clone/codex
1
0
Fork 0
mirror of https://github.com/openai/codex.git synced 2026-04-29 00:55:38 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add MCP tool wall time to model output (#17406)

Browse Source 
Include MCP wall time in the output so the model is aware of how long
it's calls are taking.
This commit is contained in:
pakrym-oai
2026-04-12 18:26:15 -07:00
committed by GitHub
parent 68a1d82a41
commit 7c1e41c8b6
6 changed files with 286 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files

61
codex-rs/core/tests/suite/rmcp_client.rs
View File

@@ -30,6 +30,7 @@ use codex_protocol::protocol::Op;
use codex_protocol::protocol::SandboxPolicy;
use codex_protocol::user_input::UserInput;
use codex_utils_cargo_bin::cargo_bin;
use core_test_support::assert_regex_match;
use core_test_support::responses;
use core_test_support::responses::mount_models_once;
use core_test_support::responses::mount_sse_once;
@@ -51,6 +52,29 @@ use tokio::time::sleep;
static OPENAI_PNG: &str = "data:image/png;base64,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";
fn assert_wall_time_line(line: &str) {
assert_regex_match(r"^Wall time: [0-9]+(?:\.[0-9]+)? seconds$", line);
}
fn split_wall_time_wrapped_output(output: &str) -> &str {
let Some((wall_time, rest)) = output.split_once('\n') else {
panic!("wall-time output should contain an Output section: {output}");
};
assert_wall_time_line(wall_time);
let Some(output) = rest.strip_prefix("Output:\n") else {
panic!("wall-time output should contain Output marker: {output}");
};
output
}
fn assert_wall_time_header(output: &str) {
let Some((wall_time, marker)) = output.split_once('\n') else {
panic!("wall-time header should contain an Output marker: {output}");
};
assert_wall_time_line(wall_time);
assert_eq!(marker, "Output:");
}
#[tokio::test(flavor = "multi_thread", worker_threads = 1)]
#[serial(mcp_test_value)]
async fn stdio_server_round_trip() -> anyhow::Result<()> {
@@ -71,7 +95,7 @@ async fn stdio_server_round_trip() -> anyhow::Result<()> {
]),
)
.await;
mount_sse_once(
let final_mock = mount_sse_once(
&server,
responses::sse(vec![
responses::ev_assistant_message("msg-1", "rmcp echo tool completed successfully."),
@@ -190,6 +214,17 @@ async fn stdio_server_round_trip() -> anyhow::Result<()> {
wait_for_event(&fixture.codex, |ev| matches!(ev, EventMsg::TurnComplete(_))).await;
let output_item = final_mock.single_request().function_call_output(call_id);
let output_text = output_item
.get("output")
.and_then(Value::as_str)
.expect("function_call_output output should be a string");
let wrapped_payload = split_wall_time_wrapped_output(output_text);
let output_json: Value = serde_json::from_str(wrapped_payload)
.expect("wrapped MCP output should preserve structured JSON");
assert_eq!(output_json["echo"], "ECHOING: ping");
assert_eq!(output_json["env"], expected_env_value);
server.verify().await;
Ok(())
@@ -362,15 +397,22 @@ async fn stdio_image_responses_round_trip() -> anyhow::Result<()> {
wait_for_event(&fixture.codex, |ev| matches!(ev, EventMsg::TurnComplete(_))).await;
let output_item = final_mock.single_request().function_call_output(call_id);
assert_eq!(output_item["type"], "function_call_output");
assert_eq!(output_item["call_id"], call_id);
let output = output_item["output"]
.as_array()
.expect("image MCP output should be content items");
assert_eq!(output.len(), 2);
assert_wall_time_header(
output[0]["text"]
.as_str()
.expect("first MCP image output item should be wall-time text"),
);
assert_eq!(
output_item,
output[1],
json!({
"type": "function_call_output",
"call_id": call_id,
"output": [{
"type": "input_image",
"image_url": OPENAI_PNG
}]
"type": "input_image",
"image_url": OPENAI_PNG
})
);
server.verify().await;
@@ -533,7 +575,8 @@ async fn stdio_image_responses_are_sanitized_for_text_only_model() -> anyhow::Re
.get("output")
.and_then(Value::as_str)
.expect("function_call_output output should be a JSON string");
let output_json: Value = serde_json::from_str(output_text)
let wrapped_payload = split_wall_time_wrapped_output(output_text);
let output_json: Value = serde_json::from_str(wrapped_payload)
.expect("function_call_output output should be valid JSON");
assert_eq!(
output_json,

38
codex-rs/core/tests/suite/truncation.rs
View File

@@ -28,6 +28,14 @@ use serde_json::json;
use std::collections::HashMap;
use std::time::Duration;
fn assert_wall_time_header(output: &str) {
let (wall_time, marker) = output
.split_once('\n')
.expect("wall-time header should contain an Output marker");
assert_regex_match(r"^Wall time: [0-9]+(?:\.[0-9]+)? seconds$", wall_time);
assert_eq!(marker, "Output:");
}
// Verifies that a standard tool call (shell_command) exceeding the model formatting
// limits is truncated before being sent back to the model.
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
@@ -400,9 +408,9 @@ async fn mcp_tool_call_output_exceeds_limit_truncated_for_model() -> Result<()>
"MCP output should not include line-based truncation header: {output}"
);
let truncated_pattern = r#"(?s)^\{"echo":\s*"ECHOING: long-message-with-newlines-.*tokens truncated.*long-message-with-newlines-.*$"#;
let truncated_pattern = r#"(?s)^Wall time: [0-9]+(?:\.[0-9]+)? seconds\nOutput:\n\{"echo":\s*"ECHOING: long-message-with-newlines-.*tokens truncated.*long-message-with-newlines-.*$"#;
assert_regex_match(truncated_pattern, &output);
assert!(output.len() < 2500, "{}", output.len());
assert!(output.len() < 2600, "{}", output.len());
Ok(())
}
@@ -502,13 +510,18 @@ async fn mcp_image_output_preserves_image_and_no_text_summary() -> Result<()> {
// Wait for completion to ensure the outbound request is captured.
wait_for_event(&fixture.codex, |ev| matches!(ev, EventMsg::TurnComplete(_))).await;
let output_item = final_mock.single_request().function_call_output(call_id);
// Expect exactly one array element: the image item; and no trailing summary text.
// Expect exactly the wall-time text and image item; no trailing truncation summary.
let output = output_item.get("output").expect("output");
assert!(output.is_array(), "expected array output");
let arr = output.as_array().unwrap();
assert_eq!(arr.len(), 1, "no truncation summary should be appended");
assert_eq!(arr.len(), 2, "no truncation summary should be appended");
assert_wall_time_header(
arr[0]["text"]
.as_str()
.expect("first MCP image output item should be wall-time text"),
);
assert_eq!(
arr[0],
arr[1],
json!({"type": "input_image", "image_url": openai_png})
);
@@ -758,22 +771,11 @@ async fn mcp_tool_call_output_not_truncated_with_custom_limit() -> Result<()> {
.function_call_output_text(call_id)
.context("function_call_output present for rmcp call")?;
let parsed: Value = serde_json::from_str(&output)?;
assert_eq!(
output.len(),
80031,
"parsed MCP output should retain its serialized length"
80065,
"MCP output should retain its serialized length plus wall-time header"
);
let expected_echo = format!("ECHOING: {large_msg}");
let echo_str = parsed["echo"]
.as_str()
.context("echo field should be a string in rmcp echo output")?;
assert_eq!(
echo_str.len(),
expected_echo.len(),
"echo length should match"
);
assert_eq!(echo_str, expected_echo);
assert!(
!output.contains("truncated"),
"output should not include truncation markers when limit is raised: {output}"