fix: restart stale codex threads
This commit is contained in:
@@ -330,13 +330,36 @@ export async function runCodexStdioTurn(options: CodexStdioTurnOptions): Promise
|
||||
client.notify("initialized", {});
|
||||
events.push({ type: "backend_status", payload: { phase: "initialize:completed", ...backendMetadata(options), protocol: codexProtocol } });
|
||||
|
||||
const threadMethod = options.threadId ? "thread/resume" : "thread/start";
|
||||
const threadParams: JsonRecord = options.threadId
|
||||
? withOptionalModel({ threadId: options.threadId, cwd: options.cwd, approvalPolicy: options.approvalPolicy, sandbox: options.sandbox }, options.model)
|
||||
: withOptionalModel({ cwd: options.cwd, approvalPolicy: options.approvalPolicy, sandbox: options.sandbox, serviceName: "agentrun" }, options.model);
|
||||
const threadResponse = requireResponseRecord(await client.request(threadMethod, threadParams, requestTimeoutMs), threadMethod);
|
||||
threadId = requireNestedId(threadResponse, threadMethod, "thread");
|
||||
events.push({ type: "backend_status", payload: { phase: `${threadMethod}:completed`, threadId } });
|
||||
const startThread = async (phasePrefix = "thread/start"): Promise<string> => {
|
||||
const response = requireResponseRecord(await client!.request("thread/start", withOptionalModel({ cwd: options.cwd, approvalPolicy: options.approvalPolicy, sandbox: options.sandbox, serviceName: "agentrun" }, options.model), requestTimeoutMs), "thread/start");
|
||||
const nextThreadId = requireNestedId(response, "thread/start", "thread");
|
||||
events.push({ type: "backend_status", payload: { phase: `${phasePrefix}:completed`, threadId: nextThreadId } });
|
||||
return nextThreadId;
|
||||
};
|
||||
|
||||
if (options.threadId) {
|
||||
try {
|
||||
const threadResponse = requireResponseRecord(await client.request("thread/resume", withOptionalModel({ threadId: options.threadId, cwd: options.cwd, approvalPolicy: options.approvalPolicy, sandbox: options.sandbox }, options.model), requestTimeoutMs), "thread/resume");
|
||||
threadId = requireNestedId(threadResponse, "thread/resume", "thread");
|
||||
events.push({ type: "backend_status", payload: { phase: "thread/resume:completed", threadId } });
|
||||
} catch (error) {
|
||||
const failure = normalizeFailure(error);
|
||||
if (!isStaleThreadResumeFailure(failure)) throw error;
|
||||
events.push({
|
||||
type: "backend_status",
|
||||
payload: {
|
||||
phase: "thread/resume:stale-thread-fallback",
|
||||
requestedThreadId: options.threadId,
|
||||
failureKind: failure.failureKind,
|
||||
message: failure.message,
|
||||
fallback: "thread/start",
|
||||
},
|
||||
});
|
||||
threadId = await startThread("thread/start:after-stale-resume");
|
||||
}
|
||||
} else {
|
||||
threadId = await startThread();
|
||||
}
|
||||
|
||||
const turnResponse = requireResponseRecord(await client.request("turn/start", withOptionalModel({ threadId, input: textInput(options.prompt), cwd: options.cwd, approvalPolicy: options.approvalPolicy }, options.model), requestTimeoutMs), "turn/start");
|
||||
turnId = requireNestedId(turnResponse, "turn/start", "turn");
|
||||
@@ -644,6 +667,12 @@ function normalizeFailure(error: unknown): CodexStdioFailure {
|
||||
return new CodexStdioFailure(classifyMessageFailureKind(message, "backend-protocol-error"), message, "codex-stdio");
|
||||
}
|
||||
|
||||
function isStaleThreadResumeFailure(error: CodexStdioFailure): boolean {
|
||||
if (error.phase !== "response:thread/resume") return false;
|
||||
const text = `${error.message}\n${JSON.stringify(error.details)}`.toLowerCase();
|
||||
return /no rollout found for thread id/u.test(text);
|
||||
}
|
||||
|
||||
function classifyCodexErrorRecord(error: JsonRecord, fallback: FailureKind): FailureKind {
|
||||
const parts: string[] = [];
|
||||
if (typeof error.message === "string") parts.push(error.message);
|
||||
|
||||
Reference in New Issue
Block a user