Greptile Summary

This PR adds a transient cooldown path to the sendChatAction handler, complementing the existing 401 suspension mechanism. When a network error, 429, or 5xx is caught, the handler silently resolves (typing indicators are best-effort), sets an exponentially-backed-off transientCooldownUntil timestamp, and suppresses subsequent calls until the cooldown expires. The existing 401 path is unchanged — it still accumulates, backs off with sleepWithAbort, and suspends after maxConsecutive401 failures. A 401 during an active transient cooldown correctly resets the transient state before incrementing the 401 counter.\n\nKey observations from the review:\n- State machine correctness: the ordering of guards (suspended → transientCooldownUntil → 401 backoff sleep → try/catch) is logically sound with no unreachable or mis-ordered branches.\n- First-failure silent swallow is intentional: callers of sendChatAction receive Promise<void> and the PR explicitly scopes typing indicators as best-effort, so this is appropriate — but reviewers should be aware this is a behavior change from the pre-PR "always throw on non-401" contract.\n- Mock refactor: switching from importOriginal-spread to a full factory correctly includes the four infra-runtime helpers transitively required by network-errors.ts (collectErrorGraphCandidates, extractErrorCode, formatErrorMessage, readErrorName).\n- Minor style note: isRateLimitError uses string-matching while the rest of the error-detection layer uses typed error_code fields via hasTelegramErrorCode. Functionally equivalent for all realistic Telegram shapes but inconsistent with the existing pattern.

Confidence Score: 5/5

Safe to merge — the transient cooldown path is correctly implemented, all key scenarios are tested, and the 401 suspension contract is preserved.

No P0 or P1 issues found. The only comment is a P2 style suggestion about using structured error_code detection for 429 instead of string-matching, which is functionally equivalent for all realistic Telegram error shapes.

No files require special attention.

This is a comment left during a code review.
Path: extensions/telegram/src/sendchataction-401-backoff.ts
Line: 74-81

Comment:
**`isRateLimitError` duplicates structured detection already available**

`isRateLimitError` relies on string-matching (`JSON.stringify`) while `isTelegramServerError` (used right below it) uses the typed `error_code` field via `hasTelegramErrorCode`. A 429 response from Telegram carries `error_code: 429`, so it would be detected consistently if `isTransientSendChatActionError` used `hasTelegramErrorCode(error, (code) => code === 429)` (or a new exported `isTelegramRateLimitError`) instead of string-matching. The current approach works for every realistic Telegram error shape (since `JSON.stringify` will always include the number `429`), but introduces a style inconsistency that could drift if error shapes change.

```suggestion
function isRateLimitError(error: unknown): boolean {
  return hasTelegramErrorCode(error, (code) => code === 429);
}
```

> Note: `hasTelegramErrorCode` is a file-private helper in `network-errors.ts`. If you prefer to keep the string-fallback for non-structured errors (e.g. raw `Error("429 Too Many Requests")`), consider exporting a dedicated helper from `network-errors.ts` that handles both paths.

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (1): Last reviewed commit: "fix(telegram): cool down transient sendC..." | Re-trigger Greptile}

Updated this PR again.

Latest follow-ups I addressed:

• narrowed the test-only runtime overrides so the sendChatAction tests no longer replace the whole runtime module surface
• switched transient 429 handling to structured error_code detection via isTelegramRateLimitError
• changed transient sendChatAction failures to keep rejecting during cooldown, so the existing typing start guard can still count failures and trip during rate limits / outages instead of treating those starts as successful

Fresh local verification:

• pnpm test -- extensions/telegram/src/sendchataction-401-backoff.test.ts
• pnpm test -- extensions/telegram/src/network-errors.test.ts
• pnpm test -- extensions/telegram/src/send.test.ts -t "sends typing"
• pnpm lint:plugins:no-monolithic-plugin-sdk-entry-imports
• pnpm test -- src/plugins/contracts/web-search-provider.contract.test.ts

I also cross-checked the earlier failing PR run against a failing main run (23655258874). A large part of the red CI surface already overlaps with main, including build-smoke, checks-fast-contracts-protocol, checks-fast-extensions, extension-fast-telegram, checks-node-channels-*, checks-node-test-*, and the Windows shards.

I’m keeping this PR scoped to the Telegram sendChatAction fix and the review follow-ups, and I’ll keep watching the new CI run.

ProjectClownfish pushed a narrow repair to this branch so the original contributor path can stay canonical.

Source PR: #55886
Validation: pnpm test -- extensions/telegram/src/sendchataction-401-backoff.test.ts; pnpm test -- extensions/telegram/src/network-errors.test.ts; pnpm test -- extensions/telegram/src/send.test.ts -t "sends typing"; pnpm lint:plugins:no-monolithic-plugin-sdk-entry-imports; pnpm check:changed
Contributor credit is preserved in the branch history and PR context.

Thanks for the context here. I swept through the related work, and this is now duplicate or superseded.

Keep open: current main still lacks the central transient Telegram sendChatAction cooldown, and this branch contains useful bug-fix work. It is not merge-ready because it conflicts with current main, narrows current rate-limit fallback behavior, edits release-owned CHANGELOG.md, and lacks real Telegram behavior proof.

Canonical path: Close this stale PR. The latest review rated it F, the branch still lacks merge-ready proof, and there has been no human follow-up after the durable review.

So I’m closing this here because the remaining work is already tracked in the canonical issue.

ClawSweeper PR egg

🎁 Pass real behavior proof to wake the egg and unlock a hatchable treat.

This pull request has been automatically marked as stale due to inactivity.
Please add updates or it will be closed.

Clownfish 🐠 reef update

Thanks for the work on this. Clownfish could not push to this branch with the permissions available, so it opened a narrow replacement PR to keep the fix swimming forward without losing the contributor trail. not your fault, just GitHub branch-permission tides.

Replacement PR: #93020
Source PR: #55886
This source PR remains open; the replacement is just the writable fix lane.
The original contribution stays credited in the replacement PR context.

fish notes: model gpt-5.5, reasoning xhigh; reviewed against 5b4b09f.

Clownfish 🐠 landed

Thanks for the report and context here. This is superseded by #93020, which has landed as the canonical Clownfish fix path for this cluster.

Closing this now that the validated fix has landed. If current main still shows a different path, reply here and we can reopen.

fish notes: model gpt-5.5, reasoning medium; reviewed against 6134c00.