Files
codex/codex-rs/ext/agent/tests/agent_service.rs
jif 6629e08702 Add an agent extension runner (#33076)
## What changed

- Add `codex-agent-extension` with an `AgentRunner` that starts a resolved agent prompt in a thread forked from its parent.
- Propagate the invocation's trace context, select the configured execution environments, submit the initial prompt, and return the spawned thread and turn identifiers.
- Reject empty agent prompts and report when the owning thread manager is no longer available.

## Testing

- Add an integration test that verifies the agent runs in a forked thread, returns the started turn identifier, completes the turn, and sends the resolved prompt to the model.

GitOrigin-RevId: ffb805efabadf758969d6611a6867db6fa9f059a
2026-07-14 14:41:30 +00:00

71 lines
2.2 KiB
Rust

use anyhow::Result;
use codex_agent_extension::AgentInvocation;
use codex_agent_extension::AgentRunner;
use codex_protocol::protocol::EventMsg;
use core_test_support::responses;
use core_test_support::skip_if_no_network;
use core_test_support::test_codex::test_codex;
use core_test_support::wait_for_event;
use pretty_assertions::assert_eq;
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn starts_resolved_agent_prompt_in_forked_thread() -> Result<()> {
skip_if_no_network!(Ok(()));
let server = responses::start_mock_server().await;
let response_mock = responses::mount_sse_once(
&server,
responses::sse(vec![
responses::ev_response_created("agent-response"),
responses::ev_completed("agent-response"),
]),
)
.await;
let test = test_codex().build_with_auto_env(&server).await?;
let parent_thread_id = test.session_configured.session_id.into();
let agent_runner = AgentRunner::new(std::sync::Arc::downgrade(&test.thread_manager));
let agent_run = agent_runner
.start(
parent_thread_id,
AgentInvocation {
config: test.config.clone(),
prompt: "Use $example-agent to inspect the current changes.".to_string(),
parent_trace: None,
},
)
.await?;
assert_ne!(agent_run.thread_id, parent_thread_id);
assert_eq!(
agent_run
.thread
.config_snapshot()
.await
.forked_from_thread_id,
Some(parent_thread_id)
);
let started = wait_for_event(&agent_run.thread, |event| {
matches!(event, EventMsg::TurnStarted(_))
})
.await;
let EventMsg::TurnStarted(started) = started else {
unreachable!("event predicate only matches turn started events");
};
assert_eq!(started.turn_id, agent_run.turn_id);
wait_for_event(&agent_run.thread, |event| {
matches!(event, EventMsg::TurnComplete(_))
})
.await;
let request = response_mock.single_request();
assert!(
request
.message_input_texts("user")
.iter()
.any(|text| text == "Use $example-agent to inspect the current changes.")
);
Ok(())
}