CLIProxyAPI

Merge pull request #1664 from ciberponk/pr/responses-compaction-compat

feat: add codex responses compatibility for compaction payloads

Luis Pater · 2026-02-25 01:21:59 +08:00

d24ea4ce2a

Merge pull request #1687 from lyd123qw2008/fix/codex-refresh-token-reused-no-retry

fix(codex): stop retrying refresh_token_reused errors

Luis Pater · 2026-02-25 01:19:30 +08:00

2c30c981ae

Merge pull request #1676 from piexian/feat/qwen-quota-handling-clean

feat(qwen): add rate limiting and quota error handling

Luis Pater · 2026-02-25 01:07:55 +08:00

f1e9a787d7

Merge pull request #1683 from dusty-du/codex/device-login-flow

Add additive Codex device-code login flow

Luis Pater · 2026-02-25 00:50:48 +08:00

c66cb0afd2

Revert "Merge pull request #1627 from thebtf/fix/reasoning-effort-clamping"

hkfires · 2026-02-24 19:47:53 +08:00

0659ffab75

Merge pull request #1654 from alexey-yanchenko/feature/pass-file-inputs

Pass file input from /chat/completions and /responses to codex and claude

Luis Pater · 2026-02-24 05:53:11 +08:00

c3e12c5e58

Merge pull request #1643 from alexey-yanchenko/fix/gemini-prompt-tokens

Fix usage convertation from gemini response to openai format

Luis Pater · 2026-02-24 05:46:13 +08:00

1825fc7503

Merge pull request #1527 from HEUDavid/feat/auth-hook

feat(auth): add post-auth hook mechanism

Luis Pater · 2026-02-24 05:33:13 +08:00

48732ba05e

test(codex): log non-retryable refresh error and cover single-attempt behavior

lyd123qw2008 · 2026-02-23 22:41:33 +08:00

3b3e0d1141

fix(codex): stop retrying refresh_token_reused errors

lyd123qw2008 · 2026-02-23 22:31:30 +08:00

7acd428507

Add additive Codex device-code login flow

test · 2026-02-23 06:30:04 -05:00

492b9c46f0

feat(qwen): add rate limiting and quota error handling

- Add 60 requests/minute rate limiting per credential using sliding window
- Detect insufficient_quota errors and set cooldown until next day (Beijing time)
- Map quota errors (HTTP 403/429) to 429 with retryAfter for conductor integration
- Cache Beijing timezone at package level to avoid repeated syscalls
- Add redactAuthID function to protect credentials in logs
- Extract wrapQwenError helper to consolidate error handling

piexian · 2026-02-23 00:38:46 +08:00

3b421c8181

Fixed: #1675

fix(gemini): add model definitions for Gemini 3.1 Pro High and Image

Luis Pater · 2026-02-23 00:12:57 +08:00

713388dd7b

Merge pull request #1522 from soilSpoon/feature/canceled

feature(proxy): Adds special handling for client cancellations in proxy error handler

Luis Pater · 2026-02-22 22:02:59 +08:00

e6c7af0fa9

fix(gemini): update min Thinking value and add Gemini 3.1 Pro Preview model definition

Luis Pater · 2026-02-22 21:51:32 +08:00

d210be06c2

refactor: simplify context_management compatibility handling

fan · 2026-02-21 22:20:48 +08:00

afc8a0f9be

Merge pull request #1662 from matchch/contribute/cache-user-id

feat: add cache-user-id toggle for Claude cloaking

Luis Pater · 2026-02-21 20:51:30 +08:00

d6ec33e8e1

fix(gemini): correct Created timestamps for Gemini 3.1 Pro Preview model definitions

Luis Pater · 2026-02-21 20:47:47 +08:00

081cfe806e

feat(gemini): add Gemini 3.1 Pro Preview model definitions

hkfires · 2026-02-21 20:42:29 +08:00

c1c62a6c04

feat: support responses compaction payload compatibility for codex translator

ciberponk · 2026-02-21 12:56:10 +08:00

d693d7993b

feat: add cache-user-id toggle for Claude cloaking

Default to generating a fresh random user_id per request instead of
reusing cached IDs. Add cache-user-id config option to opt in to the
previous caching behavior.

- Add CacheUserID field to CloakConfig
- Extract user_id cache logic to dedicated file
- Generate fresh user_id by default, cache only when enabled
- Add tests for both paths

matchch · 2026-02-21 12:31:20 +08:00

2fdf5d2793

Merge pull request #1660 from Grivn/fix/claude-token-url

fix(claude): use api.anthropic.com for OAuth token exchange

Luis Pater · 2026-02-20 21:52:08 +08:00

7b0eb41ebc

fix(claude): use api.anthropic.com for OAuth token exchange

console.anthropic.com is now protected by a Cloudflare managed challenge
that blocks all non-browser POST requests to /v1/oauth/token, causing
`-claude-login` to fail with a 403 error.

Switch to api.anthropic.com which hosts the same OAuth token endpoint
without the Cloudflare managed challenge.

Fixes #1659

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Grivn · 2026-02-20 20:11:27 +08:00

ef5901c81b

Merge pull request #1652 from thebtf/fix/claude-translator-arguments

fix(translator): handle tool call arguments in codex→claude streaming translator

Luis Pater · 2026-02-20 19:50:20 +08:00

d4829c82f7

Merge pull request #1644 from possible055/main

feat: add Gemini 3.1 Pro Preview model definition

Luis Pater · 2026-02-20 19:44:59 +08:00

a5f4166a9b

Pass file input from /chat/completions and /responses to codex and claude

Alexey Yanchenko · 2026-02-20 10:25:44 +07:00

0cbfe7f457

fix: prevent duplicate function call arguments when delta events precede done

Non-spark codex models (gpt-5.3-codex, gpt-5.2-codex) stream function call
arguments via multiple delta events followed by a done event. The done handler
unconditionally emitted the full arguments, duplicating what deltas already
streamed. This produced invalid double JSON that Claude Code couldn't parse,
causing tool calls to fail with missing parameters and infinite retry loops.

Add HasReceivedArgumentsDelta flag to track whether delta events were received.
The done handler now only emits arguments when no deltas preceded it (spark
models), while delta-based streaming continues to work for non-spark models.

Kirill Turanskiy · 2026-02-19 23:18:14 +03:00

1cc21cc45b

fix: handle response.function_call_arguments.done in codex→claude streaming translator

Some Codex models (e.g. gpt-5.3-codex-spark) send function call arguments
in a single "done" event without preceding "delta" events. The streaming
translator only handled "delta" events, causing tool call arguments to be
lost — resulting in empty tool inputs and infinite retry loops in clients
like Claude Code.

Emit the full arguments from the "done" event as a single input_json_delta
so downstream clients receive the complete tool input.

Kirill Turanskiy · 2026-02-19 23:18:14 +03:00

07cf616e2b

Merge branch 'codex/pr-1626' into dev

Luis Pater · 2026-02-19 21:33:23 +08:00

e92e2af71a

feat: add passthrough headers configuration

- Introduced `passthrough-headers` option in configuration to control forwarding of upstream response headers.
- Updated handlers to respect the passthrough headers setting.
- Added tests to verify behavior when passthrough is enabled or disabled.

Luis Pater · 2026-02-19 21:31:29 +08:00

a6bdd9a652

fix(antigravity): prevent invalid JSON when tool_result has no content

sjson.SetRaw with an empty string produces malformed JSON (e.g. "result":}).
This happens when a Claude tool_result block has no content field, causing
functionResponseResult.Raw to be "". Guard against this by falling back to
sjson.Set with an empty string only when .Raw is empty.

TinyCoder · 2026-02-19 17:08:39 +07:00

00822770ec

feat: add Gemini 3.1 Pro Preview model definition

apparition · 2026-02-19 17:43:08 +08:00

1a0ceda0fc

Fix usage convertation from gemini response to openai format

Alexey Yanchenko · 2026-02-19 15:34:59 +07:00

b9ae4ab803

fix: ensure connection-scoped headers are filtered in upstream requests

- Added `connectionScopedHeaders` utility to respect "Connection" header directives.
- Updated `FilterUpstreamHeaders` to remove connection-scoped headers dynamically.
- Refactored and tested upstream header filtering with additional validations.
- Adjusted upstream header handling during retries to replace headers safely.

Luis Pater · 2026-02-19 13:19:10 +08:00

2789396435

Merge PR #1626 into codex/pr-1626

Luis Pater · 2026-02-19 04:49:14 +08:00

61da7bd981

Merge pull request #1635 from thebtf/fix/openai-translator-tool-streaming

fix: handle tool call argument streaming in Codex→OpenAI translator

Luis Pater · 2026-02-19 04:22:12 +08:00

9c040445af

Merge pull request #1628 from thebtf/fix/masquerading-headers

fix: update Claude masquerading headers and configurable defaults

Luis Pater · 2026-02-19 04:19:59 +08:00

fff866424e

Merge pull request #1627 from thebtf/fix/reasoning-effort-clamping

fix: clamp reasoning_effort to valid OpenAI-format values

Luis Pater · 2026-02-19 04:15:19 +08:00

2d12becfd6

Merge pull request #1625 from thebtf/feat/tool-prefix-config

feat: add per-auth tool_prefix_disabled option

Luis Pater · 2026-02-19 04:07:22 +08:00

252f7e0751

Merge branch 'pr-1624' into dev

# Conflicts:
#	internal/runtime/executor/claude_executor.go
#	internal/runtime/executor/claude_executor_test.go

Luis Pater · 2026-02-19 04:05:04 +08:00

b2b17528cb

Merge pull request #1608 from thebtf/fix/tool-reference-proxy-prefix-mainline

fix: add proxy_ prefix handling for tool_reference content blocks

Luis Pater · 2026-02-19 03:50:34 +08:00

76294f0c59

Merge branch 'tui' into dev

Luis Pater · 2026-02-19 03:18:24 +08:00

7fe8246a9f

feat(tui): add standalone mode and API-based log polling

- Implemented `--standalone` mode to launch an embedded server for TUI.
- Enhanced TUI client to support API-based log polling when log hooks are unavailable.
- Added authentication gate for password input and connection handling.
- Improved localization and UX for logs, authentication, and status bar rendering.

Luis Pater · 2026-02-19 03:18:08 +08:00

93fe58e31e

chore(executor): remove unused Openai-Beta header from Codex executor

Luis Pater · 2026-02-19 02:19:48 +08:00

e5b5dc870f

feat(logging, executor): add request logging tests and WebSocket-based Codex executor

- Introduced unit tests for request logging middleware to enhance coverage.
- Added WebSocket-based Codex executor to support Responses API upgrade.
- Updated middleware logic to selectively capture request bodies for memory efficiency.
- Enhanced Codex configuration handling with new WebSocket attributes.

Luis Pater · 2026-02-19 01:57:02 +08:00

bb86a0c0c4

fix: handle tool call argument streaming in Codex→OpenAI translator

The OpenAI Chat Completions translator was silently dropping
response.function_call_arguments.delta and
response.function_call_arguments.done Codex SSE events, meaning
tool call arguments were never streamed incrementally to clients.

Add proper handling mirroring the proven Claude translator pattern:

- response.output_item.added: announce tool call (id, name, empty args)
- response.function_call_arguments.delta: stream argument chunks
- response.function_call_arguments.done: emit full args if no deltas
- response.output_item.done: defensive fallback for backward compat

State tracking via HasReceivedArgumentsDelta and HasToolCallAnnounced
ensures no duplicate argument emission and correct behavior for models
like codex-spark that skip delta events entirely.

Kirill Turanskiy · 2026-02-18 19:09:05 +03:00

5fa23c7f41

fix: update Claude masquerading headers and make them configurable

Update hardcoded X-Stainless-* and User-Agent defaults to match
Claude Code 2.1.44 / @anthropic-ai/sdk 0.74.0 (verified via
diagnostic proxy capture 2026-02-17).

Changes:
- X-Stainless-Os/Arch: dynamic via runtime.GOOS/GOARCH
- X-Stainless-Package-Version: 0.55.1 → 0.74.0
- X-Stainless-Timeout: 60 → 600
- User-Agent: claude-cli/1.0.83 (external, cli) → claude-cli/2.1.44 (external, sdk-cli)

Add claude-header-defaults config section so values can be updated
without recompilation when Claude Code releases new versions.

Kirill Turanskiy · 2026-02-18 03:38:51 +03:00

73dc0b10b8

fix: clamp reasoning_effort to valid OpenAI-format values

CPA-internal thinking levels like 'xhigh' and 'minimal' are not accepted
by OpenAI-format providers (MiniMax, etc.). The OpenAI applier now maps
non-standard levels to the nearest valid reasoning_effort value before
writing to the request body:

  xhigh   → high
  minimal → low
  auto    → medium

Kirill Turanskiy · 2026-02-18 03:36:42 +03:00

2ea95266e3

feat: passthrough upstream response headers to clients

CPA previously stripped ALL response headers from upstream AI provider
APIs, preventing clients from seeing rate-limit info, request IDs,
server-timing and other useful headers.

Changes:
- Add Headers field to Response and StreamResult structs
- Add FilterUpstreamHeaders helper (hop-by-hop + security denylist)
- Add WriteUpstreamHeaders helper (respects CPA-set headers)
- ExecuteWithAuthManager/ExecuteCountWithAuthManager now return headers
- ExecuteStreamWithAuthManager returns headers from initial connection
- All 11 provider executors populate Response.Headers
- All handler call sites write filtered upstream headers before response

Filtered headers (not forwarded):
- RFC 7230 hop-by-hop: Connection, Transfer-Encoding, Keep-Alive, etc.
- Security: Set-Cookie
- CPA-managed: Content-Length, Content-Encoding

Kirill Turanskiy · 2026-02-18 00:16:22 +03:00

1f8f198c45

feat: add per-auth tool_prefix_disabled option

Allow disabling the proxy_ tool name prefix on a per-account basis.
Users who route their own Anthropic account through CPA can set
"tool_prefix_disabled": true in their OAuth auth JSON to send tool
names unchanged to Anthropic.

Default behavior is fully preserved — prefix is applied unless
explicitly disabled.

Changes:
- Add ToolPrefixDisabled() accessor to Auth (reads metadata key
  "tool_prefix_disabled" or "tool-prefix-disabled")
- Gate all 6 prefix apply/strip points with the new flag
- Add unit tests for the accessor

Kirill Turanskiy · 2026-02-17 21:48:19 +03:00

9261b0c20b

1301 Commits