mirror of
https://github.com/alkimake/paperclip.git
synced 2026-06-14 01:50:39 +09:00
## Thinking Path > - Paperclip orchestrates AI agents for zero-human companies > - Adapter execution and retry classification decide whether agent work pauses, retries, or recovers automatically > - Transient provider failures need to be classified precisely so Paperclip does not convert retryable upstream conditions into false hard failures > - At the same time, operators need an up-to-date model list for Codex-backed agents and prompts should nudge agents toward targeted verification instead of repo-wide sweeps > - This pull request tightens transient recovery classification for Claude and Codex, updates the agent prompt guidance, and adds Codex model refresh support end-to-end > - The benefit is better automatic retry behavior plus fresher operator-facing model configuration ## What Changed - added Codex usage-limit retry-window parsing and Claude extra-usage transient classification - normalized the heartbeat transient-recovery contract across adapter executions and heartbeat scheduling - documented that deferred comment wakes only reopen completed issues for human/comment-reopen interactions, while system follow-ups leave closed work closed - updated adapter-utils prompt guidance to prefer targeted verification - added Codex model refresh support in the server route, registry, shared types, and agent config form - added adapter/server tests covering the new parsing, retry scheduling, and model-refresh behavior ## Verification - `pnpm exec vitest run --project @paperclipai/adapter-utils packages/adapter-utils/src/server-utils.test.ts` - `pnpm exec vitest run --project @paperclipai/adapter-claude-local packages/adapters/claude-local/src/server/parse.test.ts` - `pnpm exec vitest run --project @paperclipai/adapter-codex-local packages/adapters/codex-local/src/server/parse.test.ts` - `pnpm exec vitest run --project @paperclipai/server server/src/__tests__/adapter-model-refresh-routes.test.ts server/src/__tests__/adapter-models.test.ts server/src/__tests__/claude-local-execute.test.ts server/src/__tests__/codex-local-execute.test.ts server/src/__tests__/heartbeat-process-recovery.test.ts server/src/__tests__/heartbeat-retry-scheduling.test.ts` ## Risks - Moderate behavior risk: retry classification affects whether runs auto-recover or block, so mistakes here could either suppress needed retries or over-retry real failures - Low workflow risk: deferred comment wake reopening is intentionally scoped to human/comment-reopen interactions so system follow-ups do not revive completed issues unexpectedly > For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and discuss it in `#dev` before opening the PR. Feature PRs that overlap with planned core work may need to be redirected — check the roadmap first. See `CONTRIBUTING.md`. ## Model Used - OpenAI Codex GPT-5-based coding agent with tool use and code execution in the Codex CLI environment ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [ ] If this change affects the UI, I have included before/after screenshots - [ ] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Paperclip <noreply@paperclip.ing>
261 lines
7.9 KiB
TypeScript
261 lines
7.9 KiB
TypeScript
import {
|
||
asString,
|
||
asNumber,
|
||
parseObject,
|
||
parseJson,
|
||
} from "@paperclipai/adapter-utils/server-utils";
|
||
|
||
const CODEX_TRANSIENT_UPSTREAM_RE =
|
||
/(?:we(?:'|’)re\s+currently\s+experiencing\s+high\s+demand|temporary\s+errors|rate[-\s]?limit(?:ed)?|too\s+many\s+requests|\b429\b|server\s+overloaded|service\s+unavailable|try\s+again\s+later)/i;
|
||
const CODEX_REMOTE_COMPACTION_RE = /remote\s+compact\s+task/i;
|
||
const CODEX_USAGE_LIMIT_RE =
|
||
/you(?:'|’)ve hit your usage limit for .+\.\s+switch to another model now,\s+or try again at\s+([^.!\n]+)(?:[.!]|\n|$)/i;
|
||
|
||
export function parseCodexJsonl(stdout: string) {
|
||
let sessionId: string | null = null;
|
||
let finalMessage: string | null = null;
|
||
let errorMessage: string | null = null;
|
||
const usage = {
|
||
inputTokens: 0,
|
||
cachedInputTokens: 0,
|
||
outputTokens: 0,
|
||
};
|
||
|
||
for (const rawLine of stdout.split(/\r?\n/)) {
|
||
const line = rawLine.trim();
|
||
if (!line) continue;
|
||
|
||
const event = parseJson(line);
|
||
if (!event) continue;
|
||
|
||
const type = asString(event.type, "");
|
||
if (type === "thread.started") {
|
||
sessionId = asString(event.thread_id, sessionId ?? "") || sessionId;
|
||
continue;
|
||
}
|
||
|
||
if (type === "error") {
|
||
const msg = asString(event.message, "").trim();
|
||
if (msg) errorMessage = msg;
|
||
continue;
|
||
}
|
||
|
||
if (type === "item.completed") {
|
||
const item = parseObject(event.item);
|
||
if (asString(item.type, "") === "agent_message") {
|
||
const text = asString(item.text, "");
|
||
if (text) finalMessage = text;
|
||
}
|
||
continue;
|
||
}
|
||
|
||
if (type === "turn.completed") {
|
||
const usageObj = parseObject(event.usage);
|
||
usage.inputTokens = asNumber(usageObj.input_tokens, usage.inputTokens);
|
||
usage.cachedInputTokens = asNumber(usageObj.cached_input_tokens, usage.cachedInputTokens);
|
||
usage.outputTokens = asNumber(usageObj.output_tokens, usage.outputTokens);
|
||
continue;
|
||
}
|
||
|
||
if (type === "turn.failed") {
|
||
const err = parseObject(event.error);
|
||
const msg = asString(err.message, "").trim();
|
||
if (msg) errorMessage = msg;
|
||
}
|
||
}
|
||
|
||
return {
|
||
sessionId,
|
||
summary: finalMessage?.trim() ?? "",
|
||
usage,
|
||
errorMessage,
|
||
};
|
||
}
|
||
|
||
export function isCodexUnknownSessionError(stdout: string, stderr: string): boolean {
|
||
const haystack = `${stdout}\n${stderr}`
|
||
.split(/\r?\n/)
|
||
.map((line) => line.trim())
|
||
.filter(Boolean)
|
||
.join("\n");
|
||
return /unknown (session|thread)|session .* not found|thread .* not found|conversation .* not found|missing rollout path for thread|state db missing rollout path|no rollout found for thread id/i.test(
|
||
haystack,
|
||
);
|
||
}
|
||
|
||
function buildCodexErrorHaystack(input: {
|
||
stdout?: string | null;
|
||
stderr?: string | null;
|
||
errorMessage?: string | null;
|
||
}): string {
|
||
return [
|
||
input.errorMessage ?? "",
|
||
input.stdout ?? "",
|
||
input.stderr ?? "",
|
||
]
|
||
.join("\n")
|
||
.split(/\r?\n/)
|
||
.map((line) => line.trim())
|
||
.filter(Boolean)
|
||
.join("\n");
|
||
}
|
||
|
||
function readTimeZoneParts(date: Date, timeZone: string) {
|
||
const values = new Map(
|
||
new Intl.DateTimeFormat("en-US", {
|
||
timeZone,
|
||
hourCycle: "h23",
|
||
year: "numeric",
|
||
month: "2-digit",
|
||
day: "2-digit",
|
||
hour: "2-digit",
|
||
minute: "2-digit",
|
||
}).formatToParts(date).map((part) => [part.type, part.value]),
|
||
);
|
||
return {
|
||
year: Number.parseInt(values.get("year") ?? "", 10),
|
||
month: Number.parseInt(values.get("month") ?? "", 10),
|
||
day: Number.parseInt(values.get("day") ?? "", 10),
|
||
hour: Number.parseInt(values.get("hour") ?? "", 10),
|
||
minute: Number.parseInt(values.get("minute") ?? "", 10),
|
||
};
|
||
}
|
||
|
||
function normalizeResetTimeZone(timeZoneHint: string | null | undefined): string | null {
|
||
const normalized = timeZoneHint?.trim();
|
||
if (!normalized) return null;
|
||
if (/^(?:utc|gmt)$/i.test(normalized)) return "UTC";
|
||
|
||
try {
|
||
new Intl.DateTimeFormat("en-US", { timeZone: normalized }).format(new Date(0));
|
||
return normalized;
|
||
} catch {
|
||
return null;
|
||
}
|
||
}
|
||
|
||
function dateFromTimeZoneWallClock(input: {
|
||
year: number;
|
||
month: number;
|
||
day: number;
|
||
hour: number;
|
||
minute: number;
|
||
timeZone: string;
|
||
}): Date | null {
|
||
let candidate = new Date(Date.UTC(input.year, input.month - 1, input.day, input.hour, input.minute, 0, 0));
|
||
const targetUtc = Date.UTC(input.year, input.month - 1, input.day, input.hour, input.minute, 0, 0);
|
||
|
||
for (let attempt = 0; attempt < 4; attempt += 1) {
|
||
const actual = readTimeZoneParts(candidate, input.timeZone);
|
||
const actualUtc = Date.UTC(actual.year, actual.month - 1, actual.day, actual.hour, actual.minute, 0, 0);
|
||
const offsetMs = targetUtc - actualUtc;
|
||
if (offsetMs === 0) break;
|
||
candidate = new Date(candidate.getTime() + offsetMs);
|
||
}
|
||
|
||
const verified = readTimeZoneParts(candidate, input.timeZone);
|
||
if (
|
||
verified.year !== input.year ||
|
||
verified.month !== input.month ||
|
||
verified.day !== input.day ||
|
||
verified.hour !== input.hour ||
|
||
verified.minute !== input.minute
|
||
) {
|
||
return null;
|
||
}
|
||
|
||
return candidate;
|
||
}
|
||
|
||
function nextClockTimeInTimeZone(input: {
|
||
now: Date;
|
||
hour: number;
|
||
minute: number;
|
||
timeZoneHint: string;
|
||
}): Date | null {
|
||
const timeZone = normalizeResetTimeZone(input.timeZoneHint);
|
||
if (!timeZone) return null;
|
||
|
||
const nowParts = readTimeZoneParts(input.now, timeZone);
|
||
let retryAt = dateFromTimeZoneWallClock({
|
||
year: nowParts.year,
|
||
month: nowParts.month,
|
||
day: nowParts.day,
|
||
hour: input.hour,
|
||
minute: input.minute,
|
||
timeZone,
|
||
});
|
||
if (!retryAt) return null;
|
||
|
||
if (retryAt.getTime() <= input.now.getTime()) {
|
||
const nextDay = new Date(Date.UTC(nowParts.year, nowParts.month - 1, nowParts.day + 1, 0, 0, 0, 0));
|
||
retryAt = dateFromTimeZoneWallClock({
|
||
year: nextDay.getUTCFullYear(),
|
||
month: nextDay.getUTCMonth() + 1,
|
||
day: nextDay.getUTCDate(),
|
||
hour: input.hour,
|
||
minute: input.minute,
|
||
timeZone,
|
||
});
|
||
}
|
||
|
||
return retryAt;
|
||
}
|
||
|
||
function parseLocalClockTime(clockText: string, now: Date): Date | null {
|
||
const normalized = clockText.trim();
|
||
const match = normalized.match(/^(\d{1,2})(?::(\d{2}))?\s*([ap])\.?\s*m\.?(?:\s*\(([^)]+)\)|\s+([A-Z]{2,5}))?$/i);
|
||
if (!match) return null;
|
||
|
||
const hour12 = Number.parseInt(match[1] ?? "", 10);
|
||
const minute = Number.parseInt(match[2] ?? "0", 10);
|
||
if (!Number.isInteger(hour12) || hour12 < 1 || hour12 > 12) return null;
|
||
if (!Number.isInteger(minute) || minute < 0 || minute > 59) return null;
|
||
|
||
let hour24 = hour12 % 12;
|
||
if ((match[3] ?? "").toLowerCase() === "p") hour24 += 12;
|
||
|
||
const timeZoneHint = match[4] ?? match[5];
|
||
if (timeZoneHint) {
|
||
const explicitRetryAt = nextClockTimeInTimeZone({
|
||
now,
|
||
hour: hour24,
|
||
minute,
|
||
timeZoneHint,
|
||
});
|
||
if (explicitRetryAt) return explicitRetryAt;
|
||
}
|
||
|
||
const retryAt = new Date(now);
|
||
retryAt.setHours(hour24, minute, 0, 0);
|
||
if (retryAt.getTime() <= now.getTime()) {
|
||
retryAt.setDate(retryAt.getDate() + 1);
|
||
}
|
||
return retryAt;
|
||
}
|
||
|
||
export function extractCodexRetryNotBefore(input: {
|
||
stdout?: string | null;
|
||
stderr?: string | null;
|
||
errorMessage?: string | null;
|
||
}, now = new Date()): Date | null {
|
||
const haystack = buildCodexErrorHaystack(input);
|
||
const usageLimitMatch = haystack.match(CODEX_USAGE_LIMIT_RE);
|
||
if (!usageLimitMatch) return null;
|
||
return parseLocalClockTime(usageLimitMatch[1] ?? "", now);
|
||
}
|
||
|
||
export function isCodexTransientUpstreamError(input: {
|
||
stdout?: string | null;
|
||
stderr?: string | null;
|
||
errorMessage?: string | null;
|
||
}): boolean {
|
||
const haystack = buildCodexErrorHaystack(input);
|
||
|
||
if (extractCodexRetryNotBefore(input) != null) return true;
|
||
if (!CODEX_TRANSIENT_UPSTREAM_RE.test(haystack)) return false;
|
||
// Keep automatic retries scoped to the observed remote-compaction/high-demand
|
||
// failure shape, plus explicit usage-limit windows that tell us when retrying
|
||
// becomes safe again.
|
||
return CODEX_REMOTE_COMPACTION_RE.test(haystack) || /high\s+demand|temporary\s+errors/i.test(haystack);
|
||
}
|