Surface misalignment policy violations as typed errors (#38682)

## What changed

- Recognize `misalignment_policy_violation` errors from response streams and HTTP 400 or 403 responses.
- Preserve the upstream message, use a fallback for blank messages, and treat the error as non-retryable.
- Expose `misalignmentPolicyViolation` through the app-server protocol and generated schemas so turns fail with a typed terminal error.

## Testing

- Cover streamed and HTTP policy violations, fallback messages, retry behavior, and app-server turn completion.

GitOrigin-RevId: fd3485bf0be7bfe3d51c078bbc36a081692fd57f
This commit is contained in:
Francis Chalissery
2026-08-15 01:34:33 +00:00
committed by copyberry
parent 4e9a1a9073
commit eb147c0db3
32 changed files with 299 additions and 2 deletions
@@ -606,6 +606,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -7573,6 +7573,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -3691,6 +3691,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -11,6 +11,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -34,6 +34,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -115,6 +115,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -37,6 +37,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -37,6 +37,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -37,6 +37,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -115,6 +115,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -37,6 +37,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -115,6 +115,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -37,6 +37,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -37,6 +37,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -34,6 +34,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -34,6 +34,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -34,6 +34,7 @@
"usageLimitExceeded",
"serverOverloaded",
"cyberPolicy",
"misalignmentPolicyViolation",
"internalServerError",
"unauthorized",
"badRequest",
@@ -9,4 +9,4 @@ import type { NonSteerableTurnKind } from "./NonSteerableTurnKind";
* When an upstream HTTP status is available (for example, from the Responses API or a provider),
* it is forwarded in `httpStatusCode` on the relevant `codexErrorInfo` variant.
*/
export type CodexErrorInfo = "contextWindowExceeded" | "sessionBudgetExceeded" | "usageLimitExceeded" | "serverOverloaded" | "cyberPolicy" | { "httpConnectionFailed": { httpStatusCode: number | null, } } | { "responseStreamConnectionFailed": { httpStatusCode: number | null, } } | "internalServerError" | "unauthorized" | "badRequest" | "threadRollbackFailed" | "sandboxError" | { "responseStreamDisconnected": { httpStatusCode: number | null, } } | { "responseTooManyFailedAttempts": { httpStatusCode: number | null, } } | { "activeTurnNotSteerable": { turnKind: NonSteerableTurnKind, } } | "other";
export type CodexErrorInfo = "contextWindowExceeded" | "sessionBudgetExceeded" | "usageLimitExceeded" | "serverOverloaded" | "cyberPolicy" | "misalignmentPolicyViolation" | { "httpConnectionFailed": { httpStatusCode: number | null, } } | { "responseStreamConnectionFailed": { httpStatusCode: number | null, } } | "internalServerError" | "unauthorized" | "badRequest" | "threadRollbackFailed" | "sandboxError" | { "responseStreamDisconnected": { httpStatusCode: number | null, } } | { "responseTooManyFailedAttempts": { httpStatusCode: number | null, } } | { "activeTurnNotSteerable": { turnKind: NonSteerableTurnKind, } } | "other";
@@ -80,6 +80,7 @@ pub enum CodexErrorInfo {
UsageLimitExceeded,
ServerOverloaded,
CyberPolicy,
MisalignmentPolicyViolation,
HttpConnectionFailed {
#[serde(rename = "httpStatusCode")]
#[ts(rename = "httpStatusCode")]
@@ -126,6 +127,9 @@ impl From<CoreCodexErrorInfo> for CodexErrorInfo {
CoreCodexErrorInfo::UsageLimitExceeded => CodexErrorInfo::UsageLimitExceeded,
CoreCodexErrorInfo::ServerOverloaded => CodexErrorInfo::ServerOverloaded,
CoreCodexErrorInfo::CyberPolicy => CodexErrorInfo::CyberPolicy,
CoreCodexErrorInfo::MisalignmentPolicyViolation => {
CodexErrorInfo::MisalignmentPolicyViolation
}
CoreCodexErrorInfo::HttpConnectionFailed { http_status_code } => {
CodexErrorInfo::HttpConnectionFailed { http_status_code }
}
+1
View File
@@ -1662,6 +1662,7 @@ There are additional item-specific events:
- `ContextWindowExceeded`
- `SessionBudgetExceeded`
- `UsageLimitExceeded`
- `misalignmentPolicyViolation`: a non-retryable request blocked by the misalignment policy
- `HttpConnectionFailed { httpStatusCode? }`: upstream HTTP failures including 4xx/5xx
- `ResponseStreamConnectionFailed { httpStatusCode? }`: failure to connect to the response SSE stream
- `ResponseStreamDisconnected { httpStatusCode? }`: disconnect of the response SSE stream in the middle of a turn before completion
@@ -0,0 +1,149 @@
use anyhow::Result;
use app_test_support::MockResponsesConfig;
use app_test_support::TestAppServer;
use codex_app_server_protocol::ClientRequest;
use codex_app_server_protocol::CodexErrorInfo;
use codex_app_server_protocol::ErrorNotification;
use codex_app_server_protocol::ThreadStartParams;
use codex_app_server_protocol::ThreadStartResponse;
use codex_app_server_protocol::TurnCompletedNotification;
use codex_app_server_protocol::TurnError;
use codex_app_server_protocol::TurnStartParams;
use codex_app_server_protocol::TurnStartResponse;
use codex_app_server_protocol::TurnStatus;
use codex_app_server_protocol::UserInput;
use core_test_support::responses;
use core_test_support::skip_if_no_network;
use pretty_assertions::assert_eq;
use tempfile::TempDir;
use tokio::time::timeout;
use wiremock::ResponseTemplate;
const MESSAGE: &str = "This request violated the misalignment policy.";
#[tokio::test]
async fn streamed_policy_violation_completes_turn_with_typed_terminal_error() -> Result<()> {
skip_if_no_network!(Ok(()));
let response = responses::sse_response(responses::sse(vec![
responses::ev_response_created("resp-misalignment"),
serde_json::json!({
"type": "response.failed",
"response": {
"id": "resp-misalignment",
"status": "failed",
"error": {
"type": "invalid_request_error",
"code": "misalignment_policy_violation",
"message": MESSAGE
}
}
}),
]));
assert_policy_violation_completes_turn_with_typed_terminal_error(response).await
}
#[tokio::test]
async fn http_400_policy_violation_completes_turn_with_typed_terminal_error() -> Result<()> {
skip_if_no_network!(Ok(()));
let response = ResponseTemplate::new(400).set_body_json(serde_json::json!({
"error": {
"type": "invalid_request_error",
"code": "misalignment_policy_violation",
"message": MESSAGE,
}
}));
assert_policy_violation_completes_turn_with_typed_terminal_error(response).await
}
#[tokio::test]
async fn http_403_policy_violation_completes_turn_with_typed_terminal_error() -> Result<()> {
skip_if_no_network!(Ok(()));
let response = ResponseTemplate::new(403).set_body_json(serde_json::json!({
"error": {
"type": "invalid_request_error",
"code": "misalignment_policy_violation",
"message": MESSAGE,
}
}));
assert_policy_violation_completes_turn_with_typed_terminal_error(response).await
}
async fn assert_policy_violation_completes_turn_with_typed_terminal_error(
response: ResponseTemplate,
) -> Result<()> {
let server = responses::start_mock_server().await;
let response_mock = responses::mount_response_once(&server, response).await;
let codex_home = TempDir::new()?;
MockResponsesConfig::new(&server.uri()).write(codex_home.path())?;
let mut app_server = TestAppServer::builder()
.with_codex_home(codex_home.path())
.build_initialized()
.await?;
let ThreadStartResponse { thread, .. } = app_server
.start_thread(ThreadStartParams {
model: Some("mock-model".to_string()),
..Default::default()
})
.await?;
let TurnStartResponse { turn } = app_server
.request(|request_id| ClientRequest::TurnStart {
request_id,
params: TurnStartParams {
thread_id: thread.id.clone(),
client_user_message_id: None,
input: vec![UserInput::Text {
text: "trigger policy violation".to_string(),
text_elements: Vec::new(),
}],
..Default::default()
},
})
.await?;
let error: ErrorNotification = timeout(
std::time::Duration::from_secs(10),
app_server.read_notification("error"),
)
.await??;
assert_eq!(
error,
ErrorNotification {
error: TurnError {
message: MESSAGE.to_string(),
codex_error_info: Some(CodexErrorInfo::MisalignmentPolicyViolation),
additional_details: None,
},
will_retry: false,
thread_id: thread.id.clone(),
turn_id: turn.id.clone(),
}
);
let completed: TurnCompletedNotification = timeout(
std::time::Duration::from_secs(10),
app_server.read_notification("turn/completed"),
)
.await??;
assert_eq!(completed.thread_id, thread.id);
assert_eq!(completed.turn.id, turn.id);
assert_eq!(completed.turn.status, TurnStatus::Failed);
assert_eq!(
completed.turn.error,
Some(TurnError {
message: MESSAGE.to_string(),
codex_error_info: Some(CodexErrorInfo::MisalignmentPolicyViolation),
additional_details: None,
})
);
response_mock.single_request();
Ok(())
}
@@ -45,6 +45,7 @@ mod mcp_server_elicitation;
mod mcp_server_status;
mod mcp_tool;
mod memory_reset;
mod misalignment_policy;
mod model_auto_review;
mod model_list;
mod model_provider_capabilities_read;
+1
View File
@@ -2493,6 +2493,7 @@ fn websocket_error_detail(err: &ApiError) -> String {
| ApiError::RateLimit(_)
| ApiError::InvalidRequest { .. }
| ApiError::CyberPolicy { .. }
| ApiError::MisalignmentPolicyViolation { .. }
| ApiError::ServerOverloaded => format!("handshake error: {err}"),
}
}
+26
View File
@@ -48,6 +48,9 @@ pub fn map_api_error(err: ApiError) -> CodexErr {
ApiError::CyberPolicy { message } => {
CodexErr::new(CodexErrorDetails::CyberPolicy { message })
}
ApiError::MisalignmentPolicyViolation { message } => {
CodexErr::new(CodexErrorDetails::MisalignmentPolicyViolation { message })
}
ApiError::Transport(transport) => match transport {
TransportError::Http {
status,
@@ -70,6 +73,26 @@ pub fn map_api_error(err: ApiError) -> CodexErr {
return CodexErr::ServerOverloaded;
}
if (status == http::StatusCode::BAD_REQUEST
|| status == http::StatusCode::FORBIDDEN)
&& let Ok(parsed) = serde_json::from_str::<Value>(&body_text)
&& let Some(error) = parsed.get("error")
&& error.get("code").and_then(Value::as_str)
== Some(MISALIGNMENT_POLICY_VIOLATION_ERROR_CODE)
{
let message = error
.get("message")
.and_then(Value::as_str)
.filter(|message| !message.trim().is_empty())
.map(str::to_string)
.unwrap_or_else(|| {
MISALIGNMENT_POLICY_VIOLATION_FALLBACK_MESSAGE.to_string()
});
return CodexErr::new(CodexErrorDetails::MisalignmentPolicyViolation {
message,
});
}
if status == http::StatusCode::BAD_REQUEST {
if let Ok(parsed) = serde_json::from_str::<Value>(&body_text)
&& let Some(error) = parsed.get("error")
@@ -167,6 +190,9 @@ const X_ERROR_JSON_HEADER: &str = "x-error-json";
const CYBER_POLICY_ERROR_CODE: &str = "cyber_policy";
const CYBER_POLICY_FALLBACK_MESSAGE: &str =
"This request has been flagged for possible cybersecurity risk.";
const MISALIGNMENT_POLICY_VIOLATION_ERROR_CODE: &str = "misalignment_policy_violation";
const MISALIGNMENT_POLICY_VIOLATION_FALLBACK_MESSAGE: &str =
"This request was blocked due to a misalignment policy violation.";
const CLOUDFLARE_BLOCKED_MESSAGE: &str =
"Access blocked by Cloudflare. This usually happens when connecting from a restricted region";
@@ -146,6 +146,39 @@ fn map_api_error_uses_cyber_policy_fallback_for_missing_message() {
);
}
#[test]
fn map_api_error_maps_misalignment_policy_violation_from_400_body() {
assert_misalignment_policy_violation_from_http_body(http::StatusCode::BAD_REQUEST);
}
#[test]
fn map_api_error_maps_misalignment_policy_violation_from_403_body() {
assert_misalignment_policy_violation_from_http_body(http::StatusCode::FORBIDDEN);
}
fn assert_misalignment_policy_violation_from_http_body(status: http::StatusCode) {
let body = serde_json::json!({
"error": {
"message": "This request violated the misalignment policy.",
"type": "invalid_request_error",
"code": "misalignment_policy_violation"
}
})
.to_string();
let err = map_api_error(ApiError::Transport(TransportError::Http {
status,
url: Some("http://example.com/v1/responses".to_string()),
headers: None,
body: Some(body),
}));
let CodexErrorDetails::MisalignmentPolicyViolation { message } = err.details() else {
panic!("expected CodexErrorDetails::MisalignmentPolicyViolation, got {err:?}");
};
assert_eq!(message, "This request violated the misalignment policy.");
assert!(!err.is_retryable());
}
#[test]
fn map_api_error_keeps_unknown_400_errors_generic() {
let body = serde_json::json!({
+2
View File
@@ -29,6 +29,8 @@ pub enum ApiError {
InvalidRequest { message: String },
#[error("cyber policy: {message}")]
CyberPolicy { message: String },
#[error("misalignment policy violation: {message}")]
MisalignmentPolicyViolation { message: String },
#[error("server overloaded")]
ServerOverloaded,
}
+54
View File
@@ -420,6 +420,15 @@ pub fn process_responses_event(
} else if is_cyber_policy_error(&error) {
let message = cyber_policy_message(error.message);
response_error = ApiError::CyberPolicy { message };
} else if error.code.as_deref() == Some("misalignment_policy_violation") {
let message = error
.message
.filter(|message| !message.trim().is_empty())
.unwrap_or_else(|| {
"This request was blocked due to a misalignment policy violation."
.to_string()
});
response_error = ApiError::MisalignmentPolicyViolation { message };
} else if matches!(error.code.as_deref(), Some("invalid_prompt" | "bio_policy"))
{
let message = error
@@ -1148,6 +1157,51 @@ mod tests {
}
}
#[tokio::test]
async fn misalignment_policy_violation_error_is_fatal() {
let raw_error = r#"{"type":"response.failed","sequence_number":3,"response":{"id":"resp_fatal_misalignment","object":"response","status":"failed","error":{"type":"invalid_request_error","code":"misalignment_policy_violation","message":"This request violated the misalignment policy."}}}"#;
let sse = format!("event: response.failed\ndata: {raw_error}\n\n");
let events = collect_events(&[sse.as_bytes()]).await;
assert_eq!(events.len(), 1);
match &events[0] {
Err(ApiError::MisalignmentPolicyViolation { message }) => {
assert_eq!(message, "This request violated the misalignment policy.");
}
other => panic!("unexpected event: {other:?}"),
}
}
#[tokio::test]
async fn misalignment_policy_violation_uses_fallback_for_blank_message() {
for message in ["", " "] {
let raw_error = serde_json::json!({
"type": "response.failed",
"response": {
"id": "resp_fatal_misalignment",
"status": "failed",
"error": {
"type": "invalid_request_error",
"code": "misalignment_policy_violation",
"message": message,
},
},
});
let sse = format!("event: response.failed\ndata: {raw_error}\n\n");
let events = collect_events(&[sse.as_bytes()]).await;
assert_eq!(events.len(), 1);
match &events[0] {
Err(ApiError::MisalignmentPolicyViolation { message }) => assert_eq!(
message,
"This request was blocked due to a misalignment policy violation."
),
other => panic!("unexpected event: {other:?}"),
}
}
}
#[tokio::test]
async fn content_policy_errors_without_type_are_invalid_requests() {
for (code, expected_message) in [
+7 -1
View File
@@ -132,6 +132,8 @@ pub enum CodexErrorDetails {
ServerOverloaded,
#[error("{message}")]
CyberPolicy { message: String },
#[error("{message}")]
MisalignmentPolicyViolation { message: String },
#[error("{0}")]
ResponseStreamFailed(ResponseStreamFailed),
#[error("{0}")]
@@ -383,7 +385,8 @@ impl CodexErr {
| CodexErrorDetails::SessionConfiguredNotFirstEvent
| CodexErrorDetails::UsageLimitReached(_)
| CodexErrorDetails::ServerOverloaded
| CodexErrorDetails::CyberPolicy { .. } => false,
| CodexErrorDetails::CyberPolicy { .. }
| CodexErrorDetails::MisalignmentPolicyViolation { .. } => false,
CodexErrorDetails::Stream(..)
| CodexErrorDetails::Timeout
| CodexErrorDetails::RequestTimeout
@@ -426,6 +429,9 @@ impl CodexErr {
| CodexErrorDetails::UsageNotIncluded => CodexErrorInfo::UsageLimitExceeded,
CodexErrorDetails::ServerOverloaded => CodexErrorInfo::ServerOverloaded,
CodexErrorDetails::CyberPolicy { .. } => CodexErrorInfo::CyberPolicy,
CodexErrorDetails::MisalignmentPolicyViolation { .. } => {
CodexErrorInfo::MisalignmentPolicyViolation
}
CodexErrorDetails::RetryLimit(_) => CodexErrorInfo::ResponseTooManyFailedAttempts {
http_status_code: self.http_status_code_value(),
},
+2
View File
@@ -1770,6 +1770,7 @@ pub enum CodexErrorInfo {
UsageLimitExceeded,
ServerOverloaded,
CyberPolicy,
MisalignmentPolicyViolation,
HttpConnectionFailed {
http_status_code: Option<u16>,
},
@@ -1808,6 +1809,7 @@ impl CodexErrorInfo {
| Self::UsageLimitExceeded
| Self::ServerOverloaded
| Self::CyberPolicy
| Self::MisalignmentPolicyViolation
| Self::HttpConnectionFailed { .. }
| Self::ResponseStreamConnectionFailed { .. }
| Self::InternalServerError
@@ -83,6 +83,7 @@ pub fn telemetry_api_error_message(error: &ApiError) -> String {
ApiError::RateLimit(_) => "rate limit".to_string(),
ApiError::InvalidRequest { .. } => "invalid request".to_string(),
ApiError::CyberPolicy { .. } => "cyber policy".to_string(),
ApiError::MisalignmentPolicyViolation { .. } => "misalignment policy violation".to_string(),
ApiError::ServerOverloaded => "server overloaded".to_string(),
}
}