Files
codex/codex-rs/core/tests/suite/resume_warning.rs
T
Shijie RaoandGitHub fc8c6b7384 Add per-turn multi-agent mode (#28685)
## Why

Multi-agent v2 currently carries an explicit-request-only delegation
rule in its static usage hint. That provides a safe default, but it
prevents clients from selecting proactive delegation per turn without
changing static guidance or rewriting prior model context.

This change makes delegation mode a session selection that can be
updated through `turn/start`, while deriving the effective model-visible
mode separately for each turn. Eligible multi-agent v2 turns remain
explicit-request-only unless proactive mode is both selected and
enabled.

## What changed

- Add the experimental `turn/start.multiAgentMode` parameter with
`explicitRequestOnly` and `proactive` values. Omission retains the
loaded session's current optional selection.
- Add the default-off `features.multi_agent_mode` feature gate. Eligible
multi-agent v2 turns use the selected mode when enabled; an unset
selection or disabled gate resolves to `explicitRequestOnly`.
- Treat mode prompting as inapplicable for multi-agent v1 and other
unsupported session configurations, producing no multi-agent mode
developer message rather than rejecting the turn.
- Move the explicit-request-only rule out of the static v2 usage hint
and into a bounded, tagged developer context fragment.
- Emit the effective mode in initial context and only when that
effective mode changes on later turns.
- Persist the effective mode in `TurnContextItem` as the durable
baseline for resume and context-update comparisons.

Historical rollout items are not rewritten. Later mode developer
messages establish the current rule incrementally.

## Not covered

- Initial selection through `thread/start` and selected-mode reporting
from thread lifecycle/settings APIs; those are isolated in the stacked
#28792.
- A TUI control or slash command for selecting the mode.
- Persisting a preferred mode to `config.toml`; selection remains
session/turn scoped.
- Changes to multi-agent concurrency limits, tool availability, or model
catalog capability declarations.
- Rewriting historical rollout prompt items. Cold resume restores the
latest persisted effective mode when available while leaving historical
developer messages intact.

## Verification

- `CARGO_INCREMENTAL=0 just test -p codex-core multi_agent_mode`
- Focused app-server coverage verifies that `turn/start.multiAgentMode`
produces proactive developer instructions for an eligible v2 turn.

## Stack

Followed by #28792, which adds `thread/start` initialization and
lifecycle/settings observability.
2026-06-18 22:47:51 -07:00

138 lines
5.0 KiB
Rust

#![allow(clippy::unwrap_used)]
use codex_core::NewThread;
use codex_login::CodexAuth;
use codex_protocol::ThreadId;
use codex_protocol::config_types::ModeKind;
use codex_protocol::config_types::ReasoningSummary;
use codex_protocol::protocol::EventMsg;
use codex_protocol::protocol::InitialHistory;
use codex_protocol::protocol::ResumedHistory;
use codex_protocol::protocol::RolloutItem;
use codex_protocol::protocol::TurnCompleteEvent;
use codex_protocol::protocol::TurnContextItem;
use codex_protocol::protocol::TurnStartedEvent;
use codex_protocol::protocol::UserMessageEvent;
use codex_protocol::protocol::WarningEvent;
use core::time::Duration;
use core_test_support::load_default_config_for_test;
use core_test_support::wait_for_event;
use tempfile::TempDir;
fn resume_history(
config: &codex_core::config::Config,
previous_model: &str,
rollout_path: &std::path::Path,
) -> InitialHistory {
let turn_id = "resume-warning-seed-turn".to_string();
let turn_ctx = TurnContextItem {
turn_id: Some(turn_id.clone()),
cwd: config.cwd.clone(),
workspace_roots: None,
current_date: None,
timezone: None,
approval_policy: config.permissions.approval_policy.value(),
sandbox_policy: config.legacy_sandbox_policy(),
permission_profile: None,
network: None,
file_system_sandbox_policy: None,
model: previous_model.to_string(),
comp_hash: None,
personality: None,
collaboration_mode: None,
multi_agent_version: None,
multi_agent_mode: None,
realtime_active: None,
effort: config.model_reasoning_effort.clone(),
summary: config
.model_reasoning_summary
.unwrap_or(ReasoningSummary::Auto),
};
InitialHistory::Resumed(ResumedHistory {
conversation_id: ThreadId::default(),
history: vec![
RolloutItem::EventMsg(EventMsg::TurnStarted(TurnStartedEvent {
turn_id: turn_id.clone(),
trace_id: None,
started_at: None,
model_context_window: None,
collaboration_mode_kind: ModeKind::Default,
})),
RolloutItem::EventMsg(EventMsg::UserMessage(UserMessageEvent {
client_id: None,
message: "seed".to_string(),
images: None,
local_images: vec![],
text_elements: vec![],
..Default::default()
})),
RolloutItem::TurnContext(turn_ctx),
RolloutItem::EventMsg(EventMsg::TurnComplete(TurnCompleteEvent {
turn_id,
last_agent_message: None,
completed_at: None,
duration_ms: None,
time_to_first_token_ms: None,
})),
],
rollout_path: Some(rollout_path.to_path_buf()),
})
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn emits_warning_when_resumed_model_differs() {
// Arrange a config with a current model and a prior rollout recorded under a different model.
let home = TempDir::new().expect("tempdir");
let mut config = load_default_config_for_test(&home).await;
config.model = Some("current-model".to_string());
// Ensure cwd is absolute (the helper sets it to the temp dir already).
assert!(config.cwd.is_absolute());
let rollout_path = home.path().join("rollout.jsonl");
std::fs::write(&rollout_path, "").expect("create rollout placeholder");
let initial_history = resume_history(&config, "previous-model", &rollout_path);
let thread_manager = codex_core::test_support::thread_manager_with_models_provider(
CodexAuth::from_api_key("test"),
config.model_provider.clone(),
);
let auth_manager =
codex_core::test_support::auth_manager_from_auth(CodexAuth::from_api_key("test"));
// Act: resume the conversation.
let NewThread {
thread: conversation,
..
} = thread_manager
.resume_thread_with_history(
config.clone(),
initial_history,
auth_manager,
/*parent_trace*/ None,
/*supports_openai_form_elicitation*/ false,
)
.await
.expect("resume conversation");
// Assert: a Warning event is emitted describing the model mismatch.
let warning = wait_for_event(&conversation, |ev| {
matches!(
ev,
EventMsg::Warning(WarningEvent { message })
if message.contains("previous-model") && message.contains("current-model")
)
})
.await;
let EventMsg::Warning(WarningEvent { message }) = warning else {
panic!("expected warning event");
};
assert!(message.contains("previous-model"));
assert!(message.contains("current-model"));
// Drain the TurnComplete/Shutdown window to avoid leaking tasks between tests.
// The warning is emitted during initialization, so a short sleep is sufficient.
tokio::time::sleep(Duration::from_millis(50)).await;
}