feat(app-server): thread/rollback API (#8454)

Add `thread/rollback` to app-server to support IDEs undo-ing the last N
turns of a thread.

For context, an IDE partner will be supporting an "undo" capability
where the IDE (the app-server client) will be responsible for reverting
the local changes made during the last turn. To support this well, we
also need a way to drop the last turn (or more generally, the last N
turns) from the agent's context. This is what `thread/rollback` does.

**Core idea**: A Thread rollback is represented as a persisted event
message (EventMsg::ThreadRollback) in the rollout JSONL file, not by
rewriting history. On resume, both the model's context (core replay) and
the UI turn list (app-server v2's thread history builder) apply these
markers so the pruned history is consistent across live conversations
and `thread/resume`.

Implementation notes:
- Rollback only affects agent context and appends to the rollout file;
clients are responsible for reverting files on disk.
- If a thread rollback is currently in progress, subsequent
`thread/rollback` calls are rejected.
- Because we use `CodexConversation::submit` and codex core tracks
active turns, returning an error on concurrent rollbacks is communicated
via an `EventMsg::Error` with a new variant
`CodexErrorInfo::ThreadRollbackFailed`. app-server watches for that and
sends the BAD_REQUEST RPC response.

Tests cover thread rollbacks in both core and app-server, including when
`num_turns` > existing turns (which clears all turns).

**Note**: this explicitly does **not** behave like `/undo` which we just
removed from the CLI, which does the opposite of what `thread/rollback`
does. `/undo` reverts local changes via ghost commits/snapshots and does
not modify the agent's context / conversation history.
This commit is contained in:
Owen Lin
2026-01-06 13:23:48 -08:00
committed by GitHub
parent 188f79afee
commit 8b7ec31ba7
21 changed files with 1187 additions and 30 deletions

View File

@@ -6,6 +6,7 @@ use crate::protocol::v2::UserInput;
use codex_protocol::protocol::AgentReasoningEvent;
use codex_protocol::protocol::AgentReasoningRawContentEvent;
use codex_protocol::protocol::EventMsg;
use codex_protocol::protocol::ThreadRolledBackEvent;
use codex_protocol::protocol::TurnAbortedEvent;
use codex_protocol::protocol::UserMessageEvent;
@@ -57,6 +58,7 @@ impl ThreadHistoryBuilder {
EventMsg::TokenCount(_) => {}
EventMsg::EnteredReviewMode(_) => {}
EventMsg::ExitedReviewMode(_) => {}
EventMsg::ThreadRolledBack(payload) => self.handle_thread_rollback(payload),
EventMsg::UndoCompleted(_) => {}
EventMsg::TurnAborted(payload) => self.handle_turn_aborted(payload),
_ => {}
@@ -130,6 +132,23 @@ impl ThreadHistoryBuilder {
turn.status = TurnStatus::Interrupted;
}
fn handle_thread_rollback(&mut self, payload: &ThreadRolledBackEvent) {
self.finish_current_turn();
let n = usize::try_from(payload.num_turns).unwrap_or(usize::MAX);
if n >= self.turns.len() {
self.turns.clear();
} else {
self.turns.truncate(self.turns.len().saturating_sub(n));
}
// Re-number subsequent synthetic ids so the pruned history is consistent.
self.next_turn_index =
i64::try_from(self.turns.len().saturating_add(1)).unwrap_or(i64::MAX);
let item_count: usize = self.turns.iter().map(|t| t.items.len()).sum();
self.next_item_index = i64::try_from(item_count.saturating_add(1)).unwrap_or(i64::MAX);
}
fn finish_current_turn(&mut self) {
if let Some(turn) = self.current_turn.take() {
if turn.items.is_empty() {
@@ -213,6 +232,7 @@ mod tests {
use codex_protocol::protocol::AgentMessageEvent;
use codex_protocol::protocol::AgentReasoningEvent;
use codex_protocol::protocol::AgentReasoningRawContentEvent;
use codex_protocol::protocol::ThreadRolledBackEvent;
use codex_protocol::protocol::TurnAbortReason;
use codex_protocol::protocol::TurnAbortedEvent;
use codex_protocol::protocol::UserMessageEvent;
@@ -410,4 +430,95 @@ mod tests {
}
);
}
#[test]
fn drops_last_turns_on_thread_rollback() {
let events = vec![
EventMsg::UserMessage(UserMessageEvent {
message: "First".into(),
images: None,
}),
EventMsg::AgentMessage(AgentMessageEvent {
message: "A1".into(),
}),
EventMsg::UserMessage(UserMessageEvent {
message: "Second".into(),
images: None,
}),
EventMsg::AgentMessage(AgentMessageEvent {
message: "A2".into(),
}),
EventMsg::ThreadRolledBack(ThreadRolledBackEvent { num_turns: 1 }),
EventMsg::UserMessage(UserMessageEvent {
message: "Third".into(),
images: None,
}),
EventMsg::AgentMessage(AgentMessageEvent {
message: "A3".into(),
}),
];
let turns = build_turns_from_event_msgs(&events);
let expected = vec![
Turn {
id: "turn-1".into(),
status: TurnStatus::Completed,
error: None,
items: vec![
ThreadItem::UserMessage {
id: "item-1".into(),
content: vec![UserInput::Text {
text: "First".into(),
}],
},
ThreadItem::AgentMessage {
id: "item-2".into(),
text: "A1".into(),
},
],
},
Turn {
id: "turn-2".into(),
status: TurnStatus::Completed,
error: None,
items: vec![
ThreadItem::UserMessage {
id: "item-3".into(),
content: vec![UserInput::Text {
text: "Third".into(),
}],
},
ThreadItem::AgentMessage {
id: "item-4".into(),
text: "A3".into(),
},
],
},
];
assert_eq!(turns, expected);
}
#[test]
fn thread_rollback_clears_all_turns_when_num_turns_exceeds_history() {
let events = vec![
EventMsg::UserMessage(UserMessageEvent {
message: "One".into(),
images: None,
}),
EventMsg::AgentMessage(AgentMessageEvent {
message: "A1".into(),
}),
EventMsg::UserMessage(UserMessageEvent {
message: "Two".into(),
images: None,
}),
EventMsg::AgentMessage(AgentMessageEvent {
message: "A2".into(),
}),
EventMsg::ThreadRolledBack(ThreadRolledBackEvent { num_turns: 99 }),
];
let turns = build_turns_from_event_msgs(&events);
assert_eq!(turns, Vec::<Turn>::new());
}
}