fix: partial log stream + better error messages for fast-exit containers (FAR-122 follow-up) #6

Merged
farhoodliquor-paperclip[bot] merged 2 commits from fix/far-122-partial-log-stream-and-error-message into master 2026-04-22 19:51:05 +00:00
2 changed files with 170 additions and 13 deletions
+108
View File
@@ -0,0 +1,108 @@
import { describe, it, expect } from "vitest";
import { isK8s404, buildPartialRunError } from "./execute.js";
describe("isK8s404", () => {
it("returns false for non-Error values", () => {
expect(isK8s404(null)).toBe(false);
expect(isK8s404(undefined)).toBe(false);
expect(isK8s404("string error")).toBe(false);
expect(isK8s404(404)).toBe(false);
});
it("returns false for unrelated errors", () => {
expect(isK8s404(new Error("something went wrong"))).toBe(false);
expect(isK8s404(new Error("HTTP-Code: 500 Message: Internal Server Error"))).toBe(false);
});
it("detects 404 from v1.0+ message format", () => {
const err = new Error("HTTP-Code: 404 Message: Unknown API Status Code! Body: ...");
expect(isK8s404(err)).toBe(true);
});
it("detects 404 from v0.x response.statusCode", () => {
const err = Object.assign(new Error("Not Found"), {
response: { statusCode: 404 },
});
expect(isK8s404(err)).toBe(true);
});
it("detects 404 from v1.0+ response.status", () => {
const err = Object.assign(new Error("Not Found"), {
response: { status: 404 },
});
expect(isK8s404(err)).toBe(true);
});
it("detects 404 from direct statusCode property", () => {
const err = Object.assign(new Error("Not Found"), { statusCode: 404 });
expect(isK8s404(err)).toBe(true);
});
it("does not match non-404 status codes on response", () => {
const err = Object.assign(new Error("Forbidden"), {
response: { statusCode: 403 },
});
expect(isK8s404(err)).toBe(false);
});
});
describe("buildPartialRunError", () => {
const initLine = JSON.stringify({
type: "system",
subtype: "init",
model: "claude-sonnet-4-6",
session_id: "sess_abc",
});
it("returns parse-failure message when exitCode is 0", () => {
expect(buildPartialRunError(0, "", "")).toBe("Failed to parse Claude JSON output");
expect(buildPartialRunError(0, "claude-sonnet-4-6", initLine)).toBe(
"Failed to parse Claude JSON output",
);
});
it("returns generic exit message when stdout is empty", () => {
expect(buildPartialRunError(1, "", "")).toBe("Claude exited with code 1");
expect(buildPartialRunError(null, "", "")).toBe("Claude exited with code -1");
});
it("skips system/init events and returns generic message when only init captured", () => {
const msg = buildPartialRunError(1, "claude-sonnet-4-6", initLine);
expect(msg).toBe(
"Claude started but did not produce a result (model: claude-sonnet-4-6) — check API credentials, model support, and adapter config",
);
});
it("includes model from parsedStream when stdout is init-only", () => {
const msg = buildPartialRunError(null, "MiniMax-M2.7", initLine);
expect(msg).toContain("MiniMax-M2.7");
expect(msg).not.toContain("type");
expect(msg).not.toContain("system");
});
it("uses first non-system line as content when present", () => {
const stdout = [initLine, "Error: no API key configured"].join("\n");
const msg = buildPartialRunError(1, "claude-sonnet-4-6", stdout);
expect(msg).toBe("Claude exited with code 1: Error: no API key configured");
});
it("uses first non-system JSON event as content", () => {
const resultLike = JSON.stringify({ type: "result", subtype: "error", result: "rate limit" });
const stdout = [initLine, resultLike].join("\n");
const msg = buildPartialRunError(2, "claude-sonnet-4-6", stdout);
expect(msg).toContain("rate limit");
expect(msg).toContain("code 2");
});
it("null exitCode renders as -1 in message", () => {
const msg = buildPartialRunError(null, "", "Some plain error text");
expect(msg).toBe("Claude exited with code -1: Some plain error text");
});
it("skips multiple consecutive system events", () => {
const anotherSystem = JSON.stringify({ type: "system", subtype: "other" });
const stdout = [initLine, anotherSystem, "real error line"].join("\n");
const msg = buildPartialRunError(1, "model-x", stdout);
expect(msg).toBe("Claude exited with code 1: real error line");
});
});
+62 -13
View File
@@ -19,8 +19,9 @@ const MAX_LOG_RECONNECT_ATTEMPTS = 50;
/**
* Detect a Kubernetes 404 (Not Found) error from @kubernetes/client-node.
* Works for both v0.x (response.statusCode) and v1.0+ (response.status, message).
* Exported for unit tests.
*/
function isK8s404(err: unknown): boolean {
export function isK8s404(err: unknown): boolean {
if (!(err instanceof Error)) return false;
const e = err as unknown as Record<string, unknown>;
const resp = e.response as Record<string, unknown> | undefined;
@@ -29,6 +30,45 @@ function isK8s404(err: unknown): boolean {
return /HTTP-Code:\s*404\b/.test(err.message);
}
/**
* Build the error message when Claude's stdout contains no result event.
* Skips system/init event lines so the UI doesn't display the raw init JSON.
* Exported for unit tests.
*/
export function buildPartialRunError(
exitCode: number | null,
model: string,
stdout: string,
): string {
if (exitCode === 0) return "Failed to parse Claude JSON output";
// Walk stdout lines, skip system events, return the first real content line.
const firstContentLine = stdout.split(/\r?\n/)
.map((l) => l.trim())
.find((l) => {
if (!l) return false;
try {
const obj = JSON.parse(l);
if (typeof obj === "object" && obj !== null && (obj as Record<string, unknown>).type === "system") return false;
} catch {
// not JSON — treat as content
}
return true;
}) ?? "";
// If we only have system/init events and nothing else, surface the model
// name so the operator can diagnose missing credentials or unsupported model.
const initOnlyOutput = stdout.trim() !== "" && model !== "" && !firstContentLine;
if (initOnlyOutput) {
const modelHint = model ? ` (model: ${model})` : "";
return `Claude started but did not produce a result${modelHint} — check API credentials, model support, and adapter config`;
}
return firstContentLine
? `Claude exited with code ${exitCode ?? -1}: ${firstContentLine}`
: `Claude exited with code ${exitCode ?? -1}`;
}
/**
* Wait for the Job's pod to reach a terminal or running state.
* Returns the pod name once logs can be streamed, or throws on failure.
@@ -640,13 +680,27 @@ export async function execute(ctx: AdapterExecutionContext): Promise<AdapterExec
stdout = logResult.value;
}
// If the follow stream missed output (container exited quickly), do a
// one-shot log read as fallback before the pod is cleaned up.
if (!stdout.trim()) {
await onLog("stdout", `[paperclip] Log stream returned empty — reading pod logs directly...\n`);
stdout = await readPodLogs(namespace, podName, kubeconfigPath);
if (stdout.trim()) {
// One-shot log fallback: handles two failure modes with a single read.
// Mode 1 — empty stream: the follow stream returned nothing (fast exit before connection).
// Mode 2 — partial stream: we have some output but no result event (follow stream raced
// with container exit and captured only the init line before the connection dropped).
// A one-shot readPodLogs is more reliable for already-terminated containers and reads
// from the beginning of the log, giving us the full output.
// We use a cheap string scan for the result-event guard (avoids a full JSON parse here;
// the authoritative parse happens once below after all fallbacks complete).
const hasResultEvent = stdout.includes('"type":"result"');
const needsOneShot = !stdout.trim() || (stdout.trim() && !hasResultEvent);
if (needsOneShot) {
if (!stdout.trim()) {
await onLog("stdout", `[paperclip] Log stream returned empty — reading pod logs directly...\n`);
}
const oneShotLogs = await readPodLogs(namespace, podName, kubeconfigPath);
if (!stdout.trim() && oneShotLogs.trim()) {
stdout = oneShotLogs;
await onLog("stdout", stdout);
} else if (oneShotLogs && oneShotLogs.length > stdout.length) {
await onLog("stdout", `[paperclip] Log stream captured partial output — supplemental one-shot read returned more content.\n`);
stdout = oneShotLogs;
}
}
@@ -739,16 +793,11 @@ export async function execute(ctx: AdapterExecutionContext): Promise<AdapterExec
}
if (!parsed) {
const stderrLine = stdout.split(/\r?\n/).map((l) => l.trim()).find(Boolean) ?? "";
return {
exitCode,
signal: null,
timedOut: false,
errorMessage: exitCode === 0
? "Failed to parse Claude JSON output"
: stderrLine
? `Claude exited with code ${exitCode ?? -1}: ${stderrLine}`
: `Claude exited with code ${exitCode ?? -1}`,
errorMessage: buildPartialRunError(exitCode, parsedStream.model, stdout),
resultJson: { stdout },
};
}