Compare commits

...

14 Commits

Author SHA1 Message Date
jimmyfraiture
a9def0eac1 Drop old shells 2025-09-08 09:05:28 -07:00
jimmyfraiture
d0d21550b9 NIT windows 2025-09-07 19:18:47 -07:00
jimmyfraiture
f661af0647 V12 2025-09-07 19:07:00 -07:00
jimmyfraiture
e71a533ec1 V11 2025-09-07 18:53:54 -07:00
jimmyfraiture
95eeb3d1a1 V10 2025-09-07 18:43:54 -07:00
jimmyfraiture
c4c673164e V9 2025-09-07 18:28:54 -07:00
jimmyfraiture
2de51d4e30 V8 2025-09-07 17:51:23 -07:00
jimmyfraiture
85b505afb8 V7 2025-09-07 17:45:26 -07:00
jimmyfraiture
80eea492ff V6 2025-09-07 17:22:08 -07:00
jimmyfraiture
e1809a8f13 V5 2025-09-07 17:20:34 -07:00
jimmyfraiture
16236b699e V4 2025-09-07 16:53:29 -07:00
jimmyfraiture
f00c6819d1 V3 2025-09-07 15:48:58 -07:00
jimmyfraiture
909666b03d V2 2025-09-07 15:22:24 -07:00
jimmyfraiture
4bd40bd27f V1 2025-09-07 14:47:25 -07:00
16 changed files with 1341 additions and 190 deletions

View File

@@ -56,9 +56,9 @@ pub(crate) async fn stream_chat_completions(
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::FunctionCall { .. }
| ResponseItem::LocalShellCall { .. }
| ResponseItem::UnifiedExec { .. } => last_emitted_role = Some("assistant"),
ResponseItem::FunctionCallOutput { .. } => last_emitted_role = Some("tool"),
ResponseItem::Reasoning { .. } | ResponseItem::Other => {}
ResponseItem::CustomToolCall { .. } => {}
@@ -119,7 +119,9 @@ pub(crate) async fn stream_chat_completions(
// 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 { .. } => {
ResponseItem::FunctionCall { .. }
| ResponseItem::LocalShellCall { .. }
| ResponseItem::UnifiedExec { .. } => {
reasoning_by_anchor_index
.entry(idx + 1)
.and_modify(|v| v.push_str(&text))
@@ -224,6 +226,12 @@ pub(crate) async fn stream_chat_completions(
}
messages.push(msg);
}
ResponseItem::UnifiedExec { .. } => {
// The interactive shell tool is currently exposed only via the
// Responses API. Ignore these items when using Chat
// Completions to maintain feature parity with existing
// behaviour.
}
ResponseItem::FunctionCallOutput { call_id, output } => {
messages.push(json!({
"role": "tool",

View File

@@ -21,6 +21,7 @@ use codex_protocol::protocol::TurnAbortReason;
use codex_protocol::protocol::TurnAbortedEvent;
use futures::prelude::*;
use mcp_types::CallToolResult;
use serde::Deserialize;
use serde::Serialize;
use serde_json;
use tokio::sync::oneshot;
@@ -107,6 +108,7 @@ use crate::safety::assess_command_safety;
use crate::safety::assess_safety_for_untrusted_command;
use crate::shell;
use crate::turn_diff_tracker::TurnDiffTracker;
use crate::unified_exec::UnifiedExecSessionManager;
use crate::user_instructions::UserInstructions;
use crate::user_notification::UserNotification;
use crate::util::backoff;
@@ -273,6 +275,7 @@ pub(crate) struct Session {
/// Manager for external MCP servers/tools.
mcp_connection_manager: McpConnectionManager,
session_manager: ExecSessionManager,
unified_exec_manager: UnifiedExecSessionManager,
/// External notifier command (will be passed as args to exec()). When
/// `None` this feature is disabled.
@@ -463,6 +466,7 @@ impl Session {
tx_event: tx_event.clone(),
mcp_connection_manager,
session_manager: ExecSessionManager::default(),
unified_exec_manager: UnifiedExecSessionManager::default(),
notify,
state: Mutex::new(state),
rollout: Mutex::new(Some(rollout_recorder)),
@@ -1984,6 +1988,27 @@ async fn handle_response_item(
.await,
)
}
ResponseItem::UnifiedExec {
id,
session_id,
arguments,
timeout_ms,
} => {
let call_id = id.clone().unwrap_or_else(|| match &session_id {
Some(session_id) => format!("unified_exec:{session_id}:{}", Uuid::new_v4()),
None => format!("unified_exec:{}", Uuid::new_v4()),
});
Some(
handle_unified_exec_tool_call(
sess,
call_id,
session_id.clone(),
arguments,
timeout_ms,
)
.await,
)
}
ResponseItem::CustomToolCall {
id: _,
call_id,
@@ -2028,6 +2053,72 @@ async fn handle_response_item(
Ok(output)
}
async fn handle_unified_exec_tool_call(
sess: &Session,
call_id: String,
session_id: Option<String>,
arguments: Vec<String>,
timeout_ms: Option<u64>,
) -> ResponseInputItem {
let parsed_session_id = if let Some(session_id) = session_id {
match session_id.parse::<i32>() {
Ok(parsed) => Some(parsed),
Err(output) => {
return ResponseInputItem::FunctionCallOutput {
call_id: call_id.to_string(),
output: FunctionCallOutputPayload {
content: format!("invalid session_id: {session_id} due to error {output}"),
success: Some(false),
},
};
}
}
} else {
None
};
let request = crate::unified_exec::UnifiedExecRequest {
session_id: parsed_session_id,
input_chunks: &arguments,
timeout_ms,
};
let result = sess.unified_exec_manager.handle_request(request).await;
let output_payload = match result {
Ok(value) => {
#[derive(Serialize)]
struct SerializedUnifiedExecResult<'a> {
session_id: Option<String>,
output: &'a str,
}
match serde_json::to_string(&SerializedUnifiedExecResult {
session_id: value.session_id.map(|id| id.to_string()),
output: &value.output,
}) {
Ok(serialized) => FunctionCallOutputPayload {
content: serialized,
success: Some(true),
},
Err(err) => FunctionCallOutputPayload {
content: format!("failed to serialize unified exec output: {err}"),
success: Some(false),
},
}
}
Err(err) => FunctionCallOutputPayload {
content: format!("unified exec failed: {err}"),
success: Some(false),
},
};
ResponseInputItem::FunctionCallOutput {
call_id,
output: output_payload,
}
}
async fn handle_function_call(
sess: &Session,
turn_context: &TurnContext,
@@ -2055,6 +2146,38 @@ async fn handle_function_call(
)
.await
}
"unified_exec" => {
#[derive(Deserialize)]
struct UnifiedExecArgs {
input: Vec<String>,
#[serde(default)]
session_id: Option<String>,
#[serde(default)]
timeout_ms: Option<u64>,
}
let args = match serde_json::from_str::<UnifiedExecArgs>(&arguments) {
Ok(args) => args,
Err(err) => {
return ResponseInputItem::FunctionCallOutput {
call_id,
output: FunctionCallOutputPayload {
content: format!("failed to parse function arguments: {err}"),
success: Some(false),
},
};
}
};
handle_unified_exec_tool_call(
sess,
call_id,
args.session_id,
args.input,
args.timeout_ms,
)
.await
}
"view_image" => {
#[derive(serde::Deserialize)]
struct SeeImageArgs {

View File

@@ -71,6 +71,7 @@ fn is_api_message(message: &ResponseItem) -> bool {
| ResponseItem::CustomToolCall { .. }
| ResponseItem::CustomToolCallOutput { .. }
| ResponseItem::LocalShellCall { .. }
| ResponseItem::UnifiedExec { .. }
| ResponseItem::Reasoning { .. } => true,
ResponseItem::WebSearchCall { .. } | ResponseItem::Other => false,
}

View File

@@ -91,6 +91,7 @@ pub(crate) fn map_response_item_to_event_messages(
ResponseItem::FunctionCall { .. }
| ResponseItem::FunctionCallOutput { .. }
| ResponseItem::LocalShellCall { .. }
| ResponseItem::UnifiedExec { .. }
| ResponseItem::CustomToolCall { .. }
| ResponseItem::CustomToolCallOutput { .. }
| ResponseItem::Other => Vec::new(),

View File

@@ -24,6 +24,9 @@ pub(crate) struct ExecCommandSession {
/// JoinHandle for the child wait task.
wait_handle: StdMutex<Option<JoinHandle<()>>>,
/// Tracks whether the underlying process has exited.
exit_status: std::sync::Arc<std::sync::atomic::AtomicBool>,
}
impl ExecCommandSession {
@@ -34,6 +37,7 @@ impl ExecCommandSession {
reader_handle: JoinHandle<()>,
writer_handle: JoinHandle<()>,
wait_handle: JoinHandle<()>,
exit_status: std::sync::Arc<std::sync::atomic::AtomicBool>,
) -> Self {
Self {
writer_tx,
@@ -42,6 +46,7 @@ impl ExecCommandSession {
reader_handle: StdMutex::new(Some(reader_handle)),
writer_handle: StdMutex::new(Some(writer_handle)),
wait_handle: StdMutex::new(Some(wait_handle)),
exit_status,
}
}
@@ -52,6 +57,10 @@ impl ExecCommandSession {
pub(crate) fn output_receiver(&self) -> broadcast::Receiver<Vec<u8>> {
self.output_tx.subscribe()
}
pub(crate) fn has_exited(&self) -> bool {
self.exit_status.load(std::sync::atomic::Ordering::SeqCst)
}
}
impl Drop for ExecCommandSession {

View File

@@ -6,6 +6,7 @@ mod session_manager;
pub use exec_command_params::ExecCommandParams;
pub use exec_command_params::WriteStdinParams;
pub(crate) use exec_command_session::ExecCommandSession;
pub use responses_api::EXEC_COMMAND_TOOL_NAME;
pub use responses_api::WRITE_STDIN_TOOL_NAME;
pub use responses_api::create_exec_command_tool_for_responses_api;

View File

@@ -19,6 +19,7 @@ use crate::exec_command::exec_command_params::ExecCommandParams;
use crate::exec_command::exec_command_params::WriteStdinParams;
use crate::exec_command::exec_command_session::ExecCommandSession;
use crate::exec_command::session_id::SessionId;
use crate::truncate::truncate_middle;
use codex_protocol::models::FunctionCallOutputPayload;
#[derive(Debug, Default)]
@@ -327,11 +328,14 @@ async fn create_exec_command_session(
// Keep the child alive until it exits, then signal exit code.
let (exit_tx, exit_rx) = oneshot::channel::<i32>();
let exit_status = std::sync::Arc::new(std::sync::atomic::AtomicBool::new(false));
let wait_exit_status = std::sync::Arc::clone(&exit_status);
let wait_handle = tokio::task::spawn_blocking(move || {
let code = match child.wait() {
Ok(status) => status.exit_code() as i32,
Err(_) => -1,
};
wait_exit_status.store(true, std::sync::atomic::Ordering::SeqCst);
let _ = exit_tx.send(code);
});
@@ -343,116 +347,11 @@ async fn create_exec_command_session(
reader_handle,
writer_handle,
wait_handle,
exit_status,
);
Ok((session, exit_rx))
}
/// Truncate the middle of a UTF-8 string to at most `max_bytes` bytes,
/// preserving the beginning and the end. Returns the possibly truncated
/// string and `Some(original_token_count)` (estimated at 4 bytes/token)
/// if truncation occurred; otherwise returns the original string and `None`.
fn truncate_middle(s: &str, max_bytes: usize) -> (String, Option<u64>) {
// No truncation needed
if s.len() <= max_bytes {
return (s.to_string(), None);
}
let est_tokens = (s.len() as u64).div_ceil(4);
if max_bytes == 0 {
// Cannot keep any content; still return a full marker (never truncated).
return (format!("{est_tokens} tokens truncated…"), Some(est_tokens));
}
// Helper to truncate a string to a given byte length on a char boundary.
fn truncate_on_boundary(input: &str, max_len: usize) -> &str {
if input.len() <= max_len {
return input;
}
let mut end = max_len;
while end > 0 && !input.is_char_boundary(end) {
end -= 1;
}
&input[..end]
}
// Given a left/right budget, prefer newline boundaries; otherwise fall back
// to UTF-8 char boundaries.
fn pick_prefix_end(s: &str, left_budget: usize) -> usize {
if let Some(head) = s.get(..left_budget)
&& let Some(i) = head.rfind('\n')
{
return i + 1; // keep the newline so suffix starts on a fresh line
}
truncate_on_boundary(s, left_budget).len()
}
fn pick_suffix_start(s: &str, right_budget: usize) -> usize {
let start_tail = s.len().saturating_sub(right_budget);
if let Some(tail) = s.get(start_tail..)
&& let Some(i) = tail.find('\n')
{
return start_tail + i + 1; // start after newline
}
// Fall back to a char boundary at or after start_tail.
let mut idx = start_tail.min(s.len());
while idx < s.len() && !s.is_char_boundary(idx) {
idx += 1;
}
idx
}
// Refine marker length and budgets until stable. Marker is never truncated.
let mut guess_tokens = est_tokens; // worst-case: everything truncated
for _ in 0..4 {
let marker = format!("{guess_tokens} tokens truncated…");
let marker_len = marker.len();
let keep_budget = max_bytes.saturating_sub(marker_len);
if keep_budget == 0 {
// No room for any content within the cap; return a full, untruncated marker
// that reflects the entire truncated content.
return (format!("{est_tokens} tokens truncated…"), Some(est_tokens));
}
let left_budget = keep_budget / 2;
let right_budget = keep_budget - left_budget;
let prefix_end = pick_prefix_end(s, left_budget);
let mut suffix_start = pick_suffix_start(s, right_budget);
if suffix_start < prefix_end {
suffix_start = prefix_end;
}
let kept_content_bytes = prefix_end + (s.len() - suffix_start);
let truncated_content_bytes = s.len().saturating_sub(kept_content_bytes);
let new_tokens = (truncated_content_bytes as u64).div_ceil(4);
if new_tokens == guess_tokens {
let mut out = String::with_capacity(marker_len + kept_content_bytes + 1);
out.push_str(&s[..prefix_end]);
out.push_str(&marker);
// Place marker on its own line for symmetry when we keep line boundaries.
out.push('\n');
out.push_str(&s[suffix_start..]);
return (out, Some(est_tokens));
}
guess_tokens = new_tokens;
}
// Fallback: use last guess to build output.
let marker = format!("{guess_tokens} tokens truncated…");
let marker_len = marker.len();
let keep_budget = max_bytes.saturating_sub(marker_len);
if keep_budget == 0 {
return (format!("{est_tokens} tokens truncated…"), Some(est_tokens));
}
let left_budget = keep_budget / 2;
let right_budget = keep_budget - left_budget;
let prefix_end = pick_prefix_end(s, left_budget);
let suffix_start = pick_suffix_start(s, right_budget);
let mut out = String::with_capacity(marker_len + prefix_end + (s.len() - suffix_start) + 1);
out.push_str(&s[..prefix_end]);
out.push_str(&marker);
out.push('\n');
out.push_str(&s[suffix_start..]);
(out, Some(est_tokens))
}
#[cfg(test)]
mod tests {
use super::*;
@@ -616,50 +515,4 @@ Output:
abc"#;
assert_eq!(expected, text);
}
#[test]
fn truncate_middle_no_newlines_fallback() {
// A long string with no newlines that exceeds the cap.
let s = "abcdefghijklmnopqrstuvwxyz0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ";
let max_bytes = 16; // force truncation
let (out, original) = truncate_middle(s, max_bytes);
// For very small caps, we return the full, untruncated marker,
// even if it exceeds the cap.
assert_eq!(out, "…16 tokens truncated…");
// Original string length is 62 bytes => ceil(62/4) = 16 tokens.
assert_eq!(original, Some(16));
}
#[test]
fn truncate_middle_prefers_newline_boundaries() {
// Build a multi-line string of 20 numbered lines (each "NNN\n").
let mut s = String::new();
for i in 1..=20 {
s.push_str(&format!("{i:03}\n"));
}
// Total length: 20 lines * 4 bytes per line = 80 bytes.
assert_eq!(s.len(), 80);
// Choose a cap that forces truncation while leaving room for
// a few lines on each side after accounting for the marker.
let max_bytes = 64;
// Expect exact output: first 4 lines, marker, last 4 lines, and correct token estimate (80/4 = 20).
assert_eq!(
truncate_middle(&s, max_bytes),
(
r#"001
002
003
004
…12 tokens truncated…
017
018
019
020
"#
.to_string(),
Some(20)
)
);
}
}

View File

@@ -34,6 +34,8 @@ mod mcp_tool_call;
mod message_history;
mod model_provider_info;
pub mod parse_command;
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;

View File

@@ -200,6 +200,53 @@ fn create_shell_tool() -> OpenAiTool {
})
}
fn create_unified_exec_tool() -> OpenAiTool {
let mut properties = BTreeMap::new();
properties.insert(
"input".to_string(),
JsonSchema::Array {
items: Box::new(JsonSchema::String { description: None }),
description: Some(
"When no session_id is provided, treat the array as the command and arguments \
to launch. When session_id is set, concatenate the strings (in order) and write \
them to the session's stdin."
.to_string(),
),
},
);
properties.insert(
"session_id".to_string(),
JsonSchema::String {
description: Some(
"Identifier for an existing interactive session. If omitted, a new command \
is spawned."
.to_string(),
),
},
);
properties.insert(
"timeout_ms".to_string(),
JsonSchema::Number {
description: Some(
"Maximum time in milliseconds to wait for output after writing the input."
.to_string(),
),
},
);
OpenAiTool::Function(ResponsesApiTool {
name: "unified_exec".to_string(),
description:
"Runs a command in a PTY. Provide a session_id to reuse an existing interactive session.".to_string(),
strict: false,
parameters: JsonSchema::Object {
properties,
required: Some(vec!["input".to_string()]),
additional_properties: Some(false),
},
})
}
fn create_shell_tool_for_sandbox(sandbox_policy: &SandboxPolicy) -> OpenAiTool {
let mut properties = BTreeMap::new();
properties.insert(
@@ -534,25 +581,27 @@ pub(crate) fn get_openai_tools(
) -> Vec<OpenAiTool> {
let mut tools: Vec<OpenAiTool> = Vec::new();
match &config.shell_type {
ConfigShellToolType::DefaultShell => {
tools.push(create_shell_tool());
}
ConfigShellToolType::ShellWithRequest { sandbox_policy } => {
tools.push(create_shell_tool_for_sandbox(sandbox_policy));
}
ConfigShellToolType::LocalShell => {
tools.push(OpenAiTool::LocalShell {});
}
ConfigShellToolType::StreamableShell => {
tools.push(OpenAiTool::Function(
crate::exec_command::create_exec_command_tool_for_responses_api(),
));
tools.push(OpenAiTool::Function(
crate::exec_command::create_write_stdin_tool_for_responses_api(),
));
}
}
// match &config.shell_type {
// ConfigShellToolType::DefaultShell => {
// tools.push(create_shell_tool());
// }
// ConfigShellToolType::ShellWithRequest { sandbox_policy } => {
// tools.push(create_shell_tool_for_sandbox(sandbox_policy));
// }
// ConfigShellToolType::LocalShell => {
// tools.push(OpenAiTool::LocalShell {});
// }
// ConfigShellToolType::StreamableShell => {
// tools.push(OpenAiTool::Function(
// crate::exec_command::create_exec_command_tool_for_responses_api(),
// ));
// tools.push(OpenAiTool::Function(
// crate::exec_command::create_write_stdin_tool_for_responses_api(),
// ));
// }
// }
tools.push(create_unified_exec_tool());
if config.plan_tool {
tools.push(PLAN_TOOL.clone());
@@ -577,10 +626,7 @@ pub(crate) fn get_openai_tools(
if config.include_view_image_tool {
tools.push(create_view_image_tool());
}
if let Some(mcp_tools) = mcp_tools {
// Ensure deterministic ordering to maximize prompt cache hits.
// HashMap iteration order is non-deterministic, so sort by fully-qualified tool name.
let mut entries: Vec<(String, mcp_types::Tool)> = mcp_tools.into_iter().collect();
entries.sort_by(|a, b| a.0.cmp(&b.0));
@@ -647,7 +693,13 @@ mod tests {
assert_eq_tool_names(
&tools,
&["local_shell", "update_plan", "web_search", "view_image"],
&[
"local_shell",
"unified_exec",
"update_plan",
"web_search",
"view_image",
],
);
}
@@ -668,7 +720,13 @@ mod tests {
assert_eq_tool_names(
&tools,
&["shell", "update_plan", "web_search", "view_image"],
&[
"shell",
"unified_exec",
"update_plan",
"web_search",
"view_image",
],
);
}
@@ -727,6 +785,7 @@ mod tests {
&tools,
&[
"shell",
"unified_exec",
"web_search",
"view_image",
"test_server/do_something_cool",
@@ -734,7 +793,7 @@ mod tests {
);
assert_eq!(
tools[3],
tools[4],
OpenAiTool::Function(ResponsesApiTool {
name: "test_server/do_something_cool".to_string(),
parameters: JsonSchema::Object {
@@ -846,6 +905,7 @@ mod tests {
&tools,
&[
"shell",
"unified_exec",
"view_image",
"test_server/cool",
"test_server/do",
@@ -893,11 +953,17 @@ mod tests {
assert_eq_tool_names(
&tools,
&["shell", "web_search", "view_image", "dash/search"],
&[
"shell",
"unified_exec",
"web_search",
"view_image",
"dash/search",
],
);
assert_eq!(
tools[3],
tools[4],
OpenAiTool::Function(ResponsesApiTool {
name: "dash/search".to_string(),
parameters: JsonSchema::Object {
@@ -953,10 +1019,16 @@ mod tests {
assert_eq_tool_names(
&tools,
&["shell", "web_search", "view_image", "dash/paginate"],
&[
"shell",
"unified_exec",
"web_search",
"view_image",
"dash/paginate",
],
);
assert_eq!(
tools[3],
tools[4],
OpenAiTool::Function(ResponsesApiTool {
name: "dash/paginate".to_string(),
parameters: JsonSchema::Object {
@@ -1008,9 +1080,18 @@ mod tests {
)])),
);
assert_eq_tool_names(&tools, &["shell", "web_search", "view_image", "dash/tags"]);
assert_eq_tool_names(
&tools,
&[
"shell",
"unified_exec",
"web_search",
"view_image",
"dash/tags",
],
);
assert_eq!(
tools[3],
tools[4],
OpenAiTool::Function(ResponsesApiTool {
name: "dash/tags".to_string(),
parameters: JsonSchema::Object {
@@ -1065,9 +1146,18 @@ mod tests {
)])),
);
assert_eq_tool_names(&tools, &["shell", "web_search", "view_image", "dash/value"]);
assert_eq_tool_names(
&tools,
&[
"shell",
"unified_exec",
"web_search",
"view_image",
"dash/value",
],
);
assert_eq!(
tools[3],
tools[4],
OpenAiTool::Function(ResponsesApiTool {
name: "dash/value".to_string(),
parameters: JsonSchema::Object {

View File

@@ -7,6 +7,7 @@ pub(crate) fn is_persisted_response_item(item: &ResponseItem) -> bool {
ResponseItem::Message { .. }
| ResponseItem::Reasoning { .. }
| ResponseItem::LocalShellCall { .. }
| ResponseItem::UnifiedExec { .. }
| ResponseItem::FunctionCall { .. }
| ResponseItem::FunctionCallOutput { .. }
| ResponseItem::CustomToolCall { .. }

View File

@@ -0,0 +1,180 @@
//! Utilities for truncating large chunks of output while preserving a prefix
//! and suffix on UTF-8 boundaries.
/// Truncate the middle of a UTF-8 string to at most `max_bytes` bytes,
/// preserving the beginning and the end. Returns the possibly truncated
/// string and `Some(original_token_count)` (estimated at 4 bytes/token)
/// if truncation occurred; otherwise returns the original string and `None`.
pub(crate) fn truncate_middle(s: &str, max_bytes: usize) -> (String, Option<u64>) {
if s.len() <= max_bytes {
return (s.to_string(), None);
}
let est_tokens = (s.len() as u64).div_ceil(4);
if max_bytes == 0 {
return (format!("{est_tokens} tokens truncated…"), Some(est_tokens));
}
fn truncate_on_boundary(input: &str, max_len: usize) -> &str {
if input.len() <= max_len {
return input;
}
let mut end = max_len;
while end > 0 && !input.is_char_boundary(end) {
end -= 1;
}
&input[..end]
}
fn pick_prefix_end(s: &str, left_budget: usize) -> usize {
if let Some(head) = s.get(..left_budget)
&& let Some(i) = head.rfind('\n')
{
return i + 1;
}
truncate_on_boundary(s, left_budget).len()
}
fn pick_suffix_start(s: &str, right_budget: usize) -> usize {
let start_tail = s.len().saturating_sub(right_budget);
if let Some(tail) = s.get(start_tail..)
&& let Some(i) = tail.find('\n')
{
return start_tail + i + 1;
}
let mut idx = start_tail.min(s.len());
while idx < s.len() && !s.is_char_boundary(idx) {
idx += 1;
}
idx
}
let mut guess_tokens = est_tokens;
for _ in 0..4 {
let marker = format!("{guess_tokens} tokens truncated…");
let marker_len = marker.len();
let keep_budget = max_bytes.saturating_sub(marker_len);
if keep_budget == 0 {
return (format!("{est_tokens} tokens truncated…"), Some(est_tokens));
}
let left_budget = keep_budget / 2;
let right_budget = keep_budget - left_budget;
let prefix_end = pick_prefix_end(s, left_budget);
let mut suffix_start = pick_suffix_start(s, right_budget);
if suffix_start < prefix_end {
suffix_start = prefix_end;
}
let kept_content_bytes = prefix_end + (s.len() - suffix_start);
let truncated_content_bytes = s.len().saturating_sub(kept_content_bytes);
let new_tokens = (truncated_content_bytes as u64).div_ceil(4);
if new_tokens == guess_tokens {
let mut out = String::with_capacity(marker_len + kept_content_bytes + 1);
out.push_str(&s[..prefix_end]);
out.push_str(&marker);
out.push('\n');
out.push_str(&s[suffix_start..]);
return (out, Some(est_tokens));
}
guess_tokens = new_tokens;
}
let marker = format!("{guess_tokens} tokens truncated…");
let marker_len = marker.len();
let keep_budget = max_bytes.saturating_sub(marker_len);
if keep_budget == 0 {
return (format!("{est_tokens} tokens truncated…"), Some(est_tokens));
}
let left_budget = keep_budget / 2;
let right_budget = keep_budget - left_budget;
let prefix_end = pick_prefix_end(s, left_budget);
let suffix_start = pick_suffix_start(s, right_budget);
let mut out = String::with_capacity(marker_len + prefix_end + (s.len() - suffix_start) + 1);
out.push_str(&s[..prefix_end]);
out.push_str(&marker);
out.push('\n');
out.push_str(&s[suffix_start..]);
(out, Some(est_tokens))
}
#[cfg(test)]
mod tests {
use super::truncate_middle;
#[test]
fn truncate_middle_no_newlines_fallback() {
let s = "abcdefghijklmnopqrstuvwxyz0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ*";
let max_bytes = 32;
let (out, original) = truncate_middle(s, max_bytes);
assert!(out.starts_with("abc"));
assert!(out.contains("tokens truncated"));
assert!(out.ends_with("XYZ*"));
assert_eq!(original, Some((s.len() as u64).div_ceil(4)));
}
#[test]
fn truncate_middle_prefers_newline_boundaries() {
let mut s = String::new();
for i in 1..=20 {
s.push_str(&format!("{i:03}\n"));
}
assert_eq!(s.len(), 80);
let max_bytes = 64;
let (out, tokens) = truncate_middle(&s, max_bytes);
assert!(out.starts_with("001\n002\n003\n004\n"));
assert!(out.contains("tokens truncated"));
assert!(out.ends_with("017\n018\n019\n020\n"));
assert_eq!(tokens, Some(20));
}
#[test]
fn truncate_middle_handles_utf8_content() {
let s = "😀😀😀😀😀😀😀😀😀😀\nsecond line with ascii text\n";
let max_bytes = 32;
let (out, tokens) = truncate_middle(s, max_bytes);
assert!(out.contains("tokens truncated"));
assert!(!out.contains('\u{fffd}'));
assert_eq!(tokens, Some((s.len() as u64).div_ceil(4)));
}
#[test]
fn truncate_middle_prefers_newline_boundaries_2() {
// Build a multi-line string of 20 numbered lines (each "NNN\n").
let mut s = String::new();
for i in 1..=20 {
s.push_str(&format!("{i:03}\n"));
}
// Total length: 20 lines * 4 bytes per line = 80 bytes.
assert_eq!(s.len(), 80);
// Choose a cap that forces truncation while leaving room for
// a few lines on each side after accounting for the marker.
let max_bytes = 64;
// Expect exact output: first 4 lines, marker, last 4 lines, and correct token estimate (80/4 = 20).
assert_eq!(
truncate_middle(&s, max_bytes),
(
r#"001
002
003
004
…12 tokens truncated…
017
018
019
020
"#
.to_string(),
Some(20)
)
);
}
}

View File

@@ -0,0 +1,26 @@
use thiserror::Error;
#[derive(Debug, Error)]
pub(crate) enum UnifiedExecError {
#[error("Failed to create unified exec session: {pty_error}")]
CreateSession {
#[source]
pty_error: anyhow::Error,
},
#[error("Unknown session id {session_id}")]
UnknownSessionId { session_id: i32 },
#[error("failed to write to stdin")]
WriteToStdin,
#[error("missing command line for unified exec request")]
MissingCommandLine,
#[error("invalid command line: {command_line}")]
InvalidCommandLine { command_line: String },
#[error("command not found: {command}")]
CommandNotFound { command: String },
}
impl UnifiedExecError {
pub(crate) fn create_session(error: anyhow::Error) -> Self {
Self::CreateSession { pty_error: error }
}
}

View File

@@ -0,0 +1,643 @@
use portable_pty::CommandBuilder;
use portable_pty::PtySize;
use portable_pty::native_pty_system;
use std::collections::HashMap;
use std::collections::VecDeque;
use std::io::ErrorKind;
use std::io::Read;
use std::sync::Arc;
use std::sync::Mutex as StdMutex;
use std::sync::atomic::AtomicBool;
use std::sync::atomic::AtomicI32;
use std::sync::atomic::Ordering;
use tokio::sync::Mutex;
use tokio::sync::Notify;
use tokio::sync::mpsc;
use tokio::task::JoinHandle;
use tokio::time::Duration;
use tokio::time::Instant;
use crate::exec_command::ExecCommandSession;
use crate::truncate::truncate_middle;
mod errors;
mod path;
pub(crate) use errors::UnifiedExecError;
use path::command_from_chunks;
use path::join_input_chunks;
use path::resolve_command_path;
const DEFAULT_TIMEOUT_MS: u64 = 1_000;
const MAX_TIMEOUT_MS: u64 = 60_000;
const UNIFIED_EXEC_OUTPUT_MAX_BYTES: usize = 128 * 1024; // 128 KiB
#[derive(Debug)]
pub(crate) struct UnifiedExecRequest<'a> {
pub session_id: Option<i32>,
pub input_chunks: &'a [String],
pub timeout_ms: Option<u64>,
}
#[derive(Debug, Clone, PartialEq)]
pub(crate) struct UnifiedExecResult {
pub session_id: Option<i32>,
pub output: String,
}
#[derive(Debug, Default)]
pub(crate) struct UnifiedExecSessionManager {
next_session_id: AtomicI32,
sessions: Mutex<HashMap<i32, ManagedUnifiedExecSession>>,
}
#[derive(Debug)]
struct ManagedUnifiedExecSession {
session: ExecCommandSession,
output_buffer: OutputBuffer,
output_notify: Arc<Notify>,
output_task: JoinHandle<()>,
}
#[derive(Debug, Default)]
struct OutputBufferState {
chunks: VecDeque<Vec<u8>>,
total_bytes: usize,
}
impl OutputBufferState {
fn push_chunk(&mut self, chunk: Vec<u8>) {
self.total_bytes = self.total_bytes.saturating_add(chunk.len());
self.chunks.push_back(chunk);
while self.total_bytes > UNIFIED_EXEC_OUTPUT_MAX_BYTES {
match self.chunks.pop_front() {
Some(removed) => {
self.total_bytes = self.total_bytes.saturating_sub(removed.len());
}
None => break,
}
}
}
fn drain(&mut self) -> Vec<Vec<u8>> {
let drained: Vec<Vec<u8>> = self.chunks.drain(..).collect();
self.total_bytes = 0;
drained
}
}
type OutputBuffer = Arc<Mutex<OutputBufferState>>;
type OutputHandles = (OutputBuffer, Arc<Notify>);
impl ManagedUnifiedExecSession {
fn new(session: ExecCommandSession) -> Self {
let output_buffer = Arc::new(Mutex::new(OutputBufferState::default()));
let output_notify = Arc::new(Notify::new());
let mut receiver = session.output_receiver();
let buffer_clone = Arc::clone(&output_buffer);
let notify_clone = Arc::clone(&output_notify);
let output_task = tokio::spawn(async move {
while let Ok(chunk) = receiver.recv().await {
let mut guard = buffer_clone.lock().await;
guard.push_chunk(chunk);
drop(guard);
notify_clone.notify_waiters();
}
});
Self {
session,
output_buffer,
output_notify,
output_task,
}
}
fn writer_sender(&self) -> mpsc::Sender<Vec<u8>> {
self.session.writer_sender()
}
fn output_handles(&self) -> OutputHandles {
(
Arc::clone(&self.output_buffer),
Arc::clone(&self.output_notify),
)
}
fn has_exited(&self) -> bool {
self.session.has_exited()
}
}
impl Drop for ManagedUnifiedExecSession {
fn drop(&mut self) {
self.output_task.abort();
}
}
impl UnifiedExecSessionManager {
pub async fn handle_request(
&self,
request: UnifiedExecRequest<'_>,
) -> Result<UnifiedExecResult, UnifiedExecError> {
let (timeout_ms, timeout_warning) = match request.timeout_ms {
Some(requested) if requested > MAX_TIMEOUT_MS => (
MAX_TIMEOUT_MS,
Some(format!(
"Warning: requested timeout {requested}ms exceeds maximum of {MAX_TIMEOUT_MS}ms; clamping to {MAX_TIMEOUT_MS}ms.\n"
)),
),
Some(requested) => (requested, None),
None => (DEFAULT_TIMEOUT_MS, None),
};
let mut new_session: Option<ManagedUnifiedExecSession> = None;
let session_id;
let writer_tx;
let output_buffer;
let output_notify;
if let Some(existing_id) = request.session_id {
let mut sessions = self.sessions.lock().await;
match sessions.get(&existing_id) {
Some(session) => {
if session.has_exited() {
sessions.remove(&existing_id);
return Err(UnifiedExecError::UnknownSessionId {
session_id: existing_id,
});
}
let (buffer, notify) = session.output_handles();
session_id = existing_id;
writer_tx = session.writer_sender();
output_buffer = buffer;
output_notify = notify;
}
None => {
return Err(UnifiedExecError::UnknownSessionId {
session_id: existing_id,
});
}
}
drop(sessions);
} else {
let command = command_from_chunks(request.input_chunks)?;
let new_id = self.next_session_id.fetch_add(1, Ordering::SeqCst);
let session = create_unified_exec_session(&command).await?;
let managed_session = ManagedUnifiedExecSession::new(session);
let (buffer, notify) = managed_session.output_handles();
writer_tx = managed_session.writer_sender();
output_buffer = buffer;
output_notify = notify;
session_id = new_id;
new_session = Some(managed_session);
};
if request.session_id.is_some() {
let joined_input = join_input_chunks(request.input_chunks);
if !joined_input.is_empty() && writer_tx.send(joined_input.into_bytes()).await.is_err()
{
return Err(UnifiedExecError::WriteToStdin);
}
}
let mut collected: Vec<u8> = Vec::with_capacity(4096);
let start = Instant::now();
let deadline = start + Duration::from_millis(timeout_ms);
loop {
let drained_chunks;
let mut wait_for_output = None;
{
let mut guard = output_buffer.lock().await;
drained_chunks = guard.drain();
if drained_chunks.is_empty() {
wait_for_output = Some(output_notify.notified());
}
}
if drained_chunks.is_empty() {
if Instant::now() >= deadline {
break;
}
let remaining = deadline.saturating_duration_since(Instant::now());
let notified = wait_for_output.unwrap_or_else(|| output_notify.notified());
tokio::pin!(notified);
tokio::select! {
_ = &mut notified => {}
_ = tokio::time::sleep(remaining) => break,
}
continue;
}
for chunk in drained_chunks {
collected.extend_from_slice(&chunk);
}
if Instant::now() >= deadline {
break;
}
}
let (output, _maybe_tokens) = truncate_middle(
&String::from_utf8_lossy(&collected),
UNIFIED_EXEC_OUTPUT_MAX_BYTES,
);
let output = if let Some(warning) = timeout_warning {
format!("{warning}{output}")
} else {
output
};
let should_store_session = if let Some(session) = new_session.as_ref() {
!session.has_exited()
} else if request.session_id.is_some() {
let mut sessions = self.sessions.lock().await;
if let Some(existing) = sessions.get(&session_id) {
if existing.has_exited() {
sessions.remove(&session_id);
false
} else {
true
}
} else {
false
}
} else {
true
};
if should_store_session {
if let Some(session) = new_session {
self.sessions.lock().await.insert(session_id, session);
}
Ok(UnifiedExecResult {
session_id: Some(session_id),
output,
})
} else {
Ok(UnifiedExecResult {
session_id: None,
output,
})
}
}
}
async fn create_unified_exec_session(
command: &[String],
) -> Result<ExecCommandSession, UnifiedExecError> {
if command.is_empty() {
return Err(UnifiedExecError::MissingCommandLine);
}
let pty_system = native_pty_system();
let pair = pty_system
.openpty(PtySize {
rows: 24,
cols: 80,
pixel_width: 0,
pixel_height: 0,
})
.map_err(UnifiedExecError::create_session)?;
let resolved_command = resolve_command_path(&command[0])?;
let mut command_builder = CommandBuilder::new(&resolved_command);
for arg in &command[1..] {
command_builder.arg(arg);
}
let mut child = pair
.slave
.spawn_command(command_builder)
.map_err(UnifiedExecError::create_session)?;
let killer = child.clone_killer();
let (writer_tx, mut writer_rx) = mpsc::channel::<Vec<u8>>(128);
let (output_tx, _) = tokio::sync::broadcast::channel::<Vec<u8>>(256);
let mut reader = pair
.master
.try_clone_reader()
.map_err(UnifiedExecError::create_session)?;
let output_tx_clone = output_tx.clone();
let reader_handle = tokio::task::spawn_blocking(move || {
let mut buf = [0u8; 8192];
loop {
match reader.read(&mut buf) {
Ok(0) => break,
Ok(n) => {
let _ = output_tx_clone.send(buf[..n].to_vec());
}
Err(ref e) if e.kind() == ErrorKind::Interrupted => continue,
Err(ref e) if e.kind() == ErrorKind::WouldBlock => {
std::thread::sleep(Duration::from_millis(5));
continue;
}
Err(_) => break,
}
}
});
let writer = pair
.master
.take_writer()
.map_err(UnifiedExecError::create_session)?;
let writer = Arc::new(StdMutex::new(writer));
let writer_handle = tokio::spawn({
let writer = writer.clone();
async move {
while let Some(bytes) = writer_rx.recv().await {
let writer = writer.clone();
let _ = tokio::task::spawn_blocking(move || {
if let Ok(mut guard) = writer.lock() {
use std::io::Write;
let _ = guard.write_all(&bytes);
let _ = guard.flush();
}
})
.await;
}
}
});
let exit_status = Arc::new(AtomicBool::new(false));
let wait_exit_status = Arc::clone(&exit_status);
let wait_handle = tokio::task::spawn_blocking(move || {
let _ = child.wait();
wait_exit_status.store(true, Ordering::SeqCst);
});
Ok(ExecCommandSession::new(
writer_tx,
output_tx,
killer,
reader_handle,
writer_handle,
wait_handle,
exit_status,
))
}
#[cfg(test)]
mod tests {
use super::path::parse_command_line;
use super::*;
#[test]
fn parse_command_line_splits_words() {
assert_eq!(
parse_command_line("echo codex").unwrap(),
vec!["echo".to_string(), "codex".to_string()]
);
}
#[test]
fn parse_command_line_trims_whitespace() {
assert_eq!(
parse_command_line(" ls -la \n").unwrap(),
vec!["ls".to_string(), "-la".to_string()]
);
}
#[test]
fn parse_command_line_rejects_empty() {
let err = parse_command_line(" ").expect_err("expected error");
assert!(matches!(err, UnifiedExecError::MissingCommandLine));
}
#[cfg(unix)]
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn unified_exec_persists_across_requests() -> Result<(), UnifiedExecError> {
let manager = UnifiedExecSessionManager::default();
let open_shell = manager
.handle_request(UnifiedExecRequest {
session_id: None,
input_chunks: &["bash".to_string(), "-i".to_string()],
timeout_ms: Some(1_500),
})
.await?;
let session_id = open_shell.session_id.expect("expected session_id");
manager
.handle_request(UnifiedExecRequest {
session_id: Some(session_id),
input_chunks: &["export CODEX_INTERACTIVE_SHELL_VAR=codex\n".to_string()],
timeout_ms: Some(1_500),
})
.await?;
let out_2 = manager
.handle_request(UnifiedExecRequest {
session_id: Some(session_id),
input_chunks: &["echo $CODEX_INTERACTIVE_SHELL_VAR\n".to_string()],
timeout_ms: Some(1_500),
})
.await?;
assert!(out_2.output.contains("codex"));
Ok(())
}
#[cfg(unix)]
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn multi_unified_exec_sessions() -> Result<(), UnifiedExecError> {
let manager = UnifiedExecSessionManager::default();
let shell_a = manager
.handle_request(UnifiedExecRequest {
session_id: None,
input_chunks: &["/bin/bash".to_string(), "-i".to_string()],
timeout_ms: Some(1_500),
})
.await?;
let session_a = shell_a.session_id.expect("expected session id");
manager
.handle_request(UnifiedExecRequest {
session_id: Some(session_a),
input_chunks: &["export CODEX_INTERACTIVE_SHELL_VAR=codex\n".to_string()],
timeout_ms: Some(1_500),
})
.await?;
let out_2 = manager
.handle_request(UnifiedExecRequest {
session_id: None,
input_chunks: &["/bin/echo $CODEX_INTERACTIVE_SHELL_VAR\n".to_string()],
timeout_ms: Some(1_500),
})
.await?;
assert!(!out_2.output.contains("codex"));
let out_3 = manager
.handle_request(UnifiedExecRequest {
session_id: Some(session_a),
input_chunks: &["echo $CODEX_INTERACTIVE_SHELL_VAR\n".to_string()],
timeout_ms: Some(1_500),
})
.await?;
assert!(out_3.output.contains("codex"));
Ok(())
}
#[cfg(unix)]
#[tokio::test]
async fn unified_exec_timeouts() -> Result<(), UnifiedExecError> {
let manager = UnifiedExecSessionManager::default();
let open_shell = manager
.handle_request(UnifiedExecRequest {
session_id: None,
input_chunks: &["/bin/bash".to_string(), "-i".to_string()],
timeout_ms: Some(1_500),
})
.await?;
let session_id = open_shell.session_id.expect("expected session id");
manager
.handle_request(UnifiedExecRequest {
session_id: Some(session_id),
input_chunks: &["export CODEX_INTERACTIVE_SHELL_VAR=codex\n".to_string()],
timeout_ms: Some(1_500),
})
.await?;
let out_2 = manager
.handle_request(UnifiedExecRequest {
session_id: Some(session_id),
input_chunks: &["sleep 5 && echo $CODEX_INTERACTIVE_SHELL_VAR\n".to_string()],
timeout_ms: Some(10),
})
.await?;
assert!(!out_2.output.contains("codex"));
tokio::time::sleep(Duration::from_secs(7)).await;
let empty = Vec::new();
let out_3 = manager
.handle_request(UnifiedExecRequest {
session_id: Some(session_id),
input_chunks: &empty,
timeout_ms: Some(100),
})
.await?;
assert!(out_3.output.contains("codex"));
Ok(())
}
#[cfg(unix)]
#[tokio::test]
async fn requests_with_large_timeout_are_capped() -> Result<(), UnifiedExecError> {
let manager = UnifiedExecSessionManager::default();
let result = manager
.handle_request(UnifiedExecRequest {
session_id: None,
input_chunks: &["/bin/echo".to_string(), "codex".to_string()],
timeout_ms: Some(120_000),
})
.await?;
assert!(result.output.starts_with(
"Warning: requested timeout 120000ms exceeds maximum of 60000ms; clamping to 60000ms.\n"
));
assert!(result.output.contains("codex"));
Ok(())
}
#[cfg(unix)]
#[tokio::test]
async fn completed_commands_do_not_persist_sessions() -> Result<(), UnifiedExecError> {
let manager = UnifiedExecSessionManager::default();
let result = manager
.handle_request(UnifiedExecRequest {
session_id: None,
input_chunks: &["/bin/echo".to_string(), "codex".to_string()],
timeout_ms: Some(1_500),
})
.await?;
assert!(result.session_id.is_none());
assert!(result.output.contains("codex"));
assert!(manager.sessions.lock().await.is_empty());
Ok(())
}
#[cfg(unix)]
#[tokio::test]
async fn correct_path_resolution() -> Result<(), UnifiedExecError> {
let manager = UnifiedExecSessionManager::default();
let result = manager
.handle_request(UnifiedExecRequest {
session_id: None,
input_chunks: &["echo".to_string(), "codex".to_string()],
timeout_ms: Some(1_500),
})
.await?;
assert!(result.session_id.is_none());
assert!(result.output.contains("codex"));
assert!(manager.sessions.lock().await.is_empty());
Ok(())
}
#[cfg(unix)]
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn reusing_completed_session_returns_unknown_session() -> Result<(), UnifiedExecError> {
let manager = UnifiedExecSessionManager::default();
let open_shell = manager
.handle_request(UnifiedExecRequest {
session_id: None,
input_chunks: &["/bin/bash".to_string(), "-i".to_string()],
timeout_ms: Some(1_500),
})
.await?;
let session_id = open_shell.session_id.expect("expected session id");
manager
.handle_request(UnifiedExecRequest {
session_id: Some(session_id),
input_chunks: &["exit\n".to_string()],
timeout_ms: Some(1_500),
})
.await?;
tokio::time::sleep(Duration::from_millis(200)).await;
let err = manager
.handle_request(UnifiedExecRequest {
session_id: Some(session_id),
input_chunks: &[],
timeout_ms: Some(100),
})
.await
.expect_err("expected unknown session error");
match err {
UnifiedExecError::UnknownSessionId { session_id: err_id } => {
assert_eq!(err_id, session_id);
}
other => panic!("expected UnknownSessionId, got {other:?}"),
}
assert!(!manager.sessions.lock().await.contains_key(&session_id));
Ok(())
}
}

View File

@@ -0,0 +1,194 @@
use std::env;
use std::path::Path;
use std::path::PathBuf;
use super::errors::UnifiedExecError;
pub(crate) fn parse_command_line(line: &str) -> Result<Vec<String>, UnifiedExecError> {
let trimmed = line.trim();
if trimmed.is_empty() {
return Err(UnifiedExecError::MissingCommandLine);
}
match parse_command_line_impl(trimmed) {
Some(parts) if !parts.is_empty() => Ok(parts),
_ => Err(UnifiedExecError::InvalidCommandLine {
command_line: trimmed.to_string(),
}),
}
}
#[cfg(not(windows))]
fn parse_command_line_impl(trimmed: &str) -> Option<Vec<String>> {
shlex::split(trimmed)
}
#[cfg(windows)]
fn parse_command_line_impl(trimmed: &str) -> Option<Vec<String>> {
windows_split_command_line(trimmed)
}
#[cfg(windows)]
fn windows_split_command_line(input: &str) -> Option<Vec<String>> {
let mut args = Vec::new();
let mut current = String::new();
let mut in_quotes = false;
let mut backslashes = 0usize;
for ch in input.chars() {
match ch {
'\\' => {
backslashes += 1;
}
'"' => {
current.extend(std::iter::repeat_n('\\', backslashes / 2));
if backslashes % 2 == 0 {
in_quotes = !in_quotes;
} else {
current.push('"');
}
backslashes = 0;
}
c if c.is_whitespace() && !in_quotes => {
current.extend(std::iter::repeat_n('\\', backslashes));
backslashes = 0;
if !current.is_empty() {
args.push(std::mem::take(&mut current));
}
}
other => {
current.extend(std::iter::repeat_n('\\', backslashes));
backslashes = 0;
current.push(other);
}
}
}
if in_quotes {
return None;
}
current.extend(std::iter::repeat_n('\\', backslashes));
if !current.is_empty() {
args.push(current);
}
if args.is_empty() { None } else { Some(args) }
}
pub(crate) fn command_from_chunks(chunks: &[String]) -> Result<Vec<String>, UnifiedExecError> {
match chunks {
[] => Err(UnifiedExecError::MissingCommandLine),
[single] => parse_command_line(single),
_ => Ok(chunks.to_vec()),
}
}
pub(crate) fn join_input_chunks(chunks: &[String]) -> String {
match chunks {
[] => String::new(),
[single] => single.clone(),
_ => chunks.concat(),
}
}
pub(crate) fn resolve_command_path(command: &str) -> Result<String, UnifiedExecError> {
if command.is_empty() {
return Err(UnifiedExecError::MissingCommandLine);
}
if is_explicit_path(command) {
return ensure_executable(Path::new(command))
.then_some(command.to_string())
.ok_or_else(|| UnifiedExecError::CommandNotFound {
command: command.to_string(),
});
}
if let Some(resolved) = find_in_path(command) {
return Ok(resolved.to_string_lossy().to_string());
}
Err(UnifiedExecError::CommandNotFound {
command: command.to_string(),
})
}
fn is_explicit_path(command: &str) -> bool {
let path = Path::new(command);
path.is_absolute() || path.components().count() > 1
}
fn find_in_path(command: &str) -> Option<PathBuf> {
let path_var = env::var_os("PATH")?;
env::split_paths(&path_var)
.flat_map(|dir| candidate_paths(dir, command))
.find(|candidate| ensure_executable(candidate))
}
fn candidate_paths(dir: PathBuf, command: &str) -> Vec<PathBuf> {
build_platform_candidates(dir.join(command))
}
#[cfg(unix)]
fn build_platform_candidates(candidate: PathBuf) -> Vec<PathBuf> {
vec![candidate]
}
#[cfg(windows)]
fn build_platform_candidates(candidate: PathBuf) -> Vec<PathBuf> {
if candidate.extension().is_some() {
return vec![candidate];
}
let pathext = env::var("PATHEXT").unwrap_or_else(|_| ".COM;.EXE;.BAT;.CMD".to_string());
let mut candidates = Vec::new();
for ext in pathext.split(';') {
if ext.is_empty() {
continue;
}
let mut path_with_ext = candidate.clone();
let new_ext = ext.trim_start_matches('.');
path_with_ext.set_extension(new_ext);
candidates.push(path_with_ext);
}
if candidates.is_empty() {
candidates.push(candidate);
}
candidates
}
fn ensure_executable(path: &Path) -> bool {
match path.metadata() {
Ok(metadata) => metadata.is_file() && is_executable(&metadata),
Err(_) => false,
}
}
#[cfg(unix)]
fn is_executable(metadata: &std::fs::Metadata) -> bool {
use std::os::unix::fs::PermissionsExt;
metadata.permissions().mode() & 0o111 != 0
}
#[cfg(not(unix))]
fn is_executable(metadata: &std::fs::Metadata) -> bool {
metadata.is_file()
}
#[cfg(all(test, windows))]
mod tests {
use super::parse_command_line;
#[test]
fn parses_windows_command_with_spaces() {
let out = parse_command_line(r#""C:\Program Files\Git\bin\bash.exe" -i"#).unwrap();
assert_eq!(
out,
vec![
"C:\\Program Files\\Git\\bin\\bash.exe".to_string(),
"-i".to_string(),
]
);
}
}

View File

@@ -191,7 +191,13 @@ async fn prompt_tools_are_consistent_across_requests() {
let expected_instructions: &str = include_str!("../../prompt.md");
// our internal implementation is responsible for keeping tools in sync
// with the OpenAI schema, so we just verify the tool presence here
let expected_tools_names: &[&str] = &["shell", "update_plan", "apply_patch", "view_image"];
let expected_tools_names: &[&str] = &[
"shell",
"unified_exec",
"update_plan",
"apply_patch",
"view_image",
];
let body0 = requests[0].body_json::<serde_json::Value>().unwrap();
assert_eq!(
body0["instructions"],

View File

@@ -114,6 +114,19 @@ pub enum ResponseItem {
status: Option<String>,
action: WebSearchAction,
},
UnifiedExec {
#[serde(skip_serializing)]
id: Option<String>,
/// Identifier for the interactive shell session. When absent, a new
/// session should be created.
session_id: Option<String>,
/// Characters that should be written to the interactive shell's
/// standard input.
arguments: Vec<String>,
/// Maximum amount of time to wait for additional output after writing
/// to stdin.
timeout_ms: Option<u64>,
},
#[serde(other)]
Other,