_tool_call_output")
&& item.get("call_id").and_then(Value::as_str) == Some(call_id)
{
return item;
}
}
}
}
panic!("custom_tool_call_output {call_id} not found");
}
fn function_call_output<'a>(bodies: &'a [Value], call_id: &str) -> &'a Value {

View File

@@ -1,15 +1,16 @@
use std::sync::Arc;
use codex_api_client::ModelProviderInfo;
use codex_api_client::WireApi;
use codex_app_server_protocol::AuthMode;
use codex_core::ContentItem;
use codex_core::ModelClient;
use codex_core::ModelProviderInfo;
use codex_core::Prompt;
use codex_core::ResponseEvent;
use codex_core::ResponseItem;
use codex_core::WireApi;
use codex_otel::otel_event_manager::OtelEventManager;
use codex_protocol::ConversationId;
use codex_protocol::protocol::SessionSource;
use core_test_support::load_default_config_for_test;
use core_test_support::responses;
use futures::StreamExt;
@@ -17,7 +18,7 @@ use tempfile::TempDir;
use wiremock::matchers::header;
#[tokio::test]
async fn responses_stream_includes_task_type_header() {
async fn responses_stream_includes_subagent_header_on_review() {
core_test_support::skip_if_no_network!();
let server = responses::start_mock_server().await;
@@ -28,7 +29,7 @@ async fn responses_stream_includes_task_type_header() {
let request_recorder = responses::mount_sse_once_match(
&server,
header("Codex-Task-Type", "standard"),
header("x-openai-subagent", "review"),
response_body,
)
.await;
@@ -78,18 +79,21 @@ async fn responses_stream_includes_task_type_header() {
effort,
summary,
conversation_id,
SessionSource::SubAgent(codex_protocol::protocol::SubAgentSource::Review),
);
let mut prompt = Prompt::default();
prompt.input = vec![ResponseItem::Message {
id: None,
role: "user".into(),
content: vec![ContentItem::InputText {
text: "hello".into(),
let prompt = Prompt {
input: vec![ResponseItem::Message {
id: None,
role: "user".into(),
content: vec![ContentItem::InputText {
text: "hello".into(),
}],
}],
}];
..Prompt::default()
};
let mut stream = client.stream(&prompt).await.expect("stream failed");
let mut stream = client.stream_for_test(prompt).await.expect("stream failed");
while let Some(event) = stream.next().await {
if matches!(event, Ok(ResponseEvent::Completed { .. })) {
break;
@@ -98,7 +102,99 @@ async fn responses_stream_includes_task_type_header() {
let request = request_recorder.single_request();
assert_eq!(
request.header("Codex-Task-Type").as_deref(),
Some("standard")
request.header("x-openai-subagent").as_deref(),
Some("review")
);
}
#[tokio::test]
async fn responses_stream_includes_subagent_header_on_other() {
core_test_support::skip_if_no_network!();
let server = responses::start_mock_server().await;
let response_body = responses::sse(vec![
responses::ev_response_created("resp-1"),
responses::ev_completed("resp-1"),
]);
let request_recorder = responses::mount_sse_once_match(
&server,
header("x-openai-subagent", "my-task"),
response_body,
)
.await;
let provider = ModelProviderInfo {
name: "mock".into(),
base_url: Some(format!("{}/v1", server.uri())),
env_key: None,
env_key_instructions: None,
experimental_bearer_token: None,
wire_api: WireApi::Responses,
query_params: None,
http_headers: None,
env_http_headers: None,
request_max_retries: Some(0),
stream_max_retries: Some(0),
stream_idle_timeout_ms: Some(5_000),
requires_openai_auth: false,
};
let codex_home = TempDir::new().expect("failed to create TempDir");
let mut config = load_default_config_for_test(&codex_home);
config.model_provider_id = provider.name.clone();
config.model_provider = provider.clone();
let effort = config.model_reasoning_effort;
let summary = config.model_reasoning_summary;
let config = Arc::new(config);
let conversation_id = ConversationId::new();
let otel_event_manager = OtelEventManager::new(
conversation_id,
config.model.as_str(),
config.model_family.slug.as_str(),
None,
Some("test@test.com".to_string()),
Some(AuthMode::ChatGPT),
false,
"test".to_string(),
);
let client = ModelClient::new(
Arc::clone(&config),
None,
otel_event_manager,
provider,
effort,
summary,
conversation_id,
SessionSource::SubAgent(codex_protocol::protocol::SubAgentSource::Other(
"my-task".to_string(),
)),
);
let prompt = Prompt {
input: vec![ResponseItem::Message {
id: None,
role: "user".into(),
content: vec![ContentItem::InputText {
text: "hello".into(),
}],
}],
..Prompt::default()
};
let mut stream = client.stream_for_test(prompt).await.expect("stream failed");
while let Some(event) = stream.next().await {
if matches!(event, Ok(ResponseEvent::Completed { .. })) {
break;
}
}
let request = request_recorder.single_request();
assert_eq!(
request.header("x-openai-subagent").as_deref(),
Some("my-task")
);
}

File diff suppressed because it is too large Load Diff

View File

@@ -268,11 +268,22 @@ impl Expectation {
"expected non-zero exit for {path:?}"
);
for needle in *message_contains {
assert!(
result.stdout.contains(needle),
"stdout missing {needle:?}: {}",
result.stdout
);
if needle.contains('|') {
let options: Vec<&str> = needle.split('|').collect();
let matches_any =
options.iter().any(|option| result.stdout.contains(option));
assert!(
matches_any,
"stdout missing one of {options:?}: {}",
result.stdout
);
} else {
assert!(
result.stdout.contains(needle),
"stdout missing {needle:?}: {}",
result.stdout
);
}
}
assert!(
!path.exists(),
@@ -901,7 +912,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
message_contains: if cfg!(target_os = "linux") {
&["Permission denied"]
} else {
&["failed in sandbox"]
&["Permission denied|Operation not permitted|Read-only file system"]
},
},
},
@@ -1045,7 +1056,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
message_contains: if cfg!(target_os = "linux") {
&["Permission denied"]
} else {
&["failed in sandbox"]
&["Permission denied|Operation not permitted|Read-only file system"]
},
},
},

View File

@@ -1,3 +1,6 @@
use codex_api_client::ModelProviderInfo;
use codex_api_client::WireApi;
use codex_api_client::built_in_model_providers;
use codex_app_server_protocol::AuthMode;
use codex_core::CodexAuth;
use codex_core::ContentItem;
@@ -6,15 +9,13 @@ use codex_core::LocalShellAction;
use codex_core::LocalShellExecAction;
use codex_core::LocalShellStatus;
use codex_core::ModelClient;
use codex_core::ModelProviderInfo;
use codex_core::NewConversation;
use codex_core::Prompt;
use codex_core::ResponseEvent;
use codex_core::ResponseItem;
use codex_core::WireApi;
use codex_core::auth::AuthCredentialsStoreMode;
use codex_core::built_in_model_providers;
use codex_core::error::CodexErr;
use codex_core::features::Feature;
use codex_core::model_family::find_family_for_model;
use codex_core::protocol::EventMsg;
use codex_core::protocol::Op;
@@ -58,6 +59,18 @@ fn assert_message_role(request_body: &serde_json::Value, role: &str) {
assert_eq!(request_body["role"].as_str().unwrap(), role);
}
#[expect(clippy::expect_used)]
fn assert_message_equals(request_body: &serde_json::Value, text: &str) {
let content = request_body["content"][0]["text"]
.as_str()
.expect("invalid message content");
assert_eq!(
content, text,
"expected message content '{content}' to equal '{text}'"
);
}
#[expect(clippy::expect_used)]
fn assert_message_starts_with(request_body: &serde_json::Value, text: &str) {
let content = request_body["content"][0]["text"]
@@ -608,6 +621,156 @@ async fn includes_user_instructions_message_in_request() {
assert_message_ends_with(&request_body["input"][1], "</environment_context>");
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn includes_developer_instructions_message_in_request() {
skip_if_no_network!();
let server = MockServer::start().await;
let resp_mock =
responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
.await;
let model_provider = ModelProviderInfo {
base_url: Some(format!("{}/v1", server.uri())),
..built_in_model_providers()["openai"].clone()
};
let codex_home = TempDir::new().unwrap();
let mut config = load_default_config_for_test(&codex_home);
config.model_provider = model_provider;
config.user_instructions = Some("be nice".to_string());
config.developer_instructions = Some("be useful".to_string());
let conversation_manager =
ConversationManager::with_auth(CodexAuth::from_api_key("Test API Key"));
let codex = conversation_manager
.new_conversation(config)
.await
.expect("create new conversation")
.conversation;
codex
.submit(Op::UserInput {
items: vec![UserInput::Text {
text: "hello".into(),
}],
})
.await
.unwrap();
wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
let request = resp_mock.single_request();
let request_body = request.body_json();
assert!(
!request_body["instructions"]
.as_str()
.unwrap()
.contains("be nice")
);
assert_message_role(&request_body["input"][0], "developer");
assert_message_equals(&request_body["input"][0], "be useful");
assert_message_role(&request_body["input"][1], "user");
assert_message_starts_with(&request_body["input"][1], "<user_instructions>");
assert_message_ends_with(&request_body["input"][1], "</user_instructions>");
assert_message_role(&request_body["input"][2], "user");
assert_message_starts_with(&request_body["input"][2], "<environment_context>");
assert_message_ends_with(&request_body["input"][2], "</environment_context>");
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn responses_api_chaining_sets_store_and_previous_id() {
skip_if_no_network!();
let server = MockServer::start().await;
let first_response = responses::sse(vec![
responses::ev_response_created("resp-first"),
responses::ev_assistant_message("m1", "hi there"),
responses::ev_completed("resp-first"),
]);
let second_response = responses::sse(vec![
responses::ev_response_created("resp-second"),
responses::ev_assistant_message("m2", "second reply"),
responses::ev_completed("resp-second"),
]);
let response_mock =
responses::mount_sse_sequence(&server, vec![first_response, second_response]).await;
let model_provider = ModelProviderInfo {
base_url: Some(format!("{}/v1", server.uri())),
..built_in_model_providers()["openai"].clone()
};
let codex_home = TempDir::new().unwrap();
let mut config = load_default_config_for_test(&codex_home);
config.model_provider = model_provider;
config.features.enable(Feature::ResponsesApiChaining);
let conversation_manager =
ConversationManager::with_auth(CodexAuth::from_api_key("Test API Key"));
let codex = conversation_manager
.new_conversation(config)
.await
.expect("create new conversation")
.conversation;
codex
.submit(Op::UserInput {
items: vec![UserInput::Text {
text: "first turn".into(),
}],
})
.await
.unwrap();
wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
codex
.submit(Op::UserInput {
items: vec![UserInput::Text {
text: "second turn".into(),
}],
})
.await
.unwrap();
wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
let requests = response_mock.requests();
assert_eq!(
requests.len(),
2,
"expected two responses API calls for two turns"
);
let first_body = requests[0].body_json();
assert_eq!(first_body["store"], serde_json::Value::Bool(true));
assert!(
first_body.get("previous_response_id").is_none(),
"first request should not set previous_response_id"
);
let second_body = requests[1].body_json();
assert_eq!(second_body["store"], serde_json::Value::Bool(true));
assert_eq!(
second_body["previous_response_id"].as_str(),
Some("resp-first")
);
let second_input = requests[1].input();
assert_eq!(
second_input.len(),
1,
"second request should only send new user input items"
);
let user_item = &second_input[0];
assert_eq!(user_item["type"].as_str(), Some("message"));
assert_eq!(user_item["role"].as_str(), Some("user"));
let content = user_item["content"][0]["text"]
.as_str()
.expect("missing user message text");
assert_eq!(content, "second turn");
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn azure_responses_request_includes_store_and_reasoning_ids() {
skip_if_no_network!();
@@ -675,6 +838,7 @@ async fn azure_responses_request_includes_store_and_reasoning_ids() {
effort,
summary,
conversation_id,
codex_protocol::protocol::SessionSource::Exec,
);
let mut prompt = Prompt::default();
@@ -729,7 +893,7 @@ async fn azure_responses_request_includes_store_and_reasoning_ids() {
});
let mut stream = client
.stream(&prompt)
.stream_for_test(prompt)
.await
.expect("responses stream to start");
@@ -1261,6 +1425,10 @@ async fn history_dedupes_streamed_and_final_messages_across_turns() {
// Build a small SSE stream with deltas and a final assistant message.
// We emit the same body for all 3 turns; ids vary but are unused by assertions.
let sse_raw = r##"[
{"type":"response.output_item.added", "item":{
"type":"message", "role":"assistant",
"content":[{"type":"output_text","text":""}]
}},
{"type":"response.output_text.delta", "delta":"Hey "},
{"type":"response.output_text.delta", "delta":"there"},
{"type":"response.output_text.delta", "delta":"!\n"},

View File

@@ -0,0 +1,173 @@
use codex_core::protocol::AskForApproval;
use codex_core::protocol::EventMsg;
use codex_core::protocol::Op;
use codex_core::protocol::ReviewDecision;
use codex_core::protocol::ReviewRequest;
use codex_core::protocol::SandboxPolicy;
use core_test_support::responses::ev_apply_patch_function_call;
use core_test_support::responses::ev_assistant_message;
use core_test_support::responses::ev_completed;
use core_test_support::responses::ev_function_call;
use core_test_support::responses::ev_response_created;
use core_test_support::responses::mount_sse_sequence;
use core_test_support::responses::sse;
use core_test_support::responses::start_mock_server;
use core_test_support::skip_if_no_network;
use core_test_support::test_codex::test_codex;
use core_test_support::wait_for_event;
/// Delegate should surface ExecApprovalRequest from sub-agent and proceed
/// after parent submits an approval decision.
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn codex_delegate_forwards_exec_approval_and_proceeds_on_approval() {
skip_if_no_network!();
// Sub-agent turn 1: emit a shell function_call requiring approval, then complete.
let call_id = "call-exec-1";
let args = serde_json::json!({
"command": ["bash", "-lc", "rm -rf delegated"],
"timeout_ms": 1000,
"with_escalated_permissions": true,
})
.to_string();
let sse1 = sse(vec![
ev_response_created("resp-1"),
ev_function_call(call_id, "shell", &args),
ev_completed("resp-1"),
]);
// Sub-agent turn 2: return structured review output and complete.
let review_json = serde_json::json!({
"findings": [],
"overall_correctness": "ok",
"overall_explanation": "delegate approved exec",
"overall_confidence_score": 0.5
})
.to_string();
let sse2 = sse(vec![
ev_response_created("resp-2"),
ev_assistant_message("msg-1", &review_json),
ev_completed("resp-2"),
]);
let server = start_mock_server().await;
mount_sse_sequence(&server, vec![sse1, sse2]).await;
// Build a conversation configured to require approvals so the delegate
// routes ExecApprovalRequest via the parent.
let mut builder = test_codex().with_config(|config| {
config.approval_policy = AskForApproval::OnRequest;
config.sandbox_policy = SandboxPolicy::ReadOnly;
});
let test = builder.build(&server).await.expect("build test codex");
// Kick off review (sub-agent starts internally).
test.codex
.submit(Op::Review {
review_request: ReviewRequest {
prompt: "Please review".to_string(),
user_facing_hint: "review".to_string(),
},
})
.await
.expect("submit review");
// Lifecycle: Entered -> ExecApprovalRequest -> Exited(Some) -> TaskComplete.
wait_for_event(&test.codex, |ev| {
matches!(ev, EventMsg::EnteredReviewMode(_))
})
.await;
// Expect parent-side approval request (forwarded by delegate).
wait_for_event(&test.codex, |ev| {
matches!(ev, EventMsg::ExecApprovalRequest(_))
})
.await;
// Approve via parent; id "0" is the active sub_id in tests.
test.codex
.submit(Op::ExecApproval {
id: "0".into(),
decision: ReviewDecision::Approved,
})
.await
.expect("submit exec approval");
wait_for_event(&test.codex, |ev| {
matches!(ev, EventMsg::ExitedReviewMode(_))
})
.await;
wait_for_event(&test.codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
}
/// Delegate should surface ApplyPatchApprovalRequest and honor parent decision
/// so the sub-agent can proceed to completion.
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn codex_delegate_forwards_patch_approval_and_proceeds_on_decision() {
skip_if_no_network!();
let call_id = "call-patch-1";
let patch = "*** Begin Patch\n*** Add File: delegated.txt\n+hello\n*** End Patch\n";
let sse1 = sse(vec![
ev_response_created("resp-1"),
ev_apply_patch_function_call(call_id, patch),
ev_completed("resp-1"),
]);
let review_json = serde_json::json!({
"findings": [],
"overall_correctness": "ok",
"overall_explanation": "delegate patch handled",
"overall_confidence_score": 0.5
})
.to_string();
let sse2 = sse(vec![
ev_response_created("resp-2"),
ev_assistant_message("msg-1", &review_json),
ev_completed("resp-2"),
]);
let server = start_mock_server().await;
mount_sse_sequence(&server, vec![sse1, sse2]).await;
let mut builder = test_codex().with_config(|config| {
config.approval_policy = AskForApproval::OnRequest;
// Use a restricted sandbox so patch approval is required
config.sandbox_policy = SandboxPolicy::ReadOnly;
config.include_apply_patch_tool = true;
});
let test = builder.build(&server).await.expect("build test codex");
test.codex
.submit(Op::Review {
review_request: ReviewRequest {
prompt: "Please review".to_string(),
user_facing_hint: "review".to_string(),
},
})
.await
.expect("submit review");
wait_for_event(&test.codex, |ev| {
matches!(ev, EventMsg::EnteredReviewMode(_))
})
.await;
wait_for_event(&test.codex, |ev| {
matches!(ev, EventMsg::ApplyPatchApprovalRequest(_))
})
.await;
// Deny via parent so delegate can continue; id "0" is the active sub_id in tests.
test.codex
.submit(Op::PatchApproval {
id: "0".into(),
decision: ReviewDecision::Denied,
})
.await
.expect("submit patch approval");
wait_for_event(&test.codex, |ev| {
matches!(ev, EventMsg::ExitedReviewMode(_))
})
.await;
wait_for_event(&test.codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
}

View File

@@ -1,8 +1,8 @@
use codex_api_client::ModelProviderInfo;
use codex_api_client::built_in_model_providers;
use codex_core::CodexAuth;
use codex_core::ConversationManager;
use codex_core::ModelProviderInfo;
use codex_core::NewConversation;
use codex_core::built_in_model_providers;
use codex_core::protocol::ErrorEvent;
use codex_core::protocol::EventMsg;
use codex_core::protocol::Op;
@@ -261,6 +261,65 @@ async fn summarize_context_three_requests_and_instructions() {
);
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn manual_compact_uses_custom_prompt() {
skip_if_no_network!();
let server = start_mock_server().await;
let sse_stream = sse(vec![ev_completed("r1")]);
mount_sse_once(&server, sse_stream).await;
let custom_prompt = "Use this compact prompt instead";
let model_provider = ModelProviderInfo {
base_url: Some(format!("{}/v1", server.uri())),
..built_in_model_providers()["openai"].clone()
};
let home = TempDir::new().unwrap();
let mut config = load_default_config_for_test(&home);
config.model_provider = model_provider;
config.compact_prompt = Some(custom_prompt.to_string());
let conversation_manager = ConversationManager::with_auth(CodexAuth::from_api_key("dummy"));
let codex = conversation_manager
.new_conversation(config)
.await
.expect("create conversation")
.conversation;
codex.submit(Op::Compact).await.expect("trigger compact");
wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
let requests = server.received_requests().await.expect("collect requests");
let body = requests
.iter()
.find_map(|req| req.body_json::<serde_json::Value>().ok())
.expect("summary request body");
let input = body
.get("input")
.and_then(|v| v.as_array())
.expect("input array");
let mut found_custom_prompt = false;
let mut found_default_prompt = false;
for item in input {
if item["type"].as_str() != Some("message") {
continue;
}
let text = item["content"][0]["text"].as_str().unwrap_or_default();
if text == custom_prompt {
found_custom_prompt = true;
}
if text == SUMMARIZATION_PROMPT {
found_default_prompt = true;
}
}
assert!(found_custom_prompt, "custom prompt should be injected");
assert!(!found_default_prompt, "default prompt should be replaced");
}
// Windows CI only: bump to 4 workers to prevent SSE/event starvation and test timeouts.
#[cfg_attr(windows, tokio::test(flavor = "multi_thread", worker_threads = 4))]
#[cfg_attr(not(windows), tokio::test(flavor = "multi_thread", worker_threads = 2))]

View File

@@ -9,12 +9,12 @@
use super::compact::FIRST_REPLY;
use super::compact::SUMMARY_TEXT;
use codex_api_client::ModelProviderInfo;
use codex_api_client::built_in_model_providers;
use codex_core::CodexAuth;
use codex_core::CodexConversation;
use codex_core::ConversationManager;
use codex_core::ModelProviderInfo;
use codex_core::NewConversation;
use codex_core::built_in_model_providers;
use codex_core::codex::compact::SUMMARIZATION_PROMPT;
use codex_core::config::Config;
use codex_core::config::OPENAI_DEFAULT_MODEL;

View File

@@ -1,8 +1,8 @@
use codex_api_client::ModelProviderInfo;
use codex_api_client::built_in_model_providers;
use codex_core::CodexAuth;
use codex_core::ConversationManager;
use codex_core::ModelProviderInfo;
use codex_core::NewConversation;
use codex_core::built_in_model_providers;
use codex_core::parse_turn_item;
use codex_core::protocol::EventMsg;
use codex_core::protocol::Op;

View File

@@ -9,7 +9,12 @@ use codex_protocol::items::TurnItem;
use codex_protocol::user_input::UserInput;
use core_test_support::responses::ev_assistant_message;
use core_test_support::responses::ev_completed;
use core_test_support::responses::ev_message_item_added;
use core_test_support::responses::ev_output_text_delta;
use core_test_support::responses::ev_reasoning_item;
use core_test_support::responses::ev_reasoning_item_added;
use core_test_support::responses::ev_reasoning_summary_text_delta;
use core_test_support::responses::ev_reasoning_text_delta;
use core_test_support::responses::ev_response_created;
use core_test_support::responses::ev_web_search_call_added;
use core_test_support::responses::ev_web_search_call_done;
@@ -234,3 +239,181 @@ async fn web_search_item_is_emitted() -> anyhow::Result<()> {
Ok(())
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn agent_message_content_delta_has_item_metadata() -> anyhow::Result<()> {
skip_if_no_network!(Ok(()));
let server = start_mock_server().await;
let TestCodex {
codex,
session_configured,
..
} = test_codex().build(&server).await?;
let stream = sse(vec![
ev_response_created("resp-1"),
ev_message_item_added("msg-1", ""),
ev_output_text_delta("streamed response"),
ev_assistant_message("msg-1", "streamed response"),
ev_completed("resp-1"),
]);
mount_sse_once_match(&server, any(), stream).await;
codex
.submit(Op::UserInput {
items: vec![UserInput::Text {
text: "please stream text".into(),
}],
})
.await?;
let (started_turn_id, started_item) = wait_for_event_match(&codex, |ev| match ev {
EventMsg::ItemStarted(ItemStartedEvent {
turn_id,
item: TurnItem::AgentMessage(item),
..
}) => Some((turn_id.clone(), item.clone())),
_ => None,
})
.await;
let delta_event = wait_for_event_match(&codex, |ev| match ev {
EventMsg::AgentMessageContentDelta(event) => Some(event.clone()),
_ => None,
})
.await;
let legacy_delta = wait_for_event_match(&codex, |ev| match ev {
EventMsg::AgentMessageDelta(event) => Some(event.clone()),
_ => None,
})
.await;
let completed_item = wait_for_event_match(&codex, |ev| match ev {
EventMsg::ItemCompleted(ItemCompletedEvent {
item: TurnItem::AgentMessage(item),
..
}) => Some(item.clone()),
_ => None,
})
.await;
let session_id = session_configured.session_id.to_string();
assert_eq!(delta_event.thread_id, session_id);
assert_eq!(delta_event.turn_id, started_turn_id);
assert_eq!(delta_event.item_id, started_item.id);
assert_eq!(delta_event.delta, "streamed response");
assert_eq!(legacy_delta.delta, "streamed response");
assert_eq!(completed_item.id, started_item.id);
Ok(())
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn reasoning_content_delta_has_item_metadata() -> anyhow::Result<()> {
skip_if_no_network!(Ok(()));
let server = start_mock_server().await;
let TestCodex { codex, .. } = test_codex().build(&server).await?;
let stream = sse(vec![
ev_response_created("resp-1"),
ev_reasoning_item_added("reasoning-1", &[""]),
ev_reasoning_summary_text_delta("step one"),
ev_reasoning_item("reasoning-1", &["step one"], &[]),
ev_completed("resp-1"),
]);
mount_sse_once_match(&server, any(), stream).await;
codex
.submit(Op::UserInput {
items: vec![UserInput::Text {
text: "reason through it".into(),
}],
})
.await?;
let reasoning_item = wait_for_event_match(&codex, |ev| match ev {
EventMsg::ItemStarted(ItemStartedEvent {
item: TurnItem::Reasoning(item),
..
}) => Some(item.clone()),
_ => None,
})
.await;
let delta_event = wait_for_event_match(&codex, |ev| match ev {
EventMsg::ReasoningContentDelta(event) => Some(event.clone()),
_ => None,
})
.await;
let legacy_delta = wait_for_event_match(&codex, |ev| match ev {
EventMsg::AgentReasoningDelta(event) => Some(event.clone()),
_ => None,
})
.await;
assert_eq!(delta_event.item_id, reasoning_item.id);
assert_eq!(delta_event.delta, "step one");
assert_eq!(legacy_delta.delta, "step one");
Ok(())
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn reasoning_raw_content_delta_respects_flag() -> anyhow::Result<()> {
skip_if_no_network!(Ok(()));
let server = start_mock_server().await;
let TestCodex { codex, .. } = test_codex()
.with_config(|config| {
config.show_raw_agent_reasoning = true;
})
.build(&server)
.await?;
let stream = sse(vec![
ev_response_created("resp-1"),
ev_reasoning_item_added("reasoning-raw", &[""]),
ev_reasoning_text_delta("raw detail"),
ev_reasoning_item("reasoning-raw", &["complete"], &["raw detail"]),
ev_completed("resp-1"),
]);
mount_sse_once_match(&server, any(), stream).await;
codex
.submit(Op::UserInput {
items: vec![UserInput::Text {
text: "show raw reasoning".into(),
}],
})
.await?;
let reasoning_item = wait_for_event_match(&codex, |ev| match ev {
EventMsg::ItemStarted(ItemStartedEvent {
item: TurnItem::Reasoning(item),
..
}) => Some(item.clone()),
_ => None,
})
.await;
let delta_event = wait_for_event_match(&codex, |ev| match ev {
EventMsg::ReasoningRawContentDelta(event) => Some(event.clone()),
_ => None,
})
.await;
let legacy_delta = wait_for_event_match(&codex, |ev| match ev {
EventMsg::AgentReasoningRawContentDelta(event) => Some(event.clone()),
_ => None,
})
.await;
assert_eq!(delta_event.item_id, reasoning_item.id);
assert_eq!(delta_event.delta, "raw detail");
assert_eq!(legacy_delta.delta, "raw detail");
Ok(())
}

View File

@@ -5,9 +5,12 @@ mod abort_tasks;
#[cfg(not(target_os = "windows"))]
mod apply_patch_cli;
#[cfg(not(target_os = "windows"))]
mod apply_patch_freeform;
#[cfg(not(target_os = "windows"))]
mod approvals;
mod cli_stream;
mod client;
mod codex_delegate;
mod compact;
mod compact_resume_fork;
mod deprecation_notice;

View File

@@ -1,9 +1,9 @@
#![allow(clippy::unwrap_used)]
use codex_api_client::ModelProviderInfo;
use codex_api_client::built_in_model_providers;
use codex_core::CodexAuth;
use codex_core::ConversationManager;
use codex_core::ModelProviderInfo;
use codex_core::built_in_model_providers;
use codex_core::features::Feature;
use codex_core::model_family::find_family_for_model;
use codex_core::protocol::EventMsg;

View File

@@ -1,9 +1,9 @@
#![allow(clippy::unwrap_used)]
use codex_api_client::ModelProviderInfo;
use codex_api_client::built_in_model_providers;
use codex_core::CodexAuth;
use codex_core::ConversationManager;
use codex_core::ModelProviderInfo;
use codex_core::built_in_model_providers;
use codex_core::config::OPENAI_DEFAULT_MODEL;
use codex_core::features::Feature;
use codex_core::model_family::find_family_for_model;

View File

@@ -1,11 +1,11 @@
use codex_api_client::ModelProviderInfo;
use codex_api_client::built_in_model_providers;
use codex_core::CodexAuth;
use codex_core::CodexConversation;
use codex_core::ContentItem;
use codex_core::ConversationManager;
use codex_core::ModelProviderInfo;
use codex_core::REVIEW_PROMPT;
use codex_core::ResponseItem;
use codex_core::built_in_model_providers;
use codex_core::config::Config;
use codex_core::protocol::ENVIRONMENT_CONTEXT_OPEN_TAG;
use codex_core::protocol::EventMsg;
@@ -204,6 +204,85 @@ async fn review_op_with_plain_text_emits_review_fallback() {
server.verify().await;
}
/// Ensure review flow suppresses assistant-specific streaming/completion events:
/// - AgentMessageContentDelta
/// - AgentMessageDelta (legacy)
/// - ItemCompleted for TurnItem::AgentMessage
// Windows CI only: bump to 4 workers to prevent SSE/event starvation and test timeouts.
#[cfg_attr(windows, tokio::test(flavor = "multi_thread", worker_threads = 4))]
#[cfg_attr(not(windows), tokio::test(flavor = "multi_thread", worker_threads = 2))]
async fn review_filters_agent_message_related_events() {
skip_if_no_network!();
// Stream simulating a typing assistant message with deltas and finalization.
let sse_raw = r#"[
{"type":"response.output_item.added", "item":{
"type":"message", "role":"assistant", "id":"msg-1",
"content":[{"type":"output_text","text":""}]
}},
{"type":"response.output_text.delta", "delta":"Hi"},
{"type":"response.output_text.delta", "delta":" there"},
{"type":"response.output_item.done", "item":{
"type":"message", "role":"assistant", "id":"msg-1",
"content":[{"type":"output_text","text":"Hi there"}]
}},
{"type":"response.completed", "response": {"id": "__ID__"}}
]"#;
let server = start_responses_server_with_sse(sse_raw, 1).await;
let codex_home = TempDir::new().unwrap();
let codex = new_conversation_for_server(&server, &codex_home, |_| {}).await;
codex
.submit(Op::Review {
review_request: ReviewRequest {
prompt: "Filter streaming events".to_string(),
user_facing_hint: "Filter streaming events".to_string(),
},
})
.await
.unwrap();
let mut saw_entered = false;
let mut saw_exited = false;
// Drain until TaskComplete; assert filtered events never surface.
wait_for_event_with_timeout(
&codex,
|event| match event {
EventMsg::TaskComplete(_) => true,
EventMsg::EnteredReviewMode(_) => {
saw_entered = true;
false
}
EventMsg::ExitedReviewMode(_) => {
saw_exited = true;
false
}
// The following must be filtered by review flow
EventMsg::AgentMessageContentDelta(_) => {
panic!("unexpected AgentMessageContentDelta surfaced during review")
}
EventMsg::AgentMessageDelta(_) => {
panic!("unexpected AgentMessageDelta surfaced during review")
}
EventMsg::ItemCompleted(ev) => match &ev.item {
codex_protocol::items::TurnItem::AgentMessage(_) => {
panic!(
"unexpected ItemCompleted for TurnItem::AgentMessage surfaced during review"
)
}
_ => false,
},
_ => false,
},
tokio::time::Duration::from_secs(5),
)
.await;
assert!(saw_entered && saw_exited, "missing review lifecycle events");
server.verify().await;
}
/// When the model returns structured JSON in a review, ensure no AgentMessage
/// is emitted; the UI consumes the structured result via ExitedReviewMode.
// Windows CI only: bump to 4 workers to prevent SSE/event starvation and test timeouts.

View File

@@ -8,8 +8,8 @@ use std::time::Duration;
use std::time::SystemTime;
use std::time::UNIX_EPOCH;
use codex_core::config_types::McpServerConfig;
use codex_core::config_types::McpServerTransportConfig;
use codex_core::config::types::McpServerConfig;
use codex_core::config::types::McpServerTransportConfig;
use codex_core::features::Feature;
use codex_core::protocol::AskForApproval;
@@ -422,7 +422,7 @@ async fn stdio_image_completions_round_trip() -> anyhow::Result<()> {
let fixture = test_codex()
.with_config(move |config| {
config.model_provider.wire_api = codex_core::WireApi::Chat;
config.model_provider.wire_api = codex_api_client::WireApi::Chat;
config.features.enable(Feature::RmcpClient);
config.mcp_servers.insert(
server_name.to_string(),

View File

@@ -1,7 +1,7 @@
use std::time::Duration;
use codex_core::ModelProviderInfo;
use codex_core::WireApi;
use codex_api_client::ModelProviderInfo;
use codex_api_client::WireApi;
use codex_core::protocol::EventMsg;
use codex_core::protocol::Op;
use codex_protocol::user_input::UserInput;

View File

@@ -3,8 +3,8 @@
use std::time::Duration;
use codex_core::ModelProviderInfo;
use codex_core::WireApi;
use codex_api_client::ModelProviderInfo;
use codex_api_client::WireApi;
use codex_core::protocol::EventMsg;
use codex_core::protocol::Op;
use codex_protocol::user_input::UserInput;

View File

@@ -1,6 +1,7 @@
#![cfg(not(target_os = "windows"))]
#![allow(clippy::unwrap_used, clippy::expect_used)]
use anyhow::Context;
use anyhow::Result;
use codex_core::features::Feature;
use codex_core::model_family::find_family_for_model;
@@ -227,6 +228,103 @@ async fn shell_escalated_permissions_rejected_then_ok() -> Result<()> {
Ok(())
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn sandbox_denied_shell_returns_original_output() -> Result<()> {
skip_if_no_network!(Ok(()));
let server = start_mock_server().await;
let mut builder = test_codex().with_config(|config| {
config.model = "gpt-5-codex".to_string();
config.model_family =
find_family_for_model("gpt-5-codex").expect("gpt-5-codex model family");
});
let fixture = builder.build(&server).await?;
let call_id = "sandbox-denied-shell";
let target_path = fixture.workspace_path("sandbox-denied.txt");
let sentinel = "sandbox-denied sentinel output";
let command = vec![
"/bin/sh".to_string(),
"-c".to_string(),
format!(
"printf {sentinel:?}; printf {content:?} > {path:?}",
sentinel = format!("{sentinel}\n"),
content = "sandbox denied",
path = &target_path
),
];
let args = json!({
"command": command,
"timeout_ms": 1_000,
});
let responses = vec![
sse(vec![
ev_response_created("resp-1"),
ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
ev_completed("resp-1"),
]),
sse(vec![
ev_assistant_message("msg-1", "done"),
ev_completed("resp-2"),
]),
];
let mock = mount_sse_sequence(&server, responses).await;
fixture
.submit_turn_with_policy(
"run a command that should be denied by the read-only sandbox",
SandboxPolicy::ReadOnly,
)
.await?;
let output_text = mock
.function_call_output_text(call_id)
.context("shell output present")?;
let exit_code_line = output_text
.lines()
.next()
.context("exit code line present")?;
let exit_code = exit_code_line
.strip_prefix("Exit code: ")
.context("exit code prefix present")?
.trim()
.parse::<i32>()
.context("exit code is integer")?;
let body = output_text;
let body_lower = body.to_lowercase();
// Required for multi-OS.
let has_denial = body_lower.contains("permission denied")
|| body_lower.contains("operation not permitted")
|| body_lower.contains("read-only file system");
assert!(
has_denial,
"expected sandbox denial details in tool output: {body}"
);
assert!(
body.contains(sentinel),
"expected sentinel output from command to reach the model: {body}"
);
let target_path_str = target_path
.to_str()
.context("target path string representation")?;
assert!(
body.contains(target_path_str),
"expected sandbox error to mention denied path: {body}"
);
assert!(
!body_lower.contains("failed in sandbox"),
"expected original tool output, found fallback message: {body}"
);
assert_ne!(
exit_code, 0,
"sandbox denial should surface a non-zero exit code"
);
Ok(())
}
async fn collect_tools(use_unified_exec: bool) -> Result<Vec<String>> {
let server = start_mock_server().await;

View File

@@ -219,8 +219,8 @@ async fn mcp_tool_call_output_exceeds_limit_truncated_for_model() -> Result<()>
config.features.enable(Feature::RmcpClient);
config.mcp_servers.insert(
server_name.to_string(),
codex_core::config_types::McpServerConfig {
transport: codex_core::config_types::McpServerTransportConfig::Stdio {
codex_core::config::types::McpServerConfig {
transport: codex_core::config::types::McpServerTransportConfig::Stdio {
command: rmcp_test_server_bin,
args: Vec::new(),
env: None,

View File

@@ -61,6 +61,7 @@ Request `newConversation` params (subset):
- `sandbox`: `read-only` | `workspace-write` | `danger-full-access`
- `config`: map of additional config overrides
- `baseInstructions`: optional instruction override
- `compactPrompt`: optional replacement for the default compaction prompt
- `includePlanTool` / `includeApplyPatchTool`: booleans
Response: `{ conversationId, model, reasoningEffort?, rolloutPath }`

View File

@@ -24,8 +24,10 @@ codex-common = { workspace = true, features = [
"sandbox_summary",
] }
codex-core = { workspace = true }
codex-api-client = { workspace = true }
codex-ollama = { workspace = true }
codex-protocol = { workspace = true }
mcp-types = { workspace = true }
opentelemetry-appender-tracing = { workspace = true }
owo-colors = { workspace = true }
serde = { workspace = true, features = ["derive"] }

Some files were not shown because too many files have changed in this diff Show More