Skip to content

Harden upstream error handling: biz errors, count_tokens, stream truncation, setup retries #94

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
AmanCode22 merged 4 commits into from
Nov 3, 2026
Merged

Harden upstream error handling: biz errors, count_tokens, stream truncation, setup retries #94

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
40f6565
Harden upstream error handling: biz errors, count_tokens, stream trun…
Rrainich Oct 31, 2026
4ce4c3a
docs: changelog + README for the upstream hardening
Rrainich Oct 31, 2026
bc2cbfe
Fix first-turn parent_message_id and diagnose non-SSE failures
Rrainich Oct 31, 2026
ff82027
Fix ruff import-order errors flagged by CI
Rrainich Nov 3, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 14 additions & 0 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,20 @@ All notable user-visible changes. One bullet per change, each citing its PR
or issue — no citation, no bullet. Dates are the day the change landed on
main.

## [Unreleased]

### Fixed

- First message of a chat no longer sends `parent_message_id` (upstream began rejecting `0` with biz_code 26 "invalid message id" on 2026-10-31, and edges disagree on whether `null` is accepted). Follow-up turns keep the `parent = previous response id` chain. (PR #92)
- An upstream failure delivered as `application/json` on the SSE endpoint (as that biz_code 26 was) now names its code and message in the raised error and logs the response content-type with the raw non-SSE body, instead of surfacing as an anonymous empty stream. (PR #92)
- Upstream business errors (HTTP 200 with `code=40003 INVALID_TOKEN`, "messages too frequent", a killed web session) surface as a typed 401/429 and rotate the token instead of raising `TypeError` inside the route and pinning the dead account forever. The pre-turn create hops (`create_new_chat`, `create_challange_pow`) go through the same classifier, outside the per-signature lock so the retry cannot deadlock. (PR #92)
- Upstream hang-ups mid-stream log a warning and emit the terminal error frame instead of a traceback and a dead socket. (PR #92)
- A single upstream turn is bounded by `DEEPSEEKER_REQUEST_DEADLINE` (default 600 s); a slowly-dripping upstream no longer holds a request until the client gives up. (PR #92)

### Added

- `POST /v1/messages/count_tokens` (also at `/messages/count_tokens`) — Claude Code's context probes returned 404 before. (PR #92)

## [6.0] — 2026-10-21

### Fixed
Expand Down
1 change: 1 addition & 0 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -210,6 +210,7 @@ Copy `.env.example` to `.env` and edit your secret values. The table below is th
- `GET /v1/files/{file_id}/content`
- **Anthropic Base**: `http://localhost:4000`
- `POST /v1/messages` (also at `/messages`)
- `POST /v1/messages/count_tokens` (also at `/messages/count_tokens`)
- `POST /v1/files/upload`
- **Auth Key**: configured in `.env` (`DEEPSEEKER_API_KEY`)

Expand Down
Loading