LLM/CLIProxyAPIPlus
1
0
Fork 0
mirror of https://github.com/router-for-me/CLIProxyAPIPlus.git synced 2026-04-23 19:52:40 +00:00
Code Issues Packages Projects Releases Wiki Activity

Compare commits

base: LLM:v6.5.64-0
Branches Tags
LLM:main LLM:pr-59-resolve-conflicts
LLM:v6.9.28-0 LLM:v6.9.27-0 LLM:v6.9.23-0 LLM:v6.9.22-0 LLM:v6.9.19-0 LLM:v6.9.18-0 LLM:v6.9.17-0 LLM:v6.9.16-0 LLM:v6.9.15-0 LLM:v6.9.14-0 LLM:v6.9.13-1 LLM:v6.9.13-0 LLM:v6.9.12-0 LLM:v6.9.10-1 LLM:v6.9.10-0 LLM:v6.9.9-0 LLM:v6.9.8-0 LLM:v6.9.5-0 LLM:v6.9.4-2 LLM:v6.9.4-1 LLM:v6.9.4-0 LLM:v6.9.2-0 LLM:v6.9.1-0 LLM:v6.9.0-0 LLM:v6.8.55-0 LLM:v6.8.54-0 LLM:v6.8.53-0 LLM:v6.8.52-1 LLM:v6.8.52-0 LLM:v6.8.51-1 LLM:v6.8.51-0 LLM:v6.8.50-0 LLM:v6.8.49-0 LLM:v6.8.48-0 LLM:v6.8.47-0 LLM:v6.8.46-0 LLM:v6.8.45-0 LLM:v6.8.44-0 LLM:v6.8.42-1 LLM:v6.8.42-0 LLM:v6.8.41-0 LLM:v6.8.40-1 LLM:v6.8.40-0 LLM:v6.8.39-1 LLM:v6.8.39-0 LLM:v6.8.37-0 LLM:v6.8.36-0 LLM:v6.8.35-0 LLM:v6.8.34-0 LLM:v6.8.33-0 LLM:v6.8.32-0 LLM:v6.8.31-0 LLM:v6.8.30-0 LLM:v6.8.28-1 LLM:v6.8.28-0 LLM:v6.8.27-0 LLM:v6.8.26-0 LLM:v6.8.25-0 LLM:v6.8.24-0 LLM:v6.8.23-0 LLM:v6.8.22-0 LLM:v6.8.21-0 LLM:v6.8.18-1 LLM:v6.8.18-0 LLM:v6.8.16-0 LLM:v6.8.15-1 LLM:v6.8.15-0 LLM:v6.8.14-0 LLM:v6.8.13-1 LLM:v6.8.13-0 LLM:v6.8.12-0 LLM:v6.8.11-1 LLM:v6.8.11-0 LLM:v6.8.9-0 LLM:v6.8.8-0 LLM:v6.8.7-0 LLM:v6.8.6-0 LLM:v6.8.5-1 LLM:v6.8.5-0 LLM:v6.8.4-0 LLM:v6.8.3-0 LLM:v6.8.2-0 LLM:v6.8.1-1 LLM:v6.8.1-0 LLM:v6.8.0-0 LLM:v6.7.53-0 LLM:v6.7.52-1 LLM:v6.7.52-0 LLM:v6.7.51-0 LLM:v6.7.50-0 LLM:v6.7.48-1 LLM:v6.7.48-0 LLM:v6.7.45-0 LLM:v6.7.42-0 LLM:v6.7.41-0 LLM:v6.7.40-0 LLM:v6.7.39-0 LLM:v6.7.38-0 LLM:v6.7.37-0 LLM:v6.7.36-0 LLM:v6.7.34-0 LLM:v6.7.33-0 LLM:v6.7.32-0 LLM:v6.7.31-0 LLM:v6.7.26-0 LLM:v6.7.25-0 LLM:v6.7.24-0 LLM:v6.7.22-0 LLM:v6.7.20-0 LLM:v6.7.16-0 LLM:v6.7.15-0 LLM:v6.7.11-0 LLM:v6.7.9-0 LLM:v6.7.6-1 LLM:v6.7.6-0 LLM:v6.7.5-0 LLM:v6.7.4-0 LLM:v6.7.2-0 LLM:v6.7.1-0 LLM:v6.7.0-0 LLM:v6.6.108-0 LLM:v6.6.107-0 LLM:v6.6.105-0 LLM:v6.6.103-0 LLM:v6.6.100-0 LLM:v6.6.98-0 LLM:v6.6.97-0 LLM:v6.6.93-0 LLM:v6.6.92-0 LLM:v6.6.91-0 LLM:v6.6.90-0 LLM:v6.6.89-0 LLM:v6.6.88-0 LLM:v6.6.86-0 LLM:v6.6.85-0 LLM:v6.6.84-0 LLM:v6.6.83-0 LLM:v6.6.82-0 LLM:v6.6.81-0 LLM:v6.6.80-0 LLM:v6.6.78-0 LLM:v6.6.77-1 LLM:v6.6.77-0 LLM:v6.6.76-0 LLM:v6.6.73-0 LLM:v6.6.71-0 LLM:v6.6.69-0 LLM:v6.6.68-0 LLM:v6.6.67-0 LLM:v6.6.63-0 LLM:v6.6.62-0 LLM:v6.6.61-0 LLM:v6.6.60-0 LLM:v6.6.59-0 LLM:v6.6.58-0 LLM:v6.6.57-0 LLM:v6.6.56-0 LLM:v6.6.55-0 LLM:v6.6.54-0 LLM:v6.6.53-0 LLM:v6.6.52-0 LLM:v6.6.51-0 LLM:v6.6.50-2 LLM:v6.6.50-1 LLM:v6.6.50-0 LLM:v6.6.49-0 LLM:v6.6.48-0 LLM:v6.6.47-0 LLM:v6.6.46-0 LLM:v6.6.45-0 LLM:v6.6.44-0 LLM:v6.6.43-0 LLM:v6.6.42-0 LLM:v6.6.41-0 LLM:v6.6.40-0 LLM:v6.6.39-0 LLM:v6.6.38-0 LLM:v6.6.36-0 LLM:v6.6.35-0 LLM:v6.6.34-0 LLM:v6.6.33-0 LLM:v6.6.32-0 LLM:v6.6.31-0 LLM:v6.6.30-0 LLM:v6.6.29-1 LLM:v6.6.29-0 LLM:v6.6.28-0 LLM:v6.6.26-0 LLM:v6.6.25-1 LLM:v6.6.25-0 LLM:v6.6.24-0 LLM:v6.6.23-0 LLM:v6.6.22-0 LLM:v6.6.21-0 LLM:v6.6.20-0 LLM:v6.6.19-0 LLM:v6.6.18-0 LLM:v6.6.17-0 LLM:v6.6.16-0 LLM:v6.6.15-0 LLM:v6.6.14-0 LLM:v6.6.12-0 LLM:v6.6.11-1 LLM:v6.6.11-0 LLM:v6.6.10-0 LLM:v6.6.9-1 LLM:v6.6.9-0 LLM:v6.6.8-1 LLM:v6.6.8-0 LLM:v6.6.7-0 LLM:v6.6.6-0 LLM:v6.6.5-1 LLM:v6.6.5-0 LLM:v6.6.4-0 LLM:v6.6.3-0 LLM:v6.6.2-0 LLM:v6.6.1-1 LLM:v6.6.1-0 LLM:v6.6.0-1 LLM:v6.6.0-0 LLM:v6.5.65-1 LLM:v6.5.65-0 LLM:v6.5.64-0 LLM:v6.5.63-1 LLM:v6.5.62-1 LLM:v6.5.61-1 LLM:v6.5.61-0 LLM:v6.5.60-0 LLM:v6.5.59-0 LLM:v6.5.58-0 LLM:v6.5.57-0 LLM:v6.5.56-0 LLM:v6.5.55-0 LLM:v6.5.54-0 LLM:v6.5.53-2 LLM:v6.5.53-1 LLM:v6.5.53-0 LLM:v6.5.52-0 LLM:v6.5.51-0 LLM:v6.5.50-0 LLM:v6.5.49-1 LLM:v6.5.48-1 LLM:v6.5.48-0 LLM:v6.5.47-0 LLM:v6.5.46-0 LLM:v6.5.45-0 LLM:v6.5.44-0 LLM:v6.5.43-0 LLM:v6.5.42-0 LLM:v6.5.41-0 LLM:v6.5.40-0 LLM:v6.5.39-0 LLM:v6.5.38-0 LLM:v6.5.37-0 LLM:v6.5.36-0 LLM:v6.5.35-0 LLM:v6.5.34-0 LLM:v6.5.33-0 LLM:v6.5.32-0 LLM:v6.5.31-0 LLM:v6.5.30-0 LLM:v6.5.29-0
...
compare: LLM:v6.6.1-0
Branches Tags
LLM:main LLM:pr-59-resolve-conflicts
LLM:v6.9.28-0 LLM:v6.9.27-0 LLM:v6.9.23-0 LLM:v6.9.22-0 LLM:v6.9.19-0 LLM:v6.9.18-0 LLM:v6.9.17-0 LLM:v6.9.16-0 LLM:v6.9.15-0 LLM:v6.9.14-0 LLM:v6.9.13-1 LLM:v6.9.13-0 LLM:v6.9.12-0 LLM:v6.9.10-1 LLM:v6.9.10-0 LLM:v6.9.9-0 LLM:v6.9.8-0 LLM:v6.9.5-0 LLM:v6.9.4-2 LLM:v6.9.4-1 LLM:v6.9.4-0 LLM:v6.9.2-0 LLM:v6.9.1-0 LLM:v6.9.0-0 LLM:v6.8.55-0 LLM:v6.8.54-0 LLM:v6.8.53-0 LLM:v6.8.52-1 LLM:v6.8.52-0 LLM:v6.8.51-1 LLM:v6.8.51-0 LLM:v6.8.50-0 LLM:v6.8.49-0 LLM:v6.8.48-0 LLM:v6.8.47-0 LLM:v6.8.46-0 LLM:v6.8.45-0 LLM:v6.8.44-0 LLM:v6.8.42-1 LLM:v6.8.42-0 LLM:v6.8.41-0 LLM:v6.8.40-1 LLM:v6.8.40-0 LLM:v6.8.39-1 LLM:v6.8.39-0 LLM:v6.8.37-0 LLM:v6.8.36-0 LLM:v6.8.35-0 LLM:v6.8.34-0 LLM:v6.8.33-0 LLM:v6.8.32-0 LLM:v6.8.31-0 LLM:v6.8.30-0 LLM:v6.8.28-1 LLM:v6.8.28-0 LLM:v6.8.27-0 LLM:v6.8.26-0 LLM:v6.8.25-0 LLM:v6.8.24-0 LLM:v6.8.23-0 LLM:v6.8.22-0 LLM:v6.8.21-0 LLM:v6.8.18-1 LLM:v6.8.18-0 LLM:v6.8.16-0 LLM:v6.8.15-1 LLM:v6.8.15-0 LLM:v6.8.14-0 LLM:v6.8.13-1 LLM:v6.8.13-0 LLM:v6.8.12-0 LLM:v6.8.11-1 LLM:v6.8.11-0 LLM:v6.8.9-0 LLM:v6.8.8-0 LLM:v6.8.7-0 LLM:v6.8.6-0 LLM:v6.8.5-1 LLM:v6.8.5-0 LLM:v6.8.4-0 LLM:v6.8.3-0 LLM:v6.8.2-0 LLM:v6.8.1-1 LLM:v6.8.1-0 LLM:v6.8.0-0 LLM:v6.7.53-0 LLM:v6.7.52-1 LLM:v6.7.52-0 LLM:v6.7.51-0 LLM:v6.7.50-0 LLM:v6.7.48-1 LLM:v6.7.48-0 LLM:v6.7.45-0 LLM:v6.7.42-0 LLM:v6.7.41-0 LLM:v6.7.40-0 LLM:v6.7.39-0 LLM:v6.7.38-0 LLM:v6.7.37-0 LLM:v6.7.36-0 LLM:v6.7.34-0 LLM:v6.7.33-0 LLM:v6.7.32-0 LLM:v6.7.31-0 LLM:v6.7.26-0 LLM:v6.7.25-0 LLM:v6.7.24-0 LLM:v6.7.22-0 LLM:v6.7.20-0 LLM:v6.7.16-0 LLM:v6.7.15-0 LLM:v6.7.11-0 LLM:v6.7.9-0 LLM:v6.7.6-1 LLM:v6.7.6-0 LLM:v6.7.5-0 LLM:v6.7.4-0 LLM:v6.7.2-0 LLM:v6.7.1-0 LLM:v6.7.0-0 LLM:v6.6.108-0 LLM:v6.6.107-0 LLM:v6.6.105-0 LLM:v6.6.103-0 LLM:v6.6.100-0 LLM:v6.6.98-0 LLM:v6.6.97-0 LLM:v6.6.93-0 LLM:v6.6.92-0 LLM:v6.6.91-0 LLM:v6.6.90-0 LLM:v6.6.89-0 LLM:v6.6.88-0 LLM:v6.6.86-0 LLM:v6.6.85-0 LLM:v6.6.84-0 LLM:v6.6.83-0 LLM:v6.6.82-0 LLM:v6.6.81-0 LLM:v6.6.80-0 LLM:v6.6.78-0 LLM:v6.6.77-1 LLM:v6.6.77-0 LLM:v6.6.76-0 LLM:v6.6.73-0 LLM:v6.6.71-0 LLM:v6.6.69-0 LLM:v6.6.68-0 LLM:v6.6.67-0 LLM:v6.6.63-0 LLM:v6.6.62-0 LLM:v6.6.61-0 LLM:v6.6.60-0 LLM:v6.6.59-0 LLM:v6.6.58-0 LLM:v6.6.57-0 LLM:v6.6.56-0 LLM:v6.6.55-0 LLM:v6.6.54-0 LLM:v6.6.53-0 LLM:v6.6.52-0 LLM:v6.6.51-0 LLM:v6.6.50-2 LLM:v6.6.50-1 LLM:v6.6.50-0 LLM:v6.6.49-0 LLM:v6.6.48-0 LLM:v6.6.47-0 LLM:v6.6.46-0 LLM:v6.6.45-0 LLM:v6.6.44-0 LLM:v6.6.43-0 LLM:v6.6.42-0 LLM:v6.6.41-0 LLM:v6.6.40-0 LLM:v6.6.39-0 LLM:v6.6.38-0 LLM:v6.6.36-0 LLM:v6.6.35-0 LLM:v6.6.34-0 LLM:v6.6.33-0 LLM:v6.6.32-0 LLM:v6.6.31-0 LLM:v6.6.30-0 LLM:v6.6.29-1 LLM:v6.6.29-0 LLM:v6.6.28-0 LLM:v6.6.26-0 LLM:v6.6.25-1 LLM:v6.6.25-0 LLM:v6.6.24-0 LLM:v6.6.23-0 LLM:v6.6.22-0 LLM:v6.6.21-0 LLM:v6.6.20-0 LLM:v6.6.19-0 LLM:v6.6.18-0 LLM:v6.6.17-0 LLM:v6.6.16-0 LLM:v6.6.15-0 LLM:v6.6.14-0 LLM:v6.6.12-0 LLM:v6.6.11-1 LLM:v6.6.11-0 LLM:v6.6.10-0 LLM:v6.6.9-1 LLM:v6.6.9-0 LLM:v6.6.8-1 LLM:v6.6.8-0 LLM:v6.6.7-0 LLM:v6.6.6-0 LLM:v6.6.5-1 LLM:v6.6.5-0 LLM:v6.6.4-0 LLM:v6.6.3-0 LLM:v6.6.2-0 LLM:v6.6.1-1 LLM:v6.6.1-0 LLM:v6.6.0-1 LLM:v6.6.0-0 LLM:v6.5.65-1 LLM:v6.5.65-0 LLM:v6.5.64-0 LLM:v6.5.63-1 LLM:v6.5.62-1 LLM:v6.5.61-1 LLM:v6.5.61-0 LLM:v6.5.60-0 LLM:v6.5.59-0 LLM:v6.5.58-0 LLM:v6.5.57-0 LLM:v6.5.56-0 LLM:v6.5.55-0 LLM:v6.5.54-0 LLM:v6.5.53-2 LLM:v6.5.53-1 LLM:v6.5.53-0 LLM:v6.5.52-0 LLM:v6.5.51-0 LLM:v6.5.50-0 LLM:v6.5.49-1 LLM:v6.5.48-1 LLM:v6.5.48-0 LLM:v6.5.47-0 LLM:v6.5.46-0 LLM:v6.5.45-0 LLM:v6.5.44-0 LLM:v6.5.43-0 LLM:v6.5.42-0 LLM:v6.5.41-0 LLM:v6.5.40-0 LLM:v6.5.39-0 LLM:v6.5.38-0 LLM:v6.5.37-0 LLM:v6.5.36-0 LLM:v6.5.35-0 LLM:v6.5.34-0 LLM:v6.5.33-0 LLM:v6.5.32-0 LLM:v6.5.31-0 LLM:v6.5.30-0 LLM:v6.5.29-0

37 Commits
v6.5.64-0 ... v6.6.1-0

Author SHA1 Message Date
Luis Pater
242b4d5754 Merge pull request #19 from router-for-me/plus 
v6.6.1
 2025-12-11 21:35:16 +08:00
Luis Pater
4ce7c61a17 Merge branch 'main' into plus 2025-12-11 21:33:49 +08:00
Luis Pater
a74ee3f319 Merge pull request #481 from sususu98/fix/increase-buffer-size 
fix: increase buffer size for stream scanners to 50MB across multiple executors
 2025-12-11 21:20:54 +08:00
Luis Pater
564bcbaa54 Merge pull request #487 from router-for-me/amp 
fix(amp): set status on claude stream errors
 2025-12-11 21:18:19 +08:00
hkfires
88bdd25f06 fix(amp): set status on claude stream errors 2025-12-11 20:12:06 +08:00
hkfires
e79f65fd8e refactor(thinking): use parentheses for metadata suffix 2025-12-11 18:39:07 +08:00
Luis Pater
2760989401 Merge pull request #485 from router-for-me/think 
Think
 2025-12-11 18:27:00 +08:00
hkfires
facfe7c518 refactor(thinking): use bracket tags for thinking meta 
Align thinking suffix handling on a single bracket-style marker.

NormalizeThinkingModel strips a terminal `[value]` segment from
model identifiers and turns it into either a thinking budget (for
numeric values) or a reasoning effort hint (for strings). Emission
of `ThinkingIncludeThoughtsMetadataKey` is removed.

Executor helpers and the example config are updated so their
comments reference the new `[value]` suffix format instead of the
legacy dash variants.

BREAKING CHANGE: dash-based thinking suffixes (`-thinking`,
`-thinking-N`, `-reasoning`, `-nothinking`) are no longer parsed
for thinking metadata; only `[value]` annotations are recognized.
 2025-12-11 18:17:28 +08:00
hkfires
6285459c08 fix(runtime): unify claude thinking config resolution 2025-12-11 17:20:44 +08:00
hkfires
21bbceca0c docs(runtime): document reasoning effort precedence 2025-12-11 16:35:36 +08:00
hkfires
f6300c72b7 fix(runtime): validate thinking config in iflow and qwen 2025-12-11 16:21:50 +08:00
hkfires
007572b58e fix(util): do not strip thinking suffix on registered models 
NormalizeThinkingModel now checks ModelSupportsThinking before removing
"-thinking" or "-thinking-<ver>", avoiding accidental parsing of model
names where the suffix is part of the official id (e.g., kimi-k2-thinking,
qwen3-235b-a22b-thinking-2507).

The registry adds ThinkingSupport metadata for several models and
propagates it via ModelInfo (e.g., kimi-k2-thinking, deepseek-r1,
qwen3-235b-a22b-thinking-2507, minimax-m2), enabling accurate detection
of thinking-capable models and correcting base model inference.
 2025-12-11 15:52:14 +08:00
hkfires
3a81ab22fd fix(runtime): unify reasoning effort metadata overrides 2025-12-11 14:35:05 +08:00
hkfires
519da2e042 fix(runtime): validate reasoning effort levels 2025-12-11 12:36:54 +08:00
hkfires
169f4295d0 fix(util): align reasoning effort handling with registry 2025-12-11 12:20:12 +08:00
hkfires
d06d0eab2f fix(util): centralize reasoning effort normalization 2025-12-11 12:14:51 +08:00
hkfires
3ffd120ae9 feat(runtime): add thinking config normalization 2025-12-11 11:51:33 +08:00
hkfires
a03d514095 feat(registry): add thinking metadata for models 2025-12-11 11:28:44 +08:00
Luis Pater
69fccf0015 Merge pull request #18 from Ravens2121/master 
Kiro Executor Stability and API Compatibility Improvements
 2025-12-11 08:35:22 +08:00
Luis Pater
1da03bfe15 Merge pull request #479 from router-for-me/claude 
fix(claude): prevent final events when no content streamed
 2025-12-11 08:18:59 +08:00
Ravens2121
6133bac226 feat(kiro): enhance Kiro executor stability and compatibility 
## Changes Overview
This commit includes multiple improvements to the Kiro executor for better
stability, API compatibility, and code quality.

## Detailed Changes

### 1. Output Token Calculation Improvement (lines 317-330)
- Replace simple len(content)/4 estimation with tiktoken-based calculation
- Add fallback to character count estimation if tiktoken fails
- Improves token counting accuracy for usage tracking

### 2. Stream Handler Panic Recovery (lines 528-533)
- Add defer/recover block in streamToChannel goroutine
- Prevents single request crashes from affecting the entire service

### 3. Struct Field Reordering (lines 670-673)
- Reorder kiroToolResult struct fields: Content, Status, ToolUseID
- Ensures consistency with API expectations

### 4. Message Merging Function (lines 778-780, 2356-2483)
- Add mergeAdjacentMessages() to combine consecutive messages with same role
- Add helper functions: mergeMessageContent(), blockToMap(), createMergedMessage()
- Required by Kiro API which doesn't allow adjacent messages from same role

### 5. Empty Content Handling (lines 791-800)
- Add default content for empty history messages
- User messages with tool results: "Tool results provided."
- User messages without tool results: "Continue"

### 6. Assistant Last Message Handling (lines 811-830)
- Detect when last message is from assistant
- Create synthetic "Continue" user message to satisfy Kiro API requirements
- Kiro API requires currentMessage to be userInputMessage type

### 7. Duplicate Content Event Detection (lines 1650-1660)
- Track lastContentEvent to detect duplicate streaming events
- Skip redundant events to prevent duplicate content in responses
- Based on AIClient-2-API implementation for Kiro

### 8. Streaming Token Calculation Enhancement (lines 1785-1817)
- Add accumulatedContent buffer for streaming token calculation
- Use tiktoken for accurate output token counting during streaming
- Add fallback to character count estimation with proper logging

### 9. JSON Repair Enhancement (lines 2665-2818)
- Implement conservative JSON repair strategy
- First try to parse JSON directly - if valid, return unchanged
- Add bracket balancing detection and repair
- Only repair when necessary to avoid corrupting valid JSON
- Validate repaired JSON before returning

### 10. HELIOS_CHK Filtering Removal (lines 2500-2504, 3004-3039)
- Remove filterHeliosDebugInfo function
- Remove heliosDebugPattern regex
- HELIOS_CHK fields now handled by client-side processing

### 11. Comment Translation
- Translate Chinese comments to English for code consistency
- Affected areas: token calculation, buffer handling, message processing
 2025-12-11 08:13:33 +08:00
Ravens2121
f302be5ce6 Merge branch 'router-for-me:main' into master 2025-12-11 05:40:10 +08:00
Ravens2121
cd4e84a360 feat(kiro): enhance request format, stream handling, and usage tracking 
## English Description

### Request Format Fixes
- Fix conversationState field order (chatTriggerType must be first)
- Add conditional profileArn inclusion based on auth method
- builder-id auth (AWS SSO) doesn't require profileArn
- social auth (Google OAuth) requires profileArn

### Stream Processing Enhancements
- Add headersLen boundary validation to prevent slice out of bounds
- Handle incomplete tool use at EOF by flushing pending data
- Separate message_delta and message_stop events for proper streaming
- Add error logging for JSON unmarshal failures

### JSON Repair Improvements
- Add escapeNewlinesInStrings() to handle control characters in JSON strings
- Remove incorrect unquotedKeyPattern that broke valid JSON content
- Fix handling of streaming fragments with embedded newlines/tabs

### Debug Info Filtering (Optional)
- Add filterHeliosDebugInfo() to remove [HELIOS_CHK] blocks
- Pattern matches internal state tracking from Kiro/Amazon Q
- Currently disabled pending further testing

### Usage Tracking
- Add usage information extraction in message_delta response
- Include prompt_tokens, completion_tokens, total_tokens in OpenAI format

---

## 中文描述

### 请求格式修复
- 修复 conversationState 字段顺序（chatTriggerType 必须在第一位）
- 根据认证方式条件性包含 profileArn
- builder-id 认证（AWS SSO）不需要 profileArn
- social 认证（Google OAuth）需要 profileArn

### 流处理增强
- 添加 headersLen 边界验证，防止切片越界
- 在 EOF 时处理未完成的工具调用，刷新待处理数据
- 分离 message_delta 和 message_stop 事件以实现正确的流式传输
- 添加 JSON 反序列化失败的错误日志

### JSON 修复改进
- 添加 escapeNewlinesInStrings() 处理 JSON 字符串中的控制字符
- 移除错误的 unquotedKeyPattern，该模式会破坏有效的 JSON 内容
- 修复包含嵌入换行符/制表符的流式片段处理

### 调试信息过滤（可选）
- 添加 filterHeliosDebugInfo() 移除 [HELIOS_CHK] 块
- 模式匹配来自 Kiro/Amazon Q 的内部状态跟踪信息
- 目前已禁用，等待进一步测试

### 使用量跟踪
- 在 message_delta 响应中添加 usage 信息提取
- 以 OpenAI 格式包含 prompt_tokens、completion_tokens、total_tokens
 2025-12-11 05:37:22 +08:00
Luis Pater
4360ed8a7b Merge branch 'router-for-me:main' into main 2025-12-11 03:17:55 +08:00
Luis Pater
423ce97665 feat(util): implement dynamic thinking suffix normalization and refactor budget resolution logic 
- Added support for parsing and normalizing dynamic thinking model suffixes.
- Centralized budget resolution across executors and payload helpers.
- Retired legacy Gemini-specific thinking handlers in favor of unified logic.
- Updated executors to use metadata-based thinking configuration.
- Added `ResolveOriginalModel` utility for resolving normalized upstream models using request metadata.
- Updated executors (Gemini, Codex, iFlow, OpenAI, Qwen) to incorporate upstream model resolution and substitute model values in payloads and request URLs.
- Ensured fallbacks handle cases with missing or malformed metadata to derive models robustly.
- Refactored upstream model resolution to dynamically incorporate metadata for selecting and normalizing models.
- Improved handling of thinking configurations and model overrides in executors.
- Removed hardcoded thinking model entries and migrated logic to metadata-based resolution.
- Updated payload mutations to always include the resolved model.
 2025-12-11 03:10:50 +08:00
Luis Pater
b27a175fef Merge pull request #17 from Ravens2121/master 
由AI进行更改修复了Kiro供应商的Claude协议与OpenAI协议。（对比AIClient-2-API项目进行变更）
 2025-12-11 01:31:47 +08:00
Ravens
8d5f89ccfd fix(kiro): fix translator format mismatch for OpenAI protocol 
Amp-Thread-ID: https://ampcode.com/threads/T-019b092b-f2de-72a1-b428-72511c0de628
Co-authored-by: Amp <amp@ampcode.com>
 2025-12-11 01:15:00 +08:00
Ravens
084e2666cb fix(kiro): add SSE event: prefix for Claude client compatibility 
Amp-Thread-ID: https://ampcode.com/threads/T-019b08fc-ff96-766e-a942-63dd35ed28c6
Co-authored-by: Amp <amp@ampcode.com>
 2025-12-11 00:14:20 +08:00
Luis Pater
2eb2dbb266 Merge branch 'router-for-me:main' into main 2025-12-10 23:56:59 +08:00
Luis Pater
e717939edb Fixed: #478 
feat(antigravity): add support for inline image data in client responses
 2025-12-10 23:55:53 +08:00
Ravens2121
7758a86d1e Merge branch 'router-for-me:main' into main 2025-12-10 23:25:23 +08:00
sususu
76c563d161 fix(executor): increase buffer size for stream scanners to 50MB across multiple executors 2025-12-10 23:20:04 +08:00
hkfires
a89514951f fix(claude): prevent final events when no content streamed 2025-12-10 22:19:55 +08:00
Ravens2121
1770c491db Merge branch 'router-for-me:main' into main 2025-12-07 22:25:42 +08:00
Your Name
2bf9e08b31 style(kiro): convert Chinese comments to English in base64 image handling 2025-12-07 21:51:24 +08:00
Ravens2121
f56bfaa689 Merge branch 'router-for-me:main' into main 2025-12-07 21:35:16 +08:00
Your Name
5d716dc796 fix(kiro): 修复 base64 图片格式转换问题 2025-12-07 21:34:44 +08:00
29 changed files with 1612 additions and 794 deletions
Expand all files Collapse all files

2
config.example.yaml
View File

@@ -105,7 +105,7 @@ ws-auth: false
# excluded-models: # excluded-models:
# - "claude-opus-4-5-20251101" # exclude specific models (exact match) # - "claude-opus-4-5-20251101" # exclude specific models (exact match)
# - "claude-3-*" # wildcard matching prefix (e.g. claude-3-7-sonnet-20250219) # - "claude-3-*" # wildcard matching prefix (e.g. claude-3-7-sonnet-20250219)
# - "*-think" # wildcard matching suffix (e.g. claude-opus-4-5-thinking) # - "*-thinking" # wildcard matching suffix (e.g. claude-opus-4-5-thinking)
# - "*haiku*" # wildcard matching substring (e.g. claude-3-5-haiku-20241022) # - "*haiku*" # wildcard matching substring (e.g. claude-3-5-haiku-20241022)
# Kiro (AWS CodeWhisperer) configuration # Kiro (AWS CodeWhisperer) configuration

4
internal/api/modules/amp/fallback_handlers.go
View File

@@ -133,8 +133,8 @@ func (fh *FallbackHandler) WrapHandler(handler gin.HandlerFunc) gin.HandlerFunc
return return
} }
// Normalize model (handles Gemini thinking suffixes) // Normalize model (handles dynamic thinking suffixes)
normalizedModel, _ := util.NormalizeGeminiThinkingModel(modelName) normalizedModel, _ := util.NormalizeThinkingModel(modelName)
// Track resolved model for logging (may change if mapping is applied) // Track resolved model for logging (may change if mapping is applied)
resolvedModel := normalizedModel resolvedModel := normalizedModel

365
internal/registry/model_definitions.go
View File

@@ -16,6 +16,7 @@ func GetClaudeModels() []*ModelInfo {
DisplayName: "Claude 4.5 Haiku", DisplayName: "Claude 4.5 Haiku",
ContextLength: 200000, ContextLength: 200000,
MaxCompletionTokens: 64000, MaxCompletionTokens: 64000,
// Thinking: not supported for Haiku models
}, },
{ {
ID: "claude-sonnet-4-5-20250929", ID: "claude-sonnet-4-5-20250929",
@@ -26,60 +27,6 @@ func GetClaudeModels() []*ModelInfo {
DisplayName: "Claude 4.5 Sonnet", DisplayName: "Claude 4.5 Sonnet",
ContextLength: 200000, ContextLength: 200000,
MaxCompletionTokens: 64000, MaxCompletionTokens: 64000,
},
{
ID: "claude-sonnet-4-5-thinking",
Object: "model",
Created: 1759104000, // 2025-09-29
OwnedBy: "anthropic",
Type: "claude",
DisplayName: "Claude 4.5 Sonnet Thinking",
ContextLength: 200000,
MaxCompletionTokens: 64000,
Thinking: &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
},
{
ID: "claude-opus-4-5-thinking",
Object: "model",
Created: 1761955200, // 2025-11-01
OwnedBy: "anthropic",
Type: "claude",
DisplayName: "Claude 4.5 Opus Thinking",
ContextLength: 200000,
MaxCompletionTokens: 64000,
Thinking: &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
},
{
ID: "claude-opus-4-5-thinking-low",
Object: "model",
Created: 1761955200, // 2025-11-01
OwnedBy: "anthropic",
Type: "claude",
DisplayName: "Claude 4.5 Opus Thinking Low",
ContextLength: 200000,
MaxCompletionTokens: 64000,
Thinking: &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
},
{
ID: "claude-opus-4-5-thinking-medium",
Object: "model",
Created: 1761955200, // 2025-11-01
OwnedBy: "anthropic",
Type: "claude",
DisplayName: "Claude 4.5 Opus Thinking Medium",
ContextLength: 200000,
MaxCompletionTokens: 64000,
Thinking: &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
},
{
ID: "claude-opus-4-5-thinking-high",
Object: "model",
Created: 1761955200, // 2025-11-01
OwnedBy: "anthropic",
Type: "claude",
DisplayName: "Claude 4.5 Opus Thinking High",
ContextLength: 200000,
MaxCompletionTokens: 64000,
Thinking: &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true}, Thinking: &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
}, },
{ {
@@ -92,6 +39,7 @@ func GetClaudeModels() []*ModelInfo {
Description: "Premium model combining maximum intelligence with practical performance", Description: "Premium model combining maximum intelligence with practical performance",
ContextLength: 200000, ContextLength: 200000,
MaxCompletionTokens: 64000, MaxCompletionTokens: 64000,
Thinking: &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
}, },
{ {
ID: "claude-opus-4-1-20250805", ID: "claude-opus-4-1-20250805",
@@ -102,6 +50,7 @@ func GetClaudeModels() []*ModelInfo {
DisplayName: "Claude 4.1 Opus", DisplayName: "Claude 4.1 Opus",
ContextLength: 200000, ContextLength: 200000,
MaxCompletionTokens: 32000, MaxCompletionTokens: 32000,
Thinking: &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
}, },
{ {
ID: "claude-opus-4-20250514", ID: "claude-opus-4-20250514",
@@ -112,6 +61,7 @@ func GetClaudeModels() []*ModelInfo {
DisplayName: "Claude 4 Opus", DisplayName: "Claude 4 Opus",
ContextLength: 200000, ContextLength: 200000,
MaxCompletionTokens: 32000, MaxCompletionTokens: 32000,
Thinking: &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
}, },
{ {
ID: "claude-sonnet-4-20250514", ID: "claude-sonnet-4-20250514",
@@ -122,6 +72,7 @@ func GetClaudeModels() []*ModelInfo {
DisplayName: "Claude 4 Sonnet", DisplayName: "Claude 4 Sonnet",
ContextLength: 200000, ContextLength: 200000,
MaxCompletionTokens: 64000, MaxCompletionTokens: 64000,
Thinking: &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
}, },
{ {
ID: "claude-3-7-sonnet-20250219", ID: "claude-3-7-sonnet-20250219",
@@ -132,6 +83,7 @@ func GetClaudeModels() []*ModelInfo {
DisplayName: "Claude 3.7 Sonnet", DisplayName: "Claude 3.7 Sonnet",
ContextLength: 128000, ContextLength: 128000,
MaxCompletionTokens: 8192, MaxCompletionTokens: 8192,
Thinking: &ThinkingSupport{Min: 1024, Max: 100000, ZeroAllowed: false, DynamicAllowed: true},
}, },
{ {
ID: "claude-3-5-haiku-20241022", ID: "claude-3-5-haiku-20241022",
@@ -142,6 +94,7 @@ func GetClaudeModels() []*ModelInfo {
DisplayName: "Claude 3.5 Haiku", DisplayName: "Claude 3.5 Haiku",
ContextLength: 128000, ContextLength: 128000,
MaxCompletionTokens: 8192, MaxCompletionTokens: 8192,
// Thinking: not supported for Haiku models
}, },
} }
} }
@@ -529,58 +482,7 @@ func GetOpenAIModels() []*ModelInfo {
ContextLength: 400000, ContextLength: 400000,
MaxCompletionTokens: 128000, MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"}, SupportedParameters: []string{"tools"},
}, Thinking: &ThinkingSupport{Levels: []string{"minimal", "low", "medium", "high"}},
{
ID: "gpt-5-minimal",
Object: "model",
Created: 1754524800,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5-2025-08-07",
DisplayName: "GPT 5 Minimal",
Description: "Stable version of GPT 5, The best model for coding and agentic tasks across domains.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5-low",
Object: "model",
Created: 1754524800,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5-2025-08-07",
DisplayName: "GPT 5 Low",
Description: "Stable version of GPT 5, The best model for coding and agentic tasks across domains.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5-medium",
Object: "model",
Created: 1754524800,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5-2025-08-07",
DisplayName: "GPT 5 Medium",
Description: "Stable version of GPT 5, The best model for coding and agentic tasks across domains.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5-high",
Object: "model",
Created: 1754524800,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5-2025-08-07",
DisplayName: "GPT 5 High",
Description: "Stable version of GPT 5, The best model for coding and agentic tasks across domains.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
}, },
{ {
ID: "gpt-5-codex", ID: "gpt-5-codex",
@@ -594,45 +496,7 @@ func GetOpenAIModels() []*ModelInfo {
ContextLength: 400000, ContextLength: 400000,
MaxCompletionTokens: 128000, MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"}, SupportedParameters: []string{"tools"},
}, Thinking: &ThinkingSupport{Levels: []string{"low", "medium", "high"}},
{
ID: "gpt-5-codex-low",
Object: "model",
Created: 1757894400,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5-2025-09-15",
DisplayName: "GPT 5 Codex Low",
Description: "Stable version of GPT 5 Codex, The best model for coding and agentic tasks across domains.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5-codex-medium",
Object: "model",
Created: 1757894400,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5-2025-09-15",
DisplayName: "GPT 5 Codex Medium",
Description: "Stable version of GPT 5 Codex, The best model for coding and agentic tasks across domains.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5-codex-high",
Object: "model",
Created: 1757894400,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5-2025-09-15",
DisplayName: "GPT 5 Codex High",
Description: "Stable version of GPT 5 Codex, The best model for coding and agentic tasks across domains.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
}, },
{ {
ID: "gpt-5-codex-mini", ID: "gpt-5-codex-mini",
@@ -646,32 +510,7 @@ func GetOpenAIModels() []*ModelInfo {
ContextLength: 400000, ContextLength: 400000,
MaxCompletionTokens: 128000, MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"}, SupportedParameters: []string{"tools"},
}, Thinking: &ThinkingSupport{Levels: []string{"low", "medium", "high"}},
{
ID: "gpt-5-codex-mini-medium",
Object: "model",
Created: 1762473600,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5-2025-11-07",
DisplayName: "GPT 5 Codex Mini Medium",
Description: "Stable version of GPT 5 Codex Mini: cheaper, faster, but less capable version of GPT 5 Codex.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5-codex-mini-high",
Object: "model",
Created: 1762473600,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5-2025-11-07",
DisplayName: "GPT 5 Codex Mini High",
Description: "Stable version of GPT 5 Codex Mini: cheaper, faster, but less capable version of GPT 5 Codex.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
}, },
{ {
ID: "gpt-5.1", ID: "gpt-5.1",
@@ -685,58 +524,7 @@ func GetOpenAIModels() []*ModelInfo {
ContextLength: 400000, ContextLength: 400000,
MaxCompletionTokens: 128000, MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"}, SupportedParameters: []string{"tools"},
}, Thinking: &ThinkingSupport{Levels: []string{"none", "low", "medium", "high"}},
{
ID: "gpt-5.1-none",
Object: "model",
Created: 1762905600,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5.1-2025-11-12",
DisplayName: "GPT 5.1 Nothink",
Description: "Stable version of GPT 5.1, The best model for coding and agentic tasks across domains.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5.1-low",
Object: "model",
Created: 1762905600,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5.1-2025-11-12",
DisplayName: "GPT 5 Low",
Description: "Stable version of GPT 5, The best model for coding and agentic tasks across domains.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5.1-medium",
Object: "model",
Created: 1762905600,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5.1-2025-11-12",
DisplayName: "GPT 5.1 Medium",
Description: "Stable version of GPT 5.1, The best model for coding and agentic tasks across domains.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5.1-high",
Object: "model",
Created: 1762905600,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5.1-2025-11-12",
DisplayName: "GPT 5.1 High",
Description: "Stable version of GPT 5.1, The best model for coding and agentic tasks across domains.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
}, },
{ {
ID: "gpt-5.1-codex", ID: "gpt-5.1-codex",
@@ -750,45 +538,7 @@ func GetOpenAIModels() []*ModelInfo {
ContextLength: 400000, ContextLength: 400000,
MaxCompletionTokens: 128000, MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"}, SupportedParameters: []string{"tools"},
}, Thinking: &ThinkingSupport{Levels: []string{"low", "medium", "high"}},
{
ID: "gpt-5.1-codex-low",
Object: "model",
Created: 1762905600,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5.1-2025-11-12",
DisplayName: "GPT 5.1 Codex Low",
Description: "Stable version of GPT 5.1 Codex, The best model for coding and agentic tasks across domains.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5.1-codex-medium",
Object: "model",
Created: 1762905600,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5.1-2025-11-12",
DisplayName: "GPT 5.1 Codex Medium",
Description: "Stable version of GPT 5.1 Codex, The best model for coding and agentic tasks across domains.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5.1-codex-high",
Object: "model",
Created: 1762905600,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5.1-2025-11-12",
DisplayName: "GPT 5.1 Codex High",
Description: "Stable version of GPT 5.1 Codex, The best model for coding and agentic tasks across domains.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
}, },
{ {
ID: "gpt-5.1-codex-mini", ID: "gpt-5.1-codex-mini",
@@ -802,34 +552,8 @@ func GetOpenAIModels() []*ModelInfo {
ContextLength: 400000, ContextLength: 400000,
MaxCompletionTokens: 128000, MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"}, SupportedParameters: []string{"tools"},
Thinking: &ThinkingSupport{Levels: []string{"low", "medium", "high"}},
}, },
{
ID: "gpt-5.1-codex-mini-medium",
Object: "model",
Created: 1762905600,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5.1-2025-11-12",
DisplayName: "GPT 5.1 Codex Mini Medium",
Description: "Stable version of GPT 5.1 Codex Mini: cheaper, faster, but less capable version of GPT 5.1 Codex.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5.1-codex-mini-high",
Object: "model",
Created: 1762905600,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5.1-2025-11-12",
DisplayName: "GPT 5.1 Codex Mini High",
Description: "Stable version of GPT 5.1 Codex Mini: cheaper, faster, but less capable version of GPT 5.1 Codex.",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{ {
ID: "gpt-5.1-codex-max", ID: "gpt-5.1-codex-max",
Object: "model", Object: "model",
@@ -842,58 +566,7 @@ func GetOpenAIModels() []*ModelInfo {
ContextLength: 400000, ContextLength: 400000,
MaxCompletionTokens: 128000, MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"}, SupportedParameters: []string{"tools"},
}, Thinking: &ThinkingSupport{Levels: []string{"low", "medium", "high", "xhigh"}},
{
ID: "gpt-5.1-codex-max-low",
Object: "model",
Created: 1763424000,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5.1-max",
DisplayName: "GPT 5.1 Codex Max Low",
Description: "Stable version of GPT 5.1 Codex Max Low",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5.1-codex-max-medium",
Object: "model",
Created: 1763424000,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5.1-max",
DisplayName: "GPT 5.1 Codex Max Medium",
Description: "Stable version of GPT 5.1 Codex Max Medium",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5.1-codex-max-high",
Object: "model",
Created: 1763424000,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5.1-max",
DisplayName: "GPT 5.1 Codex Max High",
Description: "Stable version of GPT 5.1 Codex Max High",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
},
{
ID: "gpt-5.1-codex-max-xhigh",
Object: "model",
Created: 1763424000,
OwnedBy: "openai",
Type: "openai",
Version: "gpt-5.1-max",
DisplayName: "GPT 5.1 Codex Max XHigh",
Description: "Stable version of GPT 5.1 Codex Max XHigh",
ContextLength: 400000,
MaxCompletionTokens: 128000,
SupportedParameters: []string{"tools"},
}, },
} }
} }
@@ -950,6 +623,7 @@ func GetIFlowModels() []*ModelInfo {
DisplayName string DisplayName string
Description string Description string
Created int64 Created int64
Thinking *ThinkingSupport
}{ }{
{ID: "tstars2.0", DisplayName: "TStars-2.0", Description: "iFlow TStars-2.0 multimodal assistant", Created: 1746489600}, {ID: "tstars2.0", DisplayName: "TStars-2.0", Description: "iFlow TStars-2.0 multimodal assistant", Created: 1746489600},
{ID: "qwen3-coder-plus", DisplayName: "Qwen3-Coder-Plus", Description: "Qwen3 Coder Plus code generation", Created: 1753228800}, {ID: "qwen3-coder-plus", DisplayName: "Qwen3-Coder-Plus", Description: "Qwen3 Coder Plus code generation", Created: 1753228800},
@@ -959,17 +633,17 @@ func GetIFlowModels() []*ModelInfo {
{ID: "kimi-k2-0905", DisplayName: "Kimi-K2-Instruct-0905", Description: "Moonshot Kimi K2 instruct 0905", Created: 1757030400}, {ID: "kimi-k2-0905", DisplayName: "Kimi-K2-Instruct-0905", Description: "Moonshot Kimi K2 instruct 0905", Created: 1757030400},
{ID: "glm-4.6", DisplayName: "GLM-4.6", Description: "Zhipu GLM 4.6 general model", Created: 1759190400}, {ID: "glm-4.6", DisplayName: "GLM-4.6", Description: "Zhipu GLM 4.6 general model", Created: 1759190400},
{ID: "kimi-k2", DisplayName: "Kimi-K2", Description: "Moonshot Kimi K2 general model", Created: 1752192000}, {ID: "kimi-k2", DisplayName: "Kimi-K2", Description: "Moonshot Kimi K2 general model", Created: 1752192000},
{ID: "kimi-k2-thinking", DisplayName: "Kimi-K2-Thinking", Description: "Moonshot Kimi K2 general model", Created: 1762387200}, {ID: "kimi-k2-thinking", DisplayName: "Kimi-K2-Thinking", Description: "Moonshot Kimi K2 thinking model", Created: 1762387200, Thinking: &ThinkingSupport{Levels: []string{"low", "medium", "high"}}},
{ID: "deepseek-v3.2-chat", DisplayName: "DeepSeek-V3.2", Description: "DeepSeek V3.2", Created: 1764576000}, {ID: "deepseek-v3.2-chat", DisplayName: "DeepSeek-V3.2", Description: "DeepSeek V3.2", Created: 1764576000},
{ID: "deepseek-v3.2", DisplayName: "DeepSeek-V3.2-Exp", Description: "DeepSeek V3.2 experimental", Created: 1759104000}, {ID: "deepseek-v3.2", DisplayName: "DeepSeek-V3.2-Exp", Description: "DeepSeek V3.2 experimental", Created: 1759104000},
{ID: "deepseek-v3.1", DisplayName: "DeepSeek-V3.1-Terminus", Description: "DeepSeek V3.1 Terminus", Created: 1756339200}, {ID: "deepseek-v3.1", DisplayName: "DeepSeek-V3.1-Terminus", Description: "DeepSeek V3.1 Terminus", Created: 1756339200},
{ID: "deepseek-r1", DisplayName: "DeepSeek-R1", Description: "DeepSeek reasoning model R1", Created: 1737331200}, {ID: "deepseek-r1", DisplayName: "DeepSeek-R1", Description: "DeepSeek reasoning model R1", Created: 1737331200, Thinking: &ThinkingSupport{Levels: []string{"low", "medium", "high"}}},
{ID: "deepseek-v3", DisplayName: "DeepSeek-V3-671B", Description: "DeepSeek V3 671B", Created: 1734307200}, {ID: "deepseek-v3", DisplayName: "DeepSeek-V3-671B", Description: "DeepSeek V3 671B", Created: 1734307200},
{ID: "qwen3-32b", DisplayName: "Qwen3-32B", Description: "Qwen3 32B", Created: 1747094400}, {ID: "qwen3-32b", DisplayName: "Qwen3-32B", Description: "Qwen3 32B", Created: 1747094400},
{ID: "qwen3-235b-a22b-thinking-2507", DisplayName: "Qwen3-235B-A22B-Thinking", Description: "Qwen3 235B A22B Thinking (2507)", Created: 1753401600}, {ID: "qwen3-235b-a22b-thinking-2507", DisplayName: "Qwen3-235B-A22B-Thinking", Description: "Qwen3 235B A22B Thinking (2507)", Created: 1753401600, Thinking: &ThinkingSupport{Levels: []string{"low", "medium", "high"}}},
{ID: "qwen3-235b-a22b-instruct", DisplayName: "Qwen3-235B-A22B-Instruct", Description: "Qwen3 235B A22B Instruct", Created: 1753401600}, {ID: "qwen3-235b-a22b-instruct", DisplayName: "Qwen3-235B-A22B-Instruct", Description: "Qwen3 235B A22B Instruct", Created: 1753401600},
{ID: "qwen3-235b", DisplayName: "Qwen3-235B-A22B", Description: "Qwen3 235B A22B", Created: 1753401600}, {ID: "qwen3-235b", DisplayName: "Qwen3-235B-A22B", Description: "Qwen3 235B A22B", Created: 1753401600},
{ID: "minimax-m2", DisplayName: "MiniMax-M2", Description: "MiniMax M2", Created: 1758672000}, {ID: "minimax-m2", DisplayName: "MiniMax-M2", Description: "MiniMax M2", Created: 1758672000, Thinking: &ThinkingSupport{Levels: []string{"low", "medium", "high"}}},
} }
models := make([]*ModelInfo, 0, len(entries)) models := make([]*ModelInfo, 0, len(entries))
for _, entry := range entries { for _, entry := range entries {
@@ -981,6 +655,7 @@ func GetIFlowModels() []*ModelInfo {
Type: "iflow", Type: "iflow",
DisplayName: entry.DisplayName, DisplayName: entry.DisplayName,
Description: entry.Description, Description: entry.Description,
Thinking: entry.Thinking,
}) })
} }
return models return models

3
internal/registry/model_registry.go
View File

@@ -63,6 +63,9 @@ type ThinkingSupport struct {
ZeroAllowed bool `json:"zero_allowed,omitempty"` ZeroAllowed bool `json:"zero_allowed,omitempty"`
// DynamicAllowed indicates whether -1 is a valid value (dynamic thinking budget). // DynamicAllowed indicates whether -1 is a valid value (dynamic thinking budget).
DynamicAllowed bool `json:"dynamic_allowed,omitempty"` DynamicAllowed bool `json:"dynamic_allowed,omitempty"`
// Levels defines discrete reasoning effort levels (e.g., "low", "medium", "high").
// When set, the model uses level-based reasoning instead of token budgets.
Levels []string `json:"levels,omitempty"`
} }
// ModelRegistration tracks a model's availability // ModelRegistration tracks a model's availability

2
internal/runtime/executor/antigravity_executor.go
View File

@@ -39,7 +39,7 @@ const (
defaultAntigravityAgent = "antigravity/1.11.5 windows/amd64" defaultAntigravityAgent = "antigravity/1.11.5 windows/amd64"
antigravityAuthType = "antigravity" antigravityAuthType = "antigravity"
refreshSkew = 3000 * time.Second refreshSkew = 3000 * time.Second
streamScannerBuffer int = 20_971_520 streamScannerBuffer int = 52_428_800 // 50MB
) )
var ( var (

129
internal/runtime/executor/claude_executor.go
View File

@@ -54,15 +54,22 @@ func (e *ClaudeExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
// Use streaming translation to preserve function calling, except for claude. // Use streaming translation to preserve function calling, except for claude.
stream := from != to stream := from != to
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), stream) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), stream)
modelForUpstream := req.Model upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
if modelOverride := e.resolveUpstreamModel(req.Model, auth); modelOverride != "" { if upstreamModel == "" {
body, _ = sjson.SetBytes(body, "model", modelOverride) upstreamModel = req.Model
modelForUpstream = modelOverride
} }
// Inject thinking config based on model suffix for thinking variants if modelOverride := e.resolveUpstreamModel(upstreamModel, auth); modelOverride != "" {
body = e.injectThinkingConfig(req.Model, body) upstreamModel = modelOverride
} else if !strings.EqualFold(upstreamModel, req.Model) {
if modelOverride := e.resolveUpstreamModel(req.Model, auth); modelOverride != "" {
upstreamModel = modelOverride
}
}
body, _ = sjson.SetBytes(body, "model", upstreamModel)
// Inject thinking config based on model metadata for thinking variants
body = e.injectThinkingConfig(req.Model, req.Metadata, body)
if !strings.HasPrefix(modelForUpstream, "claude-3-5-haiku") { if !strings.HasPrefix(upstreamModel, "claude-3-5-haiku") {
body = checkSystemInstructions(body) body = checkSystemInstructions(body)
} }
body = applyPayloadConfig(e.cfg, req.Model, body) body = applyPayloadConfig(e.cfg, req.Model, body)
@@ -161,11 +168,20 @@ func (e *ClaudeExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
from := opts.SourceFormat from := opts.SourceFormat
to := sdktranslator.FromString("claude") to := sdktranslator.FromString("claude")
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
if modelOverride := e.resolveUpstreamModel(req.Model, auth); modelOverride != "" { upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
body, _ = sjson.SetBytes(body, "model", modelOverride) if upstreamModel == "" {
upstreamModel = req.Model
} }
// Inject thinking config based on model suffix for thinking variants if modelOverride := e.resolveUpstreamModel(upstreamModel, auth); modelOverride != "" {
body = e.injectThinkingConfig(req.Model, body) upstreamModel = modelOverride
} else if !strings.EqualFold(upstreamModel, req.Model) {
if modelOverride := e.resolveUpstreamModel(req.Model, auth); modelOverride != "" {
upstreamModel = modelOverride
}
}
body, _ = sjson.SetBytes(body, "model", upstreamModel)
// Inject thinking config based on model metadata for thinking variants
body = e.injectThinkingConfig(req.Model, req.Metadata, body)
body = checkSystemInstructions(body) body = checkSystemInstructions(body)
body = applyPayloadConfig(e.cfg, req.Model, body) body = applyPayloadConfig(e.cfg, req.Model, body)
@@ -238,7 +254,7 @@ func (e *ClaudeExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
// If from == to (Claude → Claude), directly forward the SSE stream without translation // If from == to (Claude → Claude), directly forward the SSE stream without translation
if from == to { if from == to {
scanner := bufio.NewScanner(decodedBody) scanner := bufio.NewScanner(decodedBody)
scanner.Buffer(nil, 20_971_520) scanner.Buffer(nil, 52_428_800) // 50MB
for scanner.Scan() { for scanner.Scan() {
line := scanner.Bytes() line := scanner.Bytes()
appendAPIResponseChunk(ctx, e.cfg, line) appendAPIResponseChunk(ctx, e.cfg, line)
@@ -261,7 +277,7 @@ func (e *ClaudeExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
// For other formats, use translation // For other formats, use translation
scanner := bufio.NewScanner(decodedBody) scanner := bufio.NewScanner(decodedBody)
scanner.Buffer(nil, 20_971_520) scanner.Buffer(nil, 52_428_800) // 50MB
var param any var param any
for scanner.Scan() { for scanner.Scan() {
line := scanner.Bytes() line := scanner.Bytes()
@@ -295,13 +311,20 @@ func (e *ClaudeExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Aut
// Use streaming translation to preserve function calling, except for claude. // Use streaming translation to preserve function calling, except for claude.
stream := from != to stream := from != to
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), stream) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), stream)
modelForUpstream := req.Model upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
if modelOverride := e.resolveUpstreamModel(req.Model, auth); modelOverride != "" { if upstreamModel == "" {
body, _ = sjson.SetBytes(body, "model", modelOverride) upstreamModel = req.Model
modelForUpstream = modelOverride
} }
if modelOverride := e.resolveUpstreamModel(upstreamModel, auth); modelOverride != "" {
upstreamModel = modelOverride
} else if !strings.EqualFold(upstreamModel, req.Model) {
if modelOverride := e.resolveUpstreamModel(req.Model, auth); modelOverride != "" {
upstreamModel = modelOverride
}
}
body, _ = sjson.SetBytes(body, "model", upstreamModel)
if !strings.HasPrefix(modelForUpstream, "claude-3-5-haiku") { if !strings.HasPrefix(upstreamModel, "claude-3-5-haiku") {
body = checkSystemInstructions(body) body = checkSystemInstructions(body)
} }
@@ -427,31 +450,15 @@ func extractAndRemoveBetas(body []byte) ([]string, []byte) {
return betas, body return betas, body
} }
// injectThinkingConfig adds thinking configuration based on model name suffix // injectThinkingConfig adds thinking configuration based on metadata using the unified flow.
func (e *ClaudeExecutor) injectThinkingConfig(modelName string, body []byte) []byte { // It uses util.ResolveClaudeThinkingConfig which internally calls ResolveThinkingConfigFromMetadata
// Only inject if thinking config is not already present // and NormalizeThinkingBudget, ensuring consistency with other executors like Gemini.
if gjson.GetBytes(body, "thinking").Exists() { func (e *ClaudeExecutor) injectThinkingConfig(modelName string, metadata map[string]any, body []byte) []byte {
budget, ok := util.ResolveClaudeThinkingConfig(modelName, metadata)
if !ok {
return body return body
} }
return util.ApplyClaudeThinkingConfig(body, budget)
var budgetTokens int
switch {
case strings.HasSuffix(modelName, "-thinking-low"):
budgetTokens = 1024
case strings.HasSuffix(modelName, "-thinking-medium"):
budgetTokens = 8192
case strings.HasSuffix(modelName, "-thinking-high"):
budgetTokens = 24576
case strings.HasSuffix(modelName, "-thinking"):
// Default thinking without suffix uses medium budget
budgetTokens = 8192
default:
return body
}
body, _ = sjson.SetBytes(body, "thinking.type", "enabled")
body, _ = sjson.SetBytes(body, "thinking.budget_tokens", budgetTokens)
return body
} }
// ensureMaxTokensForThinking ensures max_tokens > thinking.budget_tokens when thinking is enabled. // ensureMaxTokensForThinking ensures max_tokens > thinking.budget_tokens when thinking is enabled.
@@ -491,35 +498,45 @@ func ensureMaxTokensForThinking(modelName string, body []byte) []byte {
} }
func (e *ClaudeExecutor) resolveUpstreamModel(alias string, auth *cliproxyauth.Auth) string { func (e *ClaudeExecutor) resolveUpstreamModel(alias string, auth *cliproxyauth.Auth) string {
if alias == "" { trimmed := strings.TrimSpace(alias)
if trimmed == "" {
return "" return ""
} }
// Hardcoded mappings for thinking models to actual Claude model names
switch alias {
case "claude-opus-4-5-thinking", "claude-opus-4-5-thinking-low", "claude-opus-4-5-thinking-medium", "claude-opus-4-5-thinking-high":
return "claude-opus-4-5-20251101"
case "claude-sonnet-4-5-thinking":
return "claude-sonnet-4-5-20250929"
}
entry := e.resolveClaudeConfig(auth) entry := e.resolveClaudeConfig(auth)
if entry == nil { if entry == nil {
return "" return ""
} }
normalizedModel, metadata := util.NormalizeThinkingModel(trimmed)
// Candidate names to match against configured aliases/names.
candidates := []string{strings.TrimSpace(normalizedModel)}
if !strings.EqualFold(normalizedModel, trimmed) {
candidates = append(candidates, trimmed)
}
if original := util.ResolveOriginalModel(normalizedModel, metadata); original != "" && !strings.EqualFold(original, normalizedModel) {
candidates = append(candidates, original)
}
for i := range entry.Models { for i := range entry.Models {
model := entry.Models[i] model := entry.Models[i]
name := strings.TrimSpace(model.Name) name := strings.TrimSpace(model.Name)
modelAlias := strings.TrimSpace(model.Alias) modelAlias := strings.TrimSpace(model.Alias)
if modelAlias != "" {
if strings.EqualFold(modelAlias, alias) { for _, candidate := range candidates {
if candidate == "" {
continue
}
if modelAlias != "" && strings.EqualFold(modelAlias, candidate) {
if name != "" { if name != "" {
return name return name
} }
return alias return candidate
}
if name != "" && strings.EqualFold(name, candidate) {
return name
} }
continue
}
if name != "" && strings.EqualFold(name, alias) {
return name
} }
} }
return "" return ""

106
internal/runtime/executor/codex_executor.go
View File

@@ -49,14 +49,18 @@ func (e *CodexExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth) reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
defer reporter.trackFailure(ctx, &err) defer reporter.trackFailure(ctx, &err)
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
from := opts.SourceFormat from := opts.SourceFormat
to := sdktranslator.FromString("codex") to := sdktranslator.FromString("codex")
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
body = applyReasoningEffortMetadata(body, req.Metadata, req.Model, "reasoning.effort")
body = e.setReasoningEffortByAlias(req.Model, body) body = normalizeThinkingConfig(body, upstreamModel)
if errValidate := validateThinkingConfig(body, upstreamModel); errValidate != nil {
return resp, errValidate
}
body = applyPayloadConfig(e.cfg, req.Model, body) body = applyPayloadConfig(e.cfg, req.Model, body)
body, _ = sjson.SetBytes(body, "model", upstreamModel)
body, _ = sjson.SetBytes(body, "stream", true) body, _ = sjson.SetBytes(body, "stream", true)
body, _ = sjson.DeleteBytes(body, "previous_response_id") body, _ = sjson.DeleteBytes(body, "previous_response_id")
@@ -142,13 +146,20 @@ func (e *CodexExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth) reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
defer reporter.trackFailure(ctx, &err) defer reporter.trackFailure(ctx, &err)
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
from := opts.SourceFormat from := opts.SourceFormat
to := sdktranslator.FromString("codex") to := sdktranslator.FromString("codex")
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
body = e.setReasoningEffortByAlias(req.Model, body) body = applyReasoningEffortMetadata(body, req.Metadata, req.Model, "reasoning.effort")
body = normalizeThinkingConfig(body, upstreamModel)
if errValidate := validateThinkingConfig(body, upstreamModel); errValidate != nil {
return nil, errValidate
}
body = applyPayloadConfig(e.cfg, req.Model, body) body = applyPayloadConfig(e.cfg, req.Model, body)
body, _ = sjson.DeleteBytes(body, "previous_response_id") body, _ = sjson.DeleteBytes(body, "previous_response_id")
body, _ = sjson.SetBytes(body, "model", upstreamModel)
url := strings.TrimSuffix(baseURL, "/") + "/responses" url := strings.TrimSuffix(baseURL, "/") + "/responses"
httpReq, err := e.cacheHelper(ctx, from, url, req, body) httpReq, err := e.cacheHelper(ctx, from, url, req, body)
@@ -205,7 +216,7 @@ func (e *CodexExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
} }
}() }()
scanner := bufio.NewScanner(httpResp.Body) scanner := bufio.NewScanner(httpResp.Body)
scanner.Buffer(nil, 20_971_520) scanner.Buffer(nil, 52_428_800) // 50MB
var param any var param any
for scanner.Scan() { for scanner.Scan() {
line := scanner.Bytes() line := scanner.Bytes()
@@ -235,14 +246,16 @@ func (e *CodexExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
} }
func (e *CodexExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options) (cliproxyexecutor.Response, error) { func (e *CodexExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options) (cliproxyexecutor.Response, error) {
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
from := opts.SourceFormat from := opts.SourceFormat
to := sdktranslator.FromString("codex") to := sdktranslator.FromString("codex")
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
modelForCounting := req.Model modelForCounting := req.Model
body = e.setReasoningEffortByAlias(req.Model, body) body = applyReasoningEffortMetadata(body, req.Metadata, req.Model, "reasoning.effort")
body, _ = sjson.SetBytes(body, "model", upstreamModel)
body, _ = sjson.DeleteBytes(body, "previous_response_id") body, _ = sjson.DeleteBytes(body, "previous_response_id")
body, _ = sjson.SetBytes(body, "stream", false) body, _ = sjson.SetBytes(body, "stream", false)
@@ -261,83 +274,6 @@ func (e *CodexExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Auth
return cliproxyexecutor.Response{Payload: []byte(translated)}, nil return cliproxyexecutor.Response{Payload: []byte(translated)}, nil
} }
func (e *CodexExecutor) setReasoningEffortByAlias(modelName string, payload []byte) []byte {
if util.InArray([]string{"gpt-5", "gpt-5-minimal", "gpt-5-low", "gpt-5-medium", "gpt-5-high"}, modelName) {
payload, _ = sjson.SetBytes(payload, "model", "gpt-5")
switch modelName {
case "gpt-5-minimal":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "minimal")
case "gpt-5-low":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "low")
case "gpt-5-medium":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "medium")
case "gpt-5-high":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "high")
}
} else if util.InArray([]string{"gpt-5-codex", "gpt-5-codex-low", "gpt-5-codex-medium", "gpt-5-codex-high"}, modelName) {
payload, _ = sjson.SetBytes(payload, "model", "gpt-5-codex")
switch modelName {
case "gpt-5-codex-low":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "low")
case "gpt-5-codex-medium":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "medium")
case "gpt-5-codex-high":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "high")
}
} else if util.InArray([]string{"gpt-5-codex-mini", "gpt-5-codex-mini-medium", "gpt-5-codex-mini-high"}, modelName) {
payload, _ = sjson.SetBytes(payload, "model", "gpt-5-codex-mini")
switch modelName {
case "gpt-5-codex-mini-medium":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "medium")
case "gpt-5-codex-mini-high":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "high")
}
} else if util.InArray([]string{"gpt-5.1", "gpt-5.1-none", "gpt-5.1-low", "gpt-5.1-medium", "gpt-5.1-high"}, modelName) {
payload, _ = sjson.SetBytes(payload, "model", "gpt-5.1")
switch modelName {
case "gpt-5.1-none":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "none")
case "gpt-5.1-low":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "low")
case "gpt-5.1-medium":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "medium")
case "gpt-5.1-high":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "high")
}
} else if util.InArray([]string{"gpt-5.1-codex", "gpt-5.1-codex-low", "gpt-5.1-codex-medium", "gpt-5.1-codex-high"}, modelName) {
payload, _ = sjson.SetBytes(payload, "model", "gpt-5.1-codex")
switch modelName {
case "gpt-5.1-codex-low":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "low")
case "gpt-5.1-codex-medium":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "medium")
case "gpt-5.1-codex-high":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "high")
}
} else if util.InArray([]string{"gpt-5.1-codex-mini", "gpt-5.1-codex-mini-medium", "gpt-5.1-codex-mini-high"}, modelName) {
payload, _ = sjson.SetBytes(payload, "model", "gpt-5.1-codex-mini")
switch modelName {
case "gpt-5.1-codex-mini-medium":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "medium")
case "gpt-5.1-codex-mini-high":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "high")
}
} else if util.InArray([]string{"gpt-5.1-codex-max", "gpt-5.1-codex-max-low", "gpt-5.1-codex-max-medium", "gpt-5.1-codex-max-high", "gpt-5.1-codex-max-xhigh"}, modelName) {
payload, _ = sjson.SetBytes(payload, "model", "gpt-5.1-codex-max")
switch modelName {
case "gpt-5.1-codex-max-low":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "low")
case "gpt-5.1-codex-max-medium":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "medium")
case "gpt-5.1-codex-max-high":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "high")
case "gpt-5.1-codex-max-xhigh":
payload, _ = sjson.SetBytes(payload, "reasoning.effort", "xhigh")
}
}
return payload
}
func tokenizerForCodexModel(model string) (tokenizer.Codec, error) { func tokenizerForCodexModel(model string) (tokenizer.Codec, error) {
sanitized := strings.ToLower(strings.TrimSpace(model)) sanitized := strings.ToLower(strings.TrimSpace(model))
switch { switch {

2
internal/runtime/executor/gemini_cli_executor.go
View File

@@ -309,7 +309,7 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
}() }()
if opts.Alt == "" { if opts.Alt == "" {
scanner := bufio.NewScanner(resp.Body) scanner := bufio.NewScanner(resp.Body)
scanner.Buffer(nil, 20_971_520) scanner.Buffer(nil, 52_428_800) // 50MB
var param any var param any
for scanner.Scan() { for scanner.Scan() {
line := scanner.Bytes() line := scanner.Bytes()

12
internal/runtime/executor/gemini_executor.go
View File

@@ -75,6 +75,8 @@ func (e *GeminiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth) reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
defer reporter.trackFailure(ctx, &err) defer reporter.trackFailure(ctx, &err)
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
// Official Gemini API via API key or OAuth bearer // Official Gemini API via API key or OAuth bearer
from := opts.SourceFormat from := opts.SourceFormat
to := sdktranslator.FromString("gemini") to := sdktranslator.FromString("gemini")
@@ -85,6 +87,7 @@ func (e *GeminiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
body = util.StripThinkingConfigIfUnsupported(req.Model, body) body = util.StripThinkingConfigIfUnsupported(req.Model, body)
body = fixGeminiImageAspectRatio(req.Model, body) body = fixGeminiImageAspectRatio(req.Model, body)
body = applyPayloadConfig(e.cfg, req.Model, body) body = applyPayloadConfig(e.cfg, req.Model, body)
body, _ = sjson.SetBytes(body, "model", upstreamModel)
action := "generateContent" action := "generateContent"
if req.Metadata != nil { if req.Metadata != nil {
@@ -93,7 +96,7 @@ func (e *GeminiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
} }
} }
baseURL := resolveGeminiBaseURL(auth) baseURL := resolveGeminiBaseURL(auth)
url := fmt.Sprintf("%s/%s/models/%s:%s", baseURL, glAPIVersion, req.Model, action) url := fmt.Sprintf("%s/%s/models/%s:%s", baseURL, glAPIVersion, upstreamModel, action)
if opts.Alt != "" && action != "countTokens" { if opts.Alt != "" && action != "countTokens" {
url = url + fmt.Sprintf("?$alt=%s", opts.Alt) url = url + fmt.Sprintf("?$alt=%s", opts.Alt)
} }
@@ -167,6 +170,8 @@ func (e *GeminiExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth) reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
defer reporter.trackFailure(ctx, &err) defer reporter.trackFailure(ctx, &err)
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
from := opts.SourceFormat from := opts.SourceFormat
to := sdktranslator.FromString("gemini") to := sdktranslator.FromString("gemini")
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
@@ -176,9 +181,10 @@ func (e *GeminiExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
body = util.StripThinkingConfigIfUnsupported(req.Model, body) body = util.StripThinkingConfigIfUnsupported(req.Model, body)
body = fixGeminiImageAspectRatio(req.Model, body) body = fixGeminiImageAspectRatio(req.Model, body)
body = applyPayloadConfig(e.cfg, req.Model, body) body = applyPayloadConfig(e.cfg, req.Model, body)
body, _ = sjson.SetBytes(body, "model", upstreamModel)
baseURL := resolveGeminiBaseURL(auth) baseURL := resolveGeminiBaseURL(auth)
url := fmt.Sprintf("%s/%s/models/%s:%s", baseURL, glAPIVersion, req.Model, "streamGenerateContent") url := fmt.Sprintf("%s/%s/models/%s:%s", baseURL, glAPIVersion, upstreamModel, "streamGenerateContent")
if opts.Alt == "" { if opts.Alt == "" {
url = url + "?alt=sse" url = url + "?alt=sse"
} else { } else {
@@ -243,7 +249,7 @@ func (e *GeminiExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
} }
}() }()
scanner := bufio.NewScanner(httpResp.Body) scanner := bufio.NewScanner(httpResp.Body)
scanner.Buffer(nil, 20_971_520) scanner.Buffer(nil, 52_428_800) // 50MB
var param any var param any
for scanner.Scan() { for scanner.Scan() {
line := scanner.Bytes() line := scanner.Bytes()

44
internal/runtime/executor/gemini_vertex_executor.go
View File

@@ -105,10 +105,12 @@ func (e *GeminiVertexExecutor) CountTokens(ctx context.Context, auth *cliproxyau
// countTokensWithServiceAccount handles token counting using service account credentials. // countTokensWithServiceAccount handles token counting using service account credentials.
func (e *GeminiVertexExecutor) countTokensWithServiceAccount(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options, projectID, location string, saJSON []byte) (cliproxyexecutor.Response, error) { func (e *GeminiVertexExecutor) countTokensWithServiceAccount(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options, projectID, location string, saJSON []byte) (cliproxyexecutor.Response, error) {
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
from := opts.SourceFormat from := opts.SourceFormat
to := sdktranslator.FromString("gemini") to := sdktranslator.FromString("gemini")
translatedReq := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false) translatedReq := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
if budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(req.Metadata); ok && util.ModelSupportsThinking(req.Model) { if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
if budgetOverride != nil { if budgetOverride != nil {
norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride) norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
budgetOverride = &norm budgetOverride = &norm
@@ -117,13 +119,14 @@ func (e *GeminiVertexExecutor) countTokensWithServiceAccount(ctx context.Context
} }
translatedReq = util.StripThinkingConfigIfUnsupported(req.Model, translatedReq) translatedReq = util.StripThinkingConfigIfUnsupported(req.Model, translatedReq)
translatedReq = fixGeminiImageAspectRatio(req.Model, translatedReq) translatedReq = fixGeminiImageAspectRatio(req.Model, translatedReq)
translatedReq, _ = sjson.SetBytes(translatedReq, "model", upstreamModel)
respCtx := context.WithValue(ctx, "alt", opts.Alt) respCtx := context.WithValue(ctx, "alt", opts.Alt)
translatedReq, _ = sjson.DeleteBytes(translatedReq, "tools") translatedReq, _ = sjson.DeleteBytes(translatedReq, "tools")
translatedReq, _ = sjson.DeleteBytes(translatedReq, "generationConfig") translatedReq, _ = sjson.DeleteBytes(translatedReq, "generationConfig")
translatedReq, _ = sjson.DeleteBytes(translatedReq, "safetySettings") translatedReq, _ = sjson.DeleteBytes(translatedReq, "safetySettings")
baseURL := vertexBaseURL(location) baseURL := vertexBaseURL(location)
url := fmt.Sprintf("%s/%s/projects/%s/locations/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, projectID, location, req.Model, "countTokens") url := fmt.Sprintf("%s/%s/projects/%s/locations/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, projectID, location, upstreamModel, "countTokens")
httpReq, errNewReq := http.NewRequestWithContext(respCtx, http.MethodPost, url, bytes.NewReader(translatedReq)) httpReq, errNewReq := http.NewRequestWithContext(respCtx, http.MethodPost, url, bytes.NewReader(translatedReq))
if errNewReq != nil { if errNewReq != nil {
@@ -191,10 +194,12 @@ func (e *GeminiVertexExecutor) countTokensWithServiceAccount(ctx context.Context
// countTokensWithAPIKey handles token counting using API key credentials. // countTokensWithAPIKey handles token counting using API key credentials.
func (e *GeminiVertexExecutor) countTokensWithAPIKey(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options, apiKey, baseURL string) (cliproxyexecutor.Response, error) { func (e *GeminiVertexExecutor) countTokensWithAPIKey(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options, apiKey, baseURL string) (cliproxyexecutor.Response, error) {
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
from := opts.SourceFormat from := opts.SourceFormat
to := sdktranslator.FromString("gemini") to := sdktranslator.FromString("gemini")
translatedReq := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false) translatedReq := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
if budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(req.Metadata); ok && util.ModelSupportsThinking(req.Model) { if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
if budgetOverride != nil { if budgetOverride != nil {
norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride) norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
budgetOverride = &norm budgetOverride = &norm
@@ -203,6 +208,7 @@ func (e *GeminiVertexExecutor) countTokensWithAPIKey(ctx context.Context, auth *
} }
translatedReq = util.StripThinkingConfigIfUnsupported(req.Model, translatedReq) translatedReq = util.StripThinkingConfigIfUnsupported(req.Model, translatedReq)
translatedReq = fixGeminiImageAspectRatio(req.Model, translatedReq) translatedReq = fixGeminiImageAspectRatio(req.Model, translatedReq)
translatedReq, _ = sjson.SetBytes(translatedReq, "model", upstreamModel)
respCtx := context.WithValue(ctx, "alt", opts.Alt) respCtx := context.WithValue(ctx, "alt", opts.Alt)
translatedReq, _ = sjson.DeleteBytes(translatedReq, "tools") translatedReq, _ = sjson.DeleteBytes(translatedReq, "tools")
translatedReq, _ = sjson.DeleteBytes(translatedReq, "generationConfig") translatedReq, _ = sjson.DeleteBytes(translatedReq, "generationConfig")
@@ -286,10 +292,12 @@ func (e *GeminiVertexExecutor) executeWithServiceAccount(ctx context.Context, au
reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth) reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
defer reporter.trackFailure(ctx, &err) defer reporter.trackFailure(ctx, &err)
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
from := opts.SourceFormat from := opts.SourceFormat
to := sdktranslator.FromString("gemini") to := sdktranslator.FromString("gemini")
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
if budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(req.Metadata); ok && util.ModelSupportsThinking(req.Model) { if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
if budgetOverride != nil { if budgetOverride != nil {
norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride) norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
budgetOverride = &norm budgetOverride = &norm
@@ -301,6 +309,7 @@ func (e *GeminiVertexExecutor) executeWithServiceAccount(ctx context.Context, au
body = util.StripThinkingConfigIfUnsupported(req.Model, body) body = util.StripThinkingConfigIfUnsupported(req.Model, body)
body = fixGeminiImageAspectRatio(req.Model, body) body = fixGeminiImageAspectRatio(req.Model, body)
body = applyPayloadConfig(e.cfg, req.Model, body) body = applyPayloadConfig(e.cfg, req.Model, body)
body, _ = sjson.SetBytes(body, "model", upstreamModel)
action := "generateContent" action := "generateContent"
if req.Metadata != nil { if req.Metadata != nil {
@@ -309,7 +318,7 @@ func (e *GeminiVertexExecutor) executeWithServiceAccount(ctx context.Context, au
} }
} }
baseURL := vertexBaseURL(location) baseURL := vertexBaseURL(location)
url := fmt.Sprintf("%s/%s/projects/%s/locations/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, projectID, location, req.Model, action) url := fmt.Sprintf("%s/%s/projects/%s/locations/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, projectID, location, upstreamModel, action)
if opts.Alt != "" && action != "countTokens" { if opts.Alt != "" && action != "countTokens" {
url = url + fmt.Sprintf("?$alt=%s", opts.Alt) url = url + fmt.Sprintf("?$alt=%s", opts.Alt)
} }
@@ -383,10 +392,12 @@ func (e *GeminiVertexExecutor) executeWithAPIKey(ctx context.Context, auth *clip
reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth) reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
defer reporter.trackFailure(ctx, &err) defer reporter.trackFailure(ctx, &err)
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
from := opts.SourceFormat from := opts.SourceFormat
to := sdktranslator.FromString("gemini") to := sdktranslator.FromString("gemini")
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
if budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(req.Metadata); ok && util.ModelSupportsThinking(req.Model) { if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
if budgetOverride != nil { if budgetOverride != nil {
norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride) norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
budgetOverride = &norm budgetOverride = &norm
@@ -398,6 +409,7 @@ func (e *GeminiVertexExecutor) executeWithAPIKey(ctx context.Context, auth *clip
body = util.StripThinkingConfigIfUnsupported(req.Model, body) body = util.StripThinkingConfigIfUnsupported(req.Model, body)
body = fixGeminiImageAspectRatio(req.Model, body) body = fixGeminiImageAspectRatio(req.Model, body)
body = applyPayloadConfig(e.cfg, req.Model, body) body = applyPayloadConfig(e.cfg, req.Model, body)
body, _ = sjson.SetBytes(body, "model", upstreamModel)
action := "generateContent" action := "generateContent"
if req.Metadata != nil { if req.Metadata != nil {
@@ -410,7 +422,7 @@ func (e *GeminiVertexExecutor) executeWithAPIKey(ctx context.Context, auth *clip
if baseURL == "" { if baseURL == "" {
baseURL = "https://generativelanguage.googleapis.com" baseURL = "https://generativelanguage.googleapis.com"
} }
url := fmt.Sprintf("%s/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, req.Model, action) url := fmt.Sprintf("%s/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, upstreamModel, action)
if opts.Alt != "" && action != "countTokens" { if opts.Alt != "" && action != "countTokens" {
url = url + fmt.Sprintf("?$alt=%s", opts.Alt) url = url + fmt.Sprintf("?$alt=%s", opts.Alt)
} }
@@ -481,10 +493,12 @@ func (e *GeminiVertexExecutor) executeStreamWithServiceAccount(ctx context.Conte
reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth) reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
defer reporter.trackFailure(ctx, &err) defer reporter.trackFailure(ctx, &err)
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
from := opts.SourceFormat from := opts.SourceFormat
to := sdktranslator.FromString("gemini") to := sdktranslator.FromString("gemini")
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
if budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(req.Metadata); ok && util.ModelSupportsThinking(req.Model) { if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
if budgetOverride != nil { if budgetOverride != nil {
norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride) norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
budgetOverride = &norm budgetOverride = &norm
@@ -496,9 +510,10 @@ func (e *GeminiVertexExecutor) executeStreamWithServiceAccount(ctx context.Conte
body = util.StripThinkingConfigIfUnsupported(req.Model, body) body = util.StripThinkingConfigIfUnsupported(req.Model, body)
body = fixGeminiImageAspectRatio(req.Model, body) body = fixGeminiImageAspectRatio(req.Model, body)
body = applyPayloadConfig(e.cfg, req.Model, body) body = applyPayloadConfig(e.cfg, req.Model, body)
body, _ = sjson.SetBytes(body, "model", upstreamModel)
baseURL := vertexBaseURL(location) baseURL := vertexBaseURL(location)
url := fmt.Sprintf("%s/%s/projects/%s/locations/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, projectID, location, req.Model, "streamGenerateContent") url := fmt.Sprintf("%s/%s/projects/%s/locations/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, projectID, location, upstreamModel, "streamGenerateContent")
if opts.Alt == "" { if opts.Alt == "" {
url = url + "?alt=sse" url = url + "?alt=sse"
} else { } else {
@@ -564,7 +579,7 @@ func (e *GeminiVertexExecutor) executeStreamWithServiceAccount(ctx context.Conte
} }
}() }()
scanner := bufio.NewScanner(httpResp.Body) scanner := bufio.NewScanner(httpResp.Body)
scanner.Buffer(nil, 20_971_520) scanner.Buffer(nil, 52_428_800) // 50MB
var param any var param any
for scanner.Scan() { for scanner.Scan() {
line := scanner.Bytes() line := scanner.Bytes()
@@ -595,10 +610,12 @@ func (e *GeminiVertexExecutor) executeStreamWithAPIKey(ctx context.Context, auth
reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth) reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
defer reporter.trackFailure(ctx, &err) defer reporter.trackFailure(ctx, &err)
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
from := opts.SourceFormat from := opts.SourceFormat
to := sdktranslator.FromString("gemini") to := sdktranslator.FromString("gemini")
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
if budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(req.Metadata); ok && util.ModelSupportsThinking(req.Model) { if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
if budgetOverride != nil { if budgetOverride != nil {
norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride) norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
budgetOverride = &norm budgetOverride = &norm
@@ -610,12 +627,13 @@ func (e *GeminiVertexExecutor) executeStreamWithAPIKey(ctx context.Context, auth
body = util.StripThinkingConfigIfUnsupported(req.Model, body) body = util.StripThinkingConfigIfUnsupported(req.Model, body)
body = fixGeminiImageAspectRatio(req.Model, body) body = fixGeminiImageAspectRatio(req.Model, body)
body = applyPayloadConfig(e.cfg, req.Model, body) body = applyPayloadConfig(e.cfg, req.Model, body)
body, _ = sjson.SetBytes(body, "model", upstreamModel)
// For API key auth, use simpler URL format without project/location // For API key auth, use simpler URL format without project/location
if baseURL == "" { if baseURL == "" {
baseURL = "https://generativelanguage.googleapis.com" baseURL = "https://generativelanguage.googleapis.com"
} }
url := fmt.Sprintf("%s/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, req.Model, "streamGenerateContent") url := fmt.Sprintf("%s/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, upstreamModel, "streamGenerateContent")
if opts.Alt == "" { if opts.Alt == "" {
url = url + "?alt=sse" url = url + "?alt=sse"
} else { } else {
@@ -678,7 +696,7 @@ func (e *GeminiVertexExecutor) executeStreamWithAPIKey(ctx context.Context, auth
} }
}() }()
scanner := bufio.NewScanner(httpResp.Body) scanner := bufio.NewScanner(httpResp.Body)
scanner.Buffer(nil, 20_971_520) scanner.Buffer(nil, 52_428_800) // 50MB
var param any var param any
for scanner.Scan() { for scanner.Scan() {
line := scanner.Bytes() line := scanner.Bytes()

20
internal/runtime/executor/iflow_executor.go
View File

@@ -57,6 +57,15 @@ func (e *IFlowExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
from := opts.SourceFormat from := opts.SourceFormat
to := sdktranslator.FromString("openai") to := sdktranslator.FromString("openai")
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
body = applyReasoningEffortMetadata(body, req.Metadata, req.Model, "reasoning_effort")
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
if upstreamModel != "" {
body, _ = sjson.SetBytes(body, "model", upstreamModel)
}
body = normalizeThinkingConfig(body, upstreamModel)
if errValidate := validateThinkingConfig(body, upstreamModel); errValidate != nil {
return resp, errValidate
}
body = applyPayloadConfig(e.cfg, req.Model, body) body = applyPayloadConfig(e.cfg, req.Model, body)
endpoint := strings.TrimSuffix(baseURL, "/") + iflowDefaultEndpoint endpoint := strings.TrimSuffix(baseURL, "/") + iflowDefaultEndpoint
@@ -139,6 +148,15 @@ func (e *IFlowExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
to := sdktranslator.FromString("openai") to := sdktranslator.FromString("openai")
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
body = applyReasoningEffortMetadata(body, req.Metadata, req.Model, "reasoning_effort")
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
if upstreamModel != "" {
body, _ = sjson.SetBytes(body, "model", upstreamModel)
}
body = normalizeThinkingConfig(body, upstreamModel)
if errValidate := validateThinkingConfig(body, upstreamModel); errValidate != nil {
return nil, errValidate
}
// Ensure tools array exists to avoid provider quirks similar to Qwen's behaviour. // Ensure tools array exists to avoid provider quirks similar to Qwen's behaviour.
toolsResult := gjson.GetBytes(body, "tools") toolsResult := gjson.GetBytes(body, "tools")
if toolsResult.Exists() && toolsResult.IsArray() && len(toolsResult.Array()) == 0 { if toolsResult.Exists() && toolsResult.IsArray() && len(toolsResult.Array()) == 0 {
@@ -201,7 +219,7 @@ func (e *IFlowExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
}() }()
scanner := bufio.NewScanner(httpResp.Body) scanner := bufio.NewScanner(httpResp.Body)
scanner.Buffer(nil, 20_971_520) scanner.Buffer(nil, 52_428_800) // 50MB
var param any var param any
for scanner.Scan() { for scanner.Scan() {
line := scanner.Bytes() line := scanner.Bytes()

752
internal/runtime/executor/kiro_executor.go
View File

File diff suppressed because it is too large Load Diff

20
internal/runtime/executor/openai_compat_executor.go
View File

@@ -58,6 +58,15 @@ func (e *OpenAICompatExecutor) Execute(ctx context.Context, auth *cliproxyauth.A
translated = e.overrideModel(translated, modelOverride) translated = e.overrideModel(translated, modelOverride)
} }
translated = applyPayloadConfigWithRoot(e.cfg, req.Model, to.String(), "", translated) translated = applyPayloadConfigWithRoot(e.cfg, req.Model, to.String(), "", translated)
translated = applyReasoningEffortMetadata(translated, req.Metadata, req.Model, "reasoning_effort")
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
if upstreamModel != "" {
translated, _ = sjson.SetBytes(translated, "model", upstreamModel)
}
translated = normalizeThinkingConfig(translated, upstreamModel)
if errValidate := validateThinkingConfig(translated, upstreamModel); errValidate != nil {
return resp, errValidate
}
url := strings.TrimSuffix(baseURL, "/") + "/chat/completions" url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(translated)) httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(translated))
@@ -143,6 +152,15 @@ func (e *OpenAICompatExecutor) ExecuteStream(ctx context.Context, auth *cliproxy
translated = e.overrideModel(translated, modelOverride) translated = e.overrideModel(translated, modelOverride)
} }
translated = applyPayloadConfigWithRoot(e.cfg, req.Model, to.String(), "", translated) translated = applyPayloadConfigWithRoot(e.cfg, req.Model, to.String(), "", translated)
translated = applyReasoningEffortMetadata(translated, req.Metadata, req.Model, "reasoning_effort")
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
if upstreamModel != "" {
translated, _ = sjson.SetBytes(translated, "model", upstreamModel)
}
translated = normalizeThinkingConfig(translated, upstreamModel)
if errValidate := validateThinkingConfig(translated, upstreamModel); errValidate != nil {
return nil, errValidate
}
url := strings.TrimSuffix(baseURL, "/") + "/chat/completions" url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(translated)) httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(translated))
@@ -206,7 +224,7 @@ func (e *OpenAICompatExecutor) ExecuteStream(ctx context.Context, auth *cliproxy
} }
}() }()
scanner := bufio.NewScanner(httpResp.Body) scanner := bufio.NewScanner(httpResp.Body)
scanner.Buffer(nil, 20_971_520) scanner.Buffer(nil, 52_428_800) // 50MB
var param any var param any
for scanner.Scan() { for scanner.Scan() {
line := scanner.Bytes() line := scanner.Bytes()

130
internal/runtime/executor/payload_helpers.go
View File

@@ -1,6 +1,8 @@
package executor package executor
import ( import (
"fmt"
"net/http"
"strings" "strings"
"github.com/router-for-me/CLIProxyAPI/v6/internal/config" "github.com/router-for-me/CLIProxyAPI/v6/internal/config"
@@ -9,11 +11,11 @@ import (
"github.com/tidwall/sjson" "github.com/tidwall/sjson"
) )
// applyThinkingMetadata applies thinking config from model suffix metadata (e.g., -reasoning, -thinking-N) // applyThinkingMetadata applies thinking config from model suffix metadata (e.g., (high), (8192))
// for standard Gemini format payloads. It normalizes the budget when the model supports thinking. // for standard Gemini format payloads. It normalizes the budget when the model supports thinking.
func applyThinkingMetadata(payload []byte, metadata map[string]any, model string) []byte { func applyThinkingMetadata(payload []byte, metadata map[string]any, model string) []byte {
budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(metadata) budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(model, metadata)
if !ok { if !ok || (budgetOverride == nil && includeOverride == nil) {
return payload return payload
} }
if !util.ModelSupportsThinking(model) { if !util.ModelSupportsThinking(model) {
@@ -26,20 +28,44 @@ func applyThinkingMetadata(payload []byte, metadata map[string]any, model string
return util.ApplyGeminiThinkingConfig(payload, budgetOverride, includeOverride) return util.ApplyGeminiThinkingConfig(payload, budgetOverride, includeOverride)
} }
// applyThinkingMetadataCLI applies thinking config from model suffix metadata (e.g., -reasoning, -thinking-N) // applyThinkingMetadataCLI applies thinking config from model suffix metadata (e.g., (high), (8192))
// for Gemini CLI format payloads (nested under "request"). It normalizes the budget when the model supports thinking. // for Gemini CLI format payloads (nested under "request"). It normalizes the budget when the model supports thinking.
func applyThinkingMetadataCLI(payload []byte, metadata map[string]any, model string) []byte { func applyThinkingMetadataCLI(payload []byte, metadata map[string]any, model string) []byte {
budgetOverride, includeOverride, ok := util.GeminiThinkingFromMetadata(metadata) budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(model, metadata)
if !ok { if !ok || (budgetOverride == nil && includeOverride == nil) {
return payload return payload
} }
if budgetOverride != nil && util.ModelSupportsThinking(model) { if !util.ModelSupportsThinking(model) {
return payload
}
if budgetOverride != nil {
norm := util.NormalizeThinkingBudget(model, *budgetOverride) norm := util.NormalizeThinkingBudget(model, *budgetOverride)
budgetOverride = &norm budgetOverride = &norm
} }
return util.ApplyGeminiCLIThinkingConfig(payload, budgetOverride, includeOverride) return util.ApplyGeminiCLIThinkingConfig(payload, budgetOverride, includeOverride)
} }
// applyReasoningEffortMetadata applies reasoning effort overrides from metadata to the given JSON path.
// Metadata values take precedence over any existing field when the model supports thinking, intentionally
// overwriting caller-provided values to honor suffix/default metadata priority.
func applyReasoningEffortMetadata(payload []byte, metadata map[string]any, model, field string) []byte {
if len(metadata) == 0 {
return payload
}
if !util.ModelSupportsThinking(model) {
return payload
}
if field == "" {
return payload
}
if effort, ok := util.ReasoningEffortFromMetadata(metadata); ok && effort != "" {
if updated, err := sjson.SetBytes(payload, field, effort); err == nil {
return updated
}
}
return payload
}
// applyPayloadConfig applies payload default and override rules from configuration // applyPayloadConfig applies payload default and override rules from configuration
// to the given JSON payload for the specified model. // to the given JSON payload for the specified model.
// Defaults only fill missing fields, while overrides always overwrite existing values. // Defaults only fill missing fields, while overrides always overwrite existing values.
@@ -189,3 +215,93 @@ func matchModelPattern(pattern, model string) bool {
} }
return pi == len(pattern) return pi == len(pattern)
} }
// normalizeThinkingConfig normalizes thinking-related fields in the payload
// based on model capabilities. For models without thinking support, it strips
// reasoning fields. For models with level-based thinking, it validates and
// normalizes the reasoning effort level.
func normalizeThinkingConfig(payload []byte, model string) []byte {
if len(payload) == 0 || model == "" {
return payload
}
if !util.ModelSupportsThinking(model) {
return stripThinkingFields(payload)
}
if util.ModelUsesThinkingLevels(model) {
return normalizeReasoningEffortLevel(payload, model)
}
return payload
}
// stripThinkingFields removes thinking-related fields from the payload for
// models that do not support thinking.
func stripThinkingFields(payload []byte) []byte {
fieldsToRemove := []string{
"reasoning",
"reasoning_effort",
"reasoning.effort",
}
out := payload
for _, field := range fieldsToRemove {
if gjson.GetBytes(out, field).Exists() {
out, _ = sjson.DeleteBytes(out, field)
}
}
return out
}
// normalizeReasoningEffortLevel validates and normalizes the reasoning_effort
// or reasoning.effort field for level-based thinking models.
func normalizeReasoningEffortLevel(payload []byte, model string) []byte {
out := payload
if effort := gjson.GetBytes(out, "reasoning_effort"); effort.Exists() {
if normalized, ok := util.NormalizeReasoningEffortLevel(model, effort.String()); ok {
out, _ = sjson.SetBytes(out, "reasoning_effort", normalized)
}
}
if effort := gjson.GetBytes(out, "reasoning.effort"); effort.Exists() {
if normalized, ok := util.NormalizeReasoningEffortLevel(model, effort.String()); ok {
out, _ = sjson.SetBytes(out, "reasoning.effort", normalized)
}
}
return out
}
// validateThinkingConfig checks for unsupported reasoning levels on level-based models.
// Returns a statusErr with 400 when an unsupported level is supplied to avoid silently
// downgrading requests.
func validateThinkingConfig(payload []byte, model string) error {
if len(payload) == 0 || model == "" {
return nil
}
if !util.ModelSupportsThinking(model) || !util.ModelUsesThinkingLevels(model) {
return nil
}
levels := util.GetModelThinkingLevels(model)
checkField := func(path string) error {
if effort := gjson.GetBytes(payload, path); effort.Exists() {
if _, ok := util.NormalizeReasoningEffortLevel(model, effort.String()); !ok {
return statusErr{
code: http.StatusBadRequest,
msg: fmt.Sprintf("unsupported reasoning effort level %q for model %s (supported: %s)", effort.String(), model, strings.Join(levels, ", ")),
}
}
}
return nil
}
if err := checkField("reasoning_effort"); err != nil {
return err
}
if err := checkField("reasoning.effort"); err != nil {
return err
}
return nil
}

21
internal/runtime/executor/qwen_executor.go
View File

@@ -12,6 +12,7 @@ import (
qwenauth "github.com/router-for-me/CLIProxyAPI/v6/internal/auth/qwen" qwenauth "github.com/router-for-me/CLIProxyAPI/v6/internal/auth/qwen"
"github.com/router-for-me/CLIProxyAPI/v6/internal/config" "github.com/router-for-me/CLIProxyAPI/v6/internal/config"
"github.com/router-for-me/CLIProxyAPI/v6/internal/util"
cliproxyauth "github.com/router-for-me/CLIProxyAPI/v6/sdk/cliproxy/auth" cliproxyauth "github.com/router-for-me/CLIProxyAPI/v6/sdk/cliproxy/auth"
cliproxyexecutor "github.com/router-for-me/CLIProxyAPI/v6/sdk/cliproxy/executor" cliproxyexecutor "github.com/router-for-me/CLIProxyAPI/v6/sdk/cliproxy/executor"
sdktranslator "github.com/router-for-me/CLIProxyAPI/v6/sdk/translator" sdktranslator "github.com/router-for-me/CLIProxyAPI/v6/sdk/translator"
@@ -50,6 +51,15 @@ func (e *QwenExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, req
from := opts.SourceFormat from := opts.SourceFormat
to := sdktranslator.FromString("openai") to := sdktranslator.FromString("openai")
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
body = applyReasoningEffortMetadata(body, req.Metadata, req.Model, "reasoning_effort")
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
if upstreamModel != "" {
body, _ = sjson.SetBytes(body, "model", upstreamModel)
}
body = normalizeThinkingConfig(body, upstreamModel)
if errValidate := validateThinkingConfig(body, upstreamModel); errValidate != nil {
return resp, errValidate
}
body = applyPayloadConfig(e.cfg, req.Model, body) body = applyPayloadConfig(e.cfg, req.Model, body)
url := strings.TrimSuffix(baseURL, "/") + "/chat/completions" url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
@@ -121,6 +131,15 @@ func (e *QwenExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Aut
to := sdktranslator.FromString("openai") to := sdktranslator.FromString("openai")
body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true) body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
body = applyReasoningEffortMetadata(body, req.Metadata, req.Model, "reasoning_effort")
upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
if upstreamModel != "" {
body, _ = sjson.SetBytes(body, "model", upstreamModel)
}
body = normalizeThinkingConfig(body, upstreamModel)
if errValidate := validateThinkingConfig(body, upstreamModel); errValidate != nil {
return nil, errValidate
}
toolsResult := gjson.GetBytes(body, "tools") toolsResult := gjson.GetBytes(body, "tools")
// I'm addressing the Qwen3 "poisoning" issue, which is caused by the model needing a tool to be defined. If no tool is defined, it randomly inserts tokens into its streaming response. // I'm addressing the Qwen3 "poisoning" issue, which is caused by the model needing a tool to be defined. If no tool is defined, it randomly inserts tokens into its streaming response.
// This will have no real consequences. It's just to scare Qwen3. // This will have no real consequences. It's just to scare Qwen3.
@@ -181,7 +200,7 @@ func (e *QwenExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Aut
} }
}() }()
scanner := bufio.NewScanner(httpResp.Body) scanner := bufio.NewScanner(httpResp.Body)
scanner.Buffer(nil, 20_971_520) scanner.Buffer(nil, 52_428_800) // 50MB
var param any var param any
for scanner.Scan() { for scanner.Scan() {
line := scanner.Bytes() line := scanner.Bytes()

9
internal/translator/antigravity/claude/antigravity_claude_request.go
View File

@@ -123,6 +123,15 @@ func ConvertClaudeRequestToAntigravity(modelName string, inputRawJSON []byte, _
functionResponse := client.FunctionResponse{ID: toolCallID, Name: funcName, Response: map[string]interface{}{"result": responseData}} functionResponse := client.FunctionResponse{ID: toolCallID, Name: funcName, Response: map[string]interface{}{"result": responseData}}
clientContent.Parts = append(clientContent.Parts, client.Part{FunctionResponse: &functionResponse}) clientContent.Parts = append(clientContent.Parts, client.Part{FunctionResponse: &functionResponse})
} }
} else if contentTypeResult.Type == gjson.String && contentTypeResult.String() == "image" {
sourceResult := contentResult.Get("source")
if sourceResult.Get("type").String() == "base64" {
inlineData := &client.InlineData{
MimeType: sourceResult.Get("media_type").String(),
Data: sourceResult.Get("data").String(),
}
clientContent.Parts = append(clientContent.Parts, client.Part{InlineData: inlineData})
}
} }
} }
contents = append(contents, clientContent) contents = append(contents, clientContent)

23
internal/translator/antigravity/claude/antigravity_claude_response.go
View File

@@ -35,6 +35,7 @@ type Params struct {
TotalTokenCount int64 // Cached total token count from usage metadata TotalTokenCount int64 // Cached total token count from usage metadata
HasSentFinalEvents bool // Indicates if final content/message events have been sent HasSentFinalEvents bool // Indicates if final content/message events have been sent
HasToolUse bool // Indicates if tool use was observed in the stream HasToolUse bool // Indicates if tool use was observed in the stream
HasContent bool // Tracks whether any content (text, thinking, or tool use) has been output
} }
// toolUseIDCounter provides a process-wide unique counter for tool use identifiers. // toolUseIDCounter provides a process-wide unique counter for tool use identifiers.
@@ -69,11 +70,14 @@ func ConvertAntigravityResponseToClaude(_ context.Context, _ string, originalReq
if bytes.Equal(rawJSON, []byte("[DONE]")) { if bytes.Equal(rawJSON, []byte("[DONE]")) {
output := "" output := ""
appendFinalEvents(params, &output, true) // Only send final events if we have actually output content
if params.HasContent {
return []string{ appendFinalEvents(params, &output, true)
output + "event: message_stop\ndata: {\"type\":\"message_stop\"}\n\n\n", return []string{
output + "event: message_stop\ndata: {\"type\":\"message_stop\"}\n\n\n",
}
} }
return []string{}
} }
output := "" output := ""
@@ -119,10 +123,12 @@ func ConvertAntigravityResponseToClaude(_ context.Context, _ string, originalReq
output = output + "event: content_block_delta\n" output = output + "event: content_block_delta\n"
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"signature_delta","signature":""}}`, params.ResponseIndex), "delta.signature", thoughtSignature.String()) data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"signature_delta","signature":""}}`, params.ResponseIndex), "delta.signature", thoughtSignature.String())
output = output + fmt.Sprintf("data: %s\n\n\n", data) output = output + fmt.Sprintf("data: %s\n\n\n", data)
params.HasContent = true
} else if params.ResponseType == 2 { // Continue existing thinking block if already in thinking state } else if params.ResponseType == 2 { // Continue existing thinking block if already in thinking state
output = output + "event: content_block_delta\n" output = output + "event: content_block_delta\n"
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, params.ResponseIndex), "delta.thinking", partTextResult.String()) data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, params.ResponseIndex), "delta.thinking", partTextResult.String())
output = output + fmt.Sprintf("data: %s\n\n\n", data) output = output + fmt.Sprintf("data: %s\n\n\n", data)
params.HasContent = true
} else { } else {
// Transition from another state to thinking // Transition from another state to thinking
// First, close any existing content block // First, close any existing content block
@@ -146,6 +152,7 @@ func ConvertAntigravityResponseToClaude(_ context.Context, _ string, originalReq
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, params.ResponseIndex), "delta.thinking", partTextResult.String()) data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, params.ResponseIndex), "delta.thinking", partTextResult.String())
output = output + fmt.Sprintf("data: %s\n\n\n", data) output = output + fmt.Sprintf("data: %s\n\n\n", data)
params.ResponseType = 2 // Set state to thinking params.ResponseType = 2 // Set state to thinking
params.HasContent = true
} }
} else { } else {
finishReasonResult := gjson.GetBytes(rawJSON, "response.candidates.0.finishReason") finishReasonResult := gjson.GetBytes(rawJSON, "response.candidates.0.finishReason")
@@ -156,6 +163,7 @@ func ConvertAntigravityResponseToClaude(_ context.Context, _ string, originalReq
output = output + "event: content_block_delta\n" output = output + "event: content_block_delta\n"
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"text_delta","text":""}}`, params.ResponseIndex), "delta.text", partTextResult.String()) data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"text_delta","text":""}}`, params.ResponseIndex), "delta.text", partTextResult.String())
output = output + fmt.Sprintf("data: %s\n\n\n", data) output = output + fmt.Sprintf("data: %s\n\n\n", data)
params.HasContent = true
} else { } else {
// Transition from another state to text content // Transition from another state to text content
// First, close any existing content block // First, close any existing content block
@@ -179,6 +187,7 @@ func ConvertAntigravityResponseToClaude(_ context.Context, _ string, originalReq
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"text_delta","text":""}}`, params.ResponseIndex), "delta.text", partTextResult.String()) data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"text_delta","text":""}}`, params.ResponseIndex), "delta.text", partTextResult.String())
output = output + fmt.Sprintf("data: %s\n\n\n", data) output = output + fmt.Sprintf("data: %s\n\n\n", data)
params.ResponseType = 1 // Set state to content params.ResponseType = 1 // Set state to content
params.HasContent = true
} }
} }
} }
@@ -230,6 +239,7 @@ func ConvertAntigravityResponseToClaude(_ context.Context, _ string, originalReq
output = output + fmt.Sprintf("data: %s\n\n\n", data) output = output + fmt.Sprintf("data: %s\n\n\n", data)
} }
params.ResponseType = 3 params.ResponseType = 3
params.HasContent = true
} }
} }
} }
@@ -269,6 +279,11 @@ func appendFinalEvents(params *Params, output *string, force bool) {
return return
} }
// Only send final events if we have actually output content
if !params.HasContent {
return
}
if params.ResponseType != 0 { if params.ResponseType != 0 {
*output = *output + "event: content_block_stop\n" *output = *output + "event: content_block_stop\n"
*output = *output + fmt.Sprintf(`data: {"type":"content_block_stop","index":%d}`, params.ResponseIndex) *output = *output + fmt.Sprintf(`data: {"type":"content_block_stop","index":%d}`, params.ResponseIndex)

5
internal/translator/claude/gemini/claude_gemini_response.go
View File

@@ -331,9 +331,8 @@ func ConvertClaudeResponseToGeminiNonStream(_ context.Context, modelName string,
streamingEvents := make([][]byte, 0) streamingEvents := make([][]byte, 0)
scanner := bufio.NewScanner(bytes.NewReader(rawJSON)) scanner := bufio.NewScanner(bytes.NewReader(rawJSON))
// Use a smaller initial buffer (64KB) that can grow up to 20MB if needed buffer := make([]byte, 52_428_800) // 50MB
// This prevents allocating 20MB for every request regardless of size scanner.Buffer(buffer, 52_428_800)
scanner.Buffer(make([]byte, 64*1024), 20_971_520)
for scanner.Scan() { for scanner.Scan() {
line := scanner.Bytes() line := scanner.Bytes()
// log.Debug(string(line)) // log.Debug(string(line))

4
internal/translator/claude/openai/responses/claude_openai-responses_response.go
View File

@@ -445,8 +445,8 @@ func ConvertClaudeResponseToOpenAIResponsesNonStream(_ context.Context, _ string
// Use a simple scanner to iterate through raw bytes // Use a simple scanner to iterate through raw bytes
// Note: extremely large responses may require increasing the buffer // Note: extremely large responses may require increasing the buffer
scanner := bufio.NewScanner(bytes.NewReader(rawJSON)) scanner := bufio.NewScanner(bytes.NewReader(rawJSON))
buf := make([]byte, 20_971_520) buf := make([]byte, 52_428_800) // 50MB
scanner.Buffer(buf, 20_971_520) scanner.Buffer(buf, 52_428_800)
for scanner.Scan() { for scanner.Scan() {
line := scanner.Bytes() line := scanner.Bytes()
if !bytes.HasPrefix(line, dataTag) { if !bytes.HasPrefix(line, dataTag) {

59
internal/translator/gemini-cli/claude/gemini-cli_claude_response.go
View File

@@ -26,6 +26,7 @@ type Params struct {
HasFirstResponse bool // Indicates if the initial message_start event has been sent HasFirstResponse bool // Indicates if the initial message_start event has been sent
ResponseType int // Current response type: 0=none, 1=content, 2=thinking, 3=function ResponseType int // Current response type: 0=none, 1=content, 2=thinking, 3=function
ResponseIndex int // Index counter for content blocks in the streaming response ResponseIndex int // Index counter for content blocks in the streaming response
HasContent bool // Tracks whether any content (text, thinking, or tool use) has been output
} }
// toolUseIDCounter provides a process-wide unique counter for tool use identifiers. // toolUseIDCounter provides a process-wide unique counter for tool use identifiers.
@@ -57,9 +58,13 @@ func ConvertGeminiCLIResponseToClaude(_ context.Context, _ string, originalReque
} }
if bytes.Equal(rawJSON, []byte("[DONE]")) { if bytes.Equal(rawJSON, []byte("[DONE]")) {
return []string{ // Only send message_stop if we have actually output content
"event: message_stop\ndata: {\"type\":\"message_stop\"}\n\n\n", if (*param).(*Params).HasContent {
return []string{
"event: message_stop\ndata: {\"type\":\"message_stop\"}\n\n\n",
}
} }
return []string{}
} }
// Track whether tools are being used in this response chunk // Track whether tools are being used in this response chunk
@@ -107,7 +112,8 @@ func ConvertGeminiCLIResponseToClaude(_ context.Context, _ string, originalReque
if (*param).(*Params).ResponseType == 2 { if (*param).(*Params).ResponseType == 2 {
sb.WriteString("event: content_block_delta\n") sb.WriteString("event: content_block_delta\n")
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, (*param).(*Params).ResponseIndex), "delta.thinking", partTextResult.String()) data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, (*param).(*Params).ResponseIndex), "delta.thinking", partTextResult.String())
sb.WriteString(fmt.Sprintf("data: %s\n\n\n", data)) output = output + fmt.Sprintf("data: %s\n\n\n", data)
(*param).(*Params).HasContent = true
} else { } else {
// Transition from another state to thinking // Transition from another state to thinking
// First, close any existing content block // First, close any existing content block
@@ -126,6 +132,7 @@ func ConvertGeminiCLIResponseToClaude(_ context.Context, _ string, originalReque
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, (*param).(*Params).ResponseIndex), "delta.thinking", partTextResult.String()) data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, (*param).(*Params).ResponseIndex), "delta.thinking", partTextResult.String())
sb.WriteString(fmt.Sprintf("data: %s\n\n\n", data)) sb.WriteString(fmt.Sprintf("data: %s\n\n\n", data))
(*param).(*Params).ResponseType = 2 // Set state to thinking (*param).(*Params).ResponseType = 2 // Set state to thinking
(*param).(*Params).HasContent = true
} }
} else { } else {
// Process regular text content (user-visible output) // Process regular text content (user-visible output)
@@ -133,7 +140,8 @@ func ConvertGeminiCLIResponseToClaude(_ context.Context, _ string, originalReque
if (*param).(*Params).ResponseType == 1 { if (*param).(*Params).ResponseType == 1 {
sb.WriteString("event: content_block_delta\n") sb.WriteString("event: content_block_delta\n")
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"text_delta","text":""}}`, (*param).(*Params).ResponseIndex), "delta.text", partTextResult.String()) data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"text_delta","text":""}}`, (*param).(*Params).ResponseIndex), "delta.text", partTextResult.String())
sb.WriteString(fmt.Sprintf("data: %s\n\n\n", data)) output = output + fmt.Sprintf("data: %s\n\n\n", data)
(*param).(*Params).HasContent = true
} else { } else {
// Transition from another state to text content // Transition from another state to text content
// First, close any existing content block // First, close any existing content block
@@ -152,6 +160,7 @@ func ConvertGeminiCLIResponseToClaude(_ context.Context, _ string, originalReque
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"text_delta","text":""}}`, (*param).(*Params).ResponseIndex), "delta.text", partTextResult.String()) data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"text_delta","text":""}}`, (*param).(*Params).ResponseIndex), "delta.text", partTextResult.String())
sb.WriteString(fmt.Sprintf("data: %s\n\n\n", data)) sb.WriteString(fmt.Sprintf("data: %s\n\n\n", data))
(*param).(*Params).ResponseType = 1 // Set state to content (*param).(*Params).ResponseType = 1 // Set state to content
(*param).(*Params).HasContent = true
} }
} }
} else if functionCallResult.Exists() { } else if functionCallResult.Exists() {
@@ -194,6 +203,7 @@ func ConvertGeminiCLIResponseToClaude(_ context.Context, _ string, originalReque
sb.WriteString(fmt.Sprintf("data: %s\n\n\n", data)) sb.WriteString(fmt.Sprintf("data: %s\n\n\n", data))
} }
(*param).(*Params).ResponseType = 3 (*param).(*Params).ResponseType = 3
(*param).(*Params).HasContent = true
} }
} }
} }
@@ -202,28 +212,31 @@ func ConvertGeminiCLIResponseToClaude(_ context.Context, _ string, originalReque
// Process usage metadata and finish reason when present in the response // Process usage metadata and finish reason when present in the response
if usageResult.Exists() && bytes.Contains(rawJSON, []byte(`"finishReason"`)) { if usageResult.Exists() && bytes.Contains(rawJSON, []byte(`"finishReason"`)) {
if candidatesTokenCountResult := usageResult.Get("candidatesTokenCount"); candidatesTokenCountResult.Exists() { if candidatesTokenCountResult := usageResult.Get("candidatesTokenCount"); candidatesTokenCountResult.Exists() {
// Close the final content block // Only send final events if we have actually output content
sb.WriteString("event: content_block_stop\n") if (*param).(*Params).HasContent {
sb.WriteString(fmt.Sprintf(`data: {"type":"content_block_stop","index":%d}`, (*param).(*Params).ResponseIndex)) // Close the final content block
sb.WriteString("\n\n\n") output = output + "event: content_block_stop\n"
output = output + fmt.Sprintf(`data: {"type":"content_block_stop","index":%d}`, (*param).(*Params).ResponseIndex)
output = output + "\n\n\n"
// Send the final message delta with usage information and stop reason // Send the final message delta with usage information and stop reason
sb.WriteString("event: message_delta\n") output = output + "event: message_delta\n"
sb.WriteString(`data: `) output = output + `data: `
// Create the message delta template with appropriate stop reason // Create the message delta template with appropriate stop reason
template := `{"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":0,"output_tokens":0}}` template := `{"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":0,"output_tokens":0}}`
// Set tool_use stop reason if tools were used in this response // Set tool_use stop reason if tools were used in this response
if usedTool { if usedTool {
template = `{"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":0,"output_tokens":0}}` template = `{"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":0,"output_tokens":0}}`
}
// Include thinking tokens in output token count if present
thoughtsTokenCount := usageResult.Get("thoughtsTokenCount").Int()
template, _ = sjson.Set(template, "usage.output_tokens", candidatesTokenCountResult.Int()+thoughtsTokenCount)
template, _ = sjson.Set(template, "usage.input_tokens", usageResult.Get("promptTokenCount").Int())
output = output + template + "\n\n\n"
} }
// Include thinking tokens in output token count if present
thoughtsTokenCount := usageResult.Get("thoughtsTokenCount").Int()
template, _ = sjson.Set(template, "usage.output_tokens", candidatesTokenCountResult.Int()+thoughtsTokenCount)
template, _ = sjson.Set(template, "usage.input_tokens", usageResult.Get("promptTokenCount").Int())
sb.WriteString(template + "\n\n\n")
} }
} }

45
internal/translator/gemini/claude/gemini_claude_response.go
View File

@@ -25,6 +25,7 @@ type Params struct {
HasFirstResponse bool HasFirstResponse bool
ResponseType int ResponseType int
ResponseIndex int ResponseIndex int
HasContent bool // Tracks whether any content (text, thinking, or tool use) has been output
} }
// toolUseIDCounter provides a process-wide unique counter for tool use identifiers. // toolUseIDCounter provides a process-wide unique counter for tool use identifiers.
@@ -57,9 +58,13 @@ func ConvertGeminiResponseToClaude(_ context.Context, _ string, originalRequestR
} }
if bytes.Equal(rawJSON, []byte("[DONE]")) { if bytes.Equal(rawJSON, []byte("[DONE]")) {
return []string{ // Only send message_stop if we have actually output content
"event: message_stop\ndata: {\"type\":\"message_stop\"}\n\n\n", if (*param).(*Params).HasContent {
return []string{
"event: message_stop\ndata: {\"type\":\"message_stop\"}\n\n\n",
}
} }
return []string{}
} }
// Track whether tools are being used in this response chunk // Track whether tools are being used in this response chunk
@@ -108,6 +113,7 @@ func ConvertGeminiResponseToClaude(_ context.Context, _ string, originalRequestR
output = output + "event: content_block_delta\n" output = output + "event: content_block_delta\n"
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, (*param).(*Params).ResponseIndex), "delta.thinking", partTextResult.String()) data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, (*param).(*Params).ResponseIndex), "delta.thinking", partTextResult.String())
output = output + fmt.Sprintf("data: %s\n\n\n", data) output = output + fmt.Sprintf("data: %s\n\n\n", data)
(*param).(*Params).HasContent = true
} else { } else {
// Transition from another state to thinking // Transition from another state to thinking
// First, close any existing content block // First, close any existing content block
@@ -131,6 +137,7 @@ func ConvertGeminiResponseToClaude(_ context.Context, _ string, originalRequestR
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, (*param).(*Params).ResponseIndex), "delta.thinking", partTextResult.String()) data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"thinking_delta","thinking":""}}`, (*param).(*Params).ResponseIndex), "delta.thinking", partTextResult.String())
output = output + fmt.Sprintf("data: %s\n\n\n", data) output = output + fmt.Sprintf("data: %s\n\n\n", data)
(*param).(*Params).ResponseType = 2 // Set state to thinking (*param).(*Params).ResponseType = 2 // Set state to thinking
(*param).(*Params).HasContent = true
} }
} else { } else {
// Process regular text content (user-visible output) // Process regular text content (user-visible output)
@@ -139,6 +146,7 @@ func ConvertGeminiResponseToClaude(_ context.Context, _ string, originalRequestR
output = output + "event: content_block_delta\n" output = output + "event: content_block_delta\n"
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"text_delta","text":""}}`, (*param).(*Params).ResponseIndex), "delta.text", partTextResult.String()) data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"text_delta","text":""}}`, (*param).(*Params).ResponseIndex), "delta.text", partTextResult.String())
output = output + fmt.Sprintf("data: %s\n\n\n", data) output = output + fmt.Sprintf("data: %s\n\n\n", data)
(*param).(*Params).HasContent = true
} else { } else {
// Transition from another state to text content // Transition from another state to text content
// First, close any existing content block // First, close any existing content block
@@ -162,6 +170,7 @@ func ConvertGeminiResponseToClaude(_ context.Context, _ string, originalRequestR
data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"text_delta","text":""}}`, (*param).(*Params).ResponseIndex), "delta.text", partTextResult.String()) data, _ := sjson.Set(fmt.Sprintf(`{"type":"content_block_delta","index":%d,"delta":{"type":"text_delta","text":""}}`, (*param).(*Params).ResponseIndex), "delta.text", partTextResult.String())
output = output + fmt.Sprintf("data: %s\n\n\n", data) output = output + fmt.Sprintf("data: %s\n\n\n", data)
(*param).(*Params).ResponseType = 1 // Set state to content (*param).(*Params).ResponseType = 1 // Set state to content
(*param).(*Params).HasContent = true
} }
} }
} else if functionCallResult.Exists() { } else if functionCallResult.Exists() {
@@ -211,6 +220,7 @@ func ConvertGeminiResponseToClaude(_ context.Context, _ string, originalRequestR
output = output + fmt.Sprintf("data: %s\n\n\n", data) output = output + fmt.Sprintf("data: %s\n\n\n", data)
} }
(*param).(*Params).ResponseType = 3 (*param).(*Params).ResponseType = 3
(*param).(*Params).HasContent = true
} }
} }
} }
@@ -218,23 +228,26 @@ func ConvertGeminiResponseToClaude(_ context.Context, _ string, originalRequestR
usageResult := gjson.GetBytes(rawJSON, "usageMetadata") usageResult := gjson.GetBytes(rawJSON, "usageMetadata")
if usageResult.Exists() && bytes.Contains(rawJSON, []byte(`"finishReason"`)) { if usageResult.Exists() && bytes.Contains(rawJSON, []byte(`"finishReason"`)) {
if candidatesTokenCountResult := usageResult.Get("candidatesTokenCount"); candidatesTokenCountResult.Exists() { if candidatesTokenCountResult := usageResult.Get("candidatesTokenCount"); candidatesTokenCountResult.Exists() {
output = output + "event: content_block_stop\n" // Only send final events if we have actually output content
output = output + fmt.Sprintf(`data: {"type":"content_block_stop","index":%d}`, (*param).(*Params).ResponseIndex) if (*param).(*Params).HasContent {
output = output + "\n\n\n" output = output + "event: content_block_stop\n"
output = output + fmt.Sprintf(`data: {"type":"content_block_stop","index":%d}`, (*param).(*Params).ResponseIndex)
output = output + "\n\n\n"
output = output + "event: message_delta\n" output = output + "event: message_delta\n"
output = output + `data: ` output = output + `data: `
template := `{"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":0,"output_tokens":0}}` template := `{"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":0,"output_tokens":0}}`
if usedTool { if usedTool {
template = `{"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":0,"output_tokens":0}}` template = `{"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":0,"output_tokens":0}}`
}
thoughtsTokenCount := usageResult.Get("thoughtsTokenCount").Int()
template, _ = sjson.Set(template, "usage.output_tokens", candidatesTokenCountResult.Int()+thoughtsTokenCount)
template, _ = sjson.Set(template, "usage.input_tokens", usageResult.Get("promptTokenCount").Int())
output = output + template + "\n\n\n"
} }
thoughtsTokenCount := usageResult.Get("thoughtsTokenCount").Int()
template, _ = sjson.Set(template, "usage.output_tokens", candidatesTokenCountResult.Int()+thoughtsTokenCount)
template, _ = sjson.Set(template, "usage.input_tokens", usageResult.Get("promptTokenCount").Int())
output = output + template + "\n\n\n"
} }
} }

5
internal/translator/kiro/claude/kiro_claude.go
View File

@@ -1,5 +1,6 @@
// Package claude provides translation between Kiro and Claude formats. // Package claude provides translation between Kiro and Claude formats.
// Since Kiro uses Claude-compatible format internally, translations are mostly pass-through. // Since Kiro executor generates Claude-compatible SSE format internally (with event: prefix),
// translations are pass-through.
package claude package claude
import ( import (
@@ -14,6 +15,8 @@ func ConvertClaudeRequestToKiro(modelName string, inputRawJSON []byte, stream bo
} }
// ConvertKiroResponseToClaude converts Kiro streaming response to Claude format. // ConvertKiroResponseToClaude converts Kiro streaming response to Claude format.
// Kiro executor already generates complete SSE format with "event:" prefix,
// so this is a simple pass-through.
func ConvertKiroResponseToClaude(ctx context.Context, model string, originalRequest, request, rawResponse []byte, param *any) []string { func ConvertKiroResponseToClaude(ctx context.Context, model string, originalRequest, request, rawResponse []byte, param *any) []string {
return []string{string(rawResponse)} return []string{string(rawResponse)}
} }

63
internal/translator/kiro/openai/chat-completions/kiro_openai_response.go
View File

@@ -4,6 +4,7 @@ package chat_completions
import ( import (
"context" "context"
"encoding/json" "encoding/json"
"strings"
"time" "time"
"github.com/google/uuid" "github.com/google/uuid"
@@ -13,15 +14,58 @@ import (
// ConvertKiroResponseToOpenAI converts Kiro streaming response to OpenAI SSE format. // ConvertKiroResponseToOpenAI converts Kiro streaming response to OpenAI SSE format.
// Handles Claude SSE events: content_block_start, content_block_delta, input_json_delta, // Handles Claude SSE events: content_block_start, content_block_delta, input_json_delta,
// content_block_stop, message_delta, and message_stop. // content_block_stop, message_delta, and message_stop.
// Input may be in SSE format: "event: xxx\ndata: {...}" or raw JSON.
func ConvertKiroResponseToOpenAI(ctx context.Context, model string, originalRequest, request, rawResponse []byte, param *any) []string { func ConvertKiroResponseToOpenAI(ctx context.Context, model string, originalRequest, request, rawResponse []byte, param *any) []string {
root := gjson.ParseBytes(rawResponse) raw := string(rawResponse)
var results []string
// Handle SSE format: extract JSON from "data: " lines
// Input format: "event: message_start\ndata: {...}"
lines := strings.Split(raw, "\n")
for _, line := range lines {
line = strings.TrimSpace(line)
if strings.HasPrefix(line, "data: ") {
jsonPart := strings.TrimPrefix(line, "data: ")
chunks := convertClaudeEventToOpenAI(jsonPart, model)
results = append(results, chunks...)
} else if strings.HasPrefix(line, "{") {
// Raw JSON (backward compatibility)
chunks := convertClaudeEventToOpenAI(line, model)
results = append(results, chunks...)
}
}
return results
}
// convertClaudeEventToOpenAI converts a single Claude JSON event to OpenAI format
func convertClaudeEventToOpenAI(jsonStr string, model string) []string {
root := gjson.Parse(jsonStr)
var results []string var results []string
eventType := root.Get("type").String() eventType := root.Get("type").String()
switch eventType { switch eventType {
case "message_start": case "message_start":
// Initial message event - could emit initial chunk if needed // Initial message event - emit initial chunk with role
response := map[string]interface{}{
"id": "chatcmpl-" + uuid.New().String()[:24],
"object": "chat.completion.chunk",
"created": time.Now().Unix(),
"model": model,
"choices": []map[string]interface{}{
{
"index": 0,
"delta": map[string]interface{}{
"role": "assistant",
"content": "",
},
"finish_reason": nil,
},
},
}
result, _ := json.Marshal(response)
results = append(results, string(result))
return results return results
case "content_block_start": case "content_block_start":
@@ -127,7 +171,7 @@ func ConvertKiroResponseToOpenAI(ctx context.Context, model string, originalRequ
return results return results
case "message_delta": case "message_delta":
// Final message delta with stop_reason // Final message delta with stop_reason and usage
stopReason := root.Get("delta.stop_reason").String() stopReason := root.Get("delta.stop_reason").String()
if stopReason != "" { if stopReason != "" {
finishReason := "stop" finishReason := "stop"
@@ -152,6 +196,19 @@ func ConvertKiroResponseToOpenAI(ctx context.Context, model string, originalRequ
}, },
}, },
} }
// Extract and include usage information from message_delta event
usage := root.Get("usage")
if usage.Exists() {
inputTokens := usage.Get("input_tokens").Int()
outputTokens := usage.Get("output_tokens").Int()
response["usage"] = map[string]interface{}{
"prompt_tokens": inputTokens,
"completion_tokens": outputTokens,
"total_tokens": inputTokens + outputTokens,
}
}
result, _ := json.Marshal(response) result, _ := json.Marshal(response)
results = append(results, string(result)) results = append(results, string(result))
} }

46
internal/util/claude_thinking.go Normal file
View File

@@ -0,0 +1,46 @@
package util
import (
"github.com/tidwall/gjson"
"github.com/tidwall/sjson"
)
// ApplyClaudeThinkingConfig applies thinking configuration to a Claude API request payload.
// It sets the thinking.type to "enabled" and thinking.budget_tokens to the specified budget.
// If budget is nil or the payload already has thinking config, it returns the payload unchanged.
func ApplyClaudeThinkingConfig(body []byte, budget *int) []byte {
if budget == nil {
return body
}
if gjson.GetBytes(body, "thinking").Exists() {
return body
}
if *budget <= 0 {
return body
}
updated := body
updated, _ = sjson.SetBytes(updated, "thinking.type", "enabled")
updated, _ = sjson.SetBytes(updated, "thinking.budget_tokens", *budget)
return updated
}
// ResolveClaudeThinkingConfig resolves thinking configuration from metadata for Claude models.
// It uses the unified ResolveThinkingConfigFromMetadata and normalizes the budget.
// Returns the normalized budget (nil if thinking should not be enabled) and whether it matched.
func ResolveClaudeThinkingConfig(modelName string, metadata map[string]any) (*int, bool) {
budget, include, matched := ResolveThinkingConfigFromMetadata(modelName, metadata)
if !matched {
return nil, false
}
if include != nil && !*include {
return nil, true
}
if budget == nil {
return nil, true
}
normalized := NormalizeThinkingBudget(modelName, *budget)
if normalized <= 0 {
return nil, true
}
return &normalized, true
}

150
internal/util/gemini_thinking.go
View File

@@ -1,8 +1,6 @@
package util package util
import ( import (
"encoding/json"
"strconv"
"strings" "strings"
"github.com/tidwall/gjson" "github.com/tidwall/gjson"
@@ -15,80 +13,6 @@ const (
GeminiOriginalModelMetadataKey = "gemini_original_model" GeminiOriginalModelMetadataKey = "gemini_original_model"
) )
func ParseGeminiThinkingSuffix(model string) (string, *int, *bool, bool) {
if model == "" {
return model, nil, nil, false
}
lower := strings.ToLower(model)
if !strings.HasPrefix(lower, "gemini-") {
return model, nil, nil, false
}
if strings.HasSuffix(lower, "-nothinking") {
base := model[:len(model)-len("-nothinking")]
budgetValue := 0
if strings.HasPrefix(lower, "gemini-2.5-pro") {
budgetValue = 128
}
include := false
return base, &budgetValue, &include, true
}
// Handle "-reasoning" suffix: enables thinking with dynamic budget (-1)
// Maps: gemini-2.5-flash-reasoning -> gemini-2.5-flash with thinkingBudget=-1
if strings.HasSuffix(lower, "-reasoning") {
base := model[:len(model)-len("-reasoning")]
budgetValue := -1 // Dynamic budget
include := true
return base, &budgetValue, &include, true
}
idx := strings.LastIndex(lower, "-thinking-")
if idx == -1 {
return model, nil, nil, false
}
digits := model[idx+len("-thinking-"):]
if digits == "" {
return model, nil, nil, false
}
end := len(digits)
for i := 0; i < len(digits); i++ {
if digits[i] < '0' || digits[i] > '9' {
end = i
break
}
}
if end == 0 {
return model, nil, nil, false
}
valueStr := digits[:end]
value, err := strconv.Atoi(valueStr)
if err != nil {
return model, nil, nil, false
}
base := model[:idx]
budgetValue := value
return base, &budgetValue, nil, true
}
func NormalizeGeminiThinkingModel(modelName string) (string, map[string]any) {
baseModel, budget, include, matched := ParseGeminiThinkingSuffix(modelName)
if !matched {
return baseModel, nil
}
metadata := map[string]any{
GeminiOriginalModelMetadataKey: modelName,
}
if budget != nil {
metadata[GeminiThinkingBudgetMetadataKey] = *budget
}
if include != nil {
metadata[GeminiIncludeThoughtsMetadataKey] = *include
}
return baseModel, metadata
}
func ApplyGeminiThinkingConfig(body []byte, budget *int, includeThoughts *bool) []byte { func ApplyGeminiThinkingConfig(body []byte, budget *int, includeThoughts *bool) []byte {
if budget == nil && includeThoughts == nil { if budget == nil && includeThoughts == nil {
return body return body
@@ -133,80 +57,6 @@ func ApplyGeminiCLIThinkingConfig(body []byte, budget *int, includeThoughts *boo
return updated return updated
} }
func GeminiThinkingFromMetadata(metadata map[string]any) (*int, *bool, bool) {
if len(metadata) == 0 {
return nil, nil, false
}
var (
budgetPtr *int
includePtr *bool
matched bool
)
if rawBudget, ok := metadata[GeminiThinkingBudgetMetadataKey]; ok {
switch v := rawBudget.(type) {
case int:
budget := v
budgetPtr = &budget
matched = true
case int32:
budget := int(v)
budgetPtr = &budget
matched = true
case int64:
budget := int(v)
budgetPtr = &budget
matched = true
case float64:
budget := int(v)
budgetPtr = &budget
matched = true
case json.Number:
if val, err := v.Int64(); err == nil {
budget := int(val)
budgetPtr = &budget
matched = true
}
}
}
if rawInclude, ok := metadata[GeminiIncludeThoughtsMetadataKey]; ok {
switch v := rawInclude.(type) {
case bool:
include := v
includePtr = &include
matched = true
case string:
if parsed, err := strconv.ParseBool(v); err == nil {
include := parsed
includePtr = &include
matched = true
}
case json.Number:
if val, err := v.Int64(); err == nil {
include := val != 0
includePtr = &include
matched = true
}
case int:
include := v != 0
includePtr = &include
matched = true
case int32:
include := v != 0
includePtr = &include
matched = true
case int64:
include := v != 0
includePtr = &include
matched = true
case float64:
include := v != 0
includePtr = &include
matched = true
}
}
return budgetPtr, includePtr, matched
}
// modelsWithDefaultThinking lists models that should have thinking enabled by default // modelsWithDefaultThinking lists models that should have thinking enabled by default
// when no explicit thinkingConfig is provided. // when no explicit thinkingConfig is provided.
var modelsWithDefaultThinking = map[string]bool{ var modelsWithDefaultThinking = map[string]bool{

38
internal/util/thinking.go
View File

@@ -1,6 +1,8 @@
package util package util
import ( import (
"strings"
"github.com/router-for-me/CLIProxyAPI/v6/internal/registry" "github.com/router-for-me/CLIProxyAPI/v6/internal/registry"
) )
@@ -67,3 +69,39 @@ func thinkingRangeFromRegistry(model string) (found bool, min int, max int, zero
} }
return true, info.Thinking.Min, info.Thinking.Max, info.Thinking.ZeroAllowed, info.Thinking.DynamicAllowed return true, info.Thinking.Min, info.Thinking.Max, info.Thinking.ZeroAllowed, info.Thinking.DynamicAllowed
} }
// GetModelThinkingLevels returns the discrete reasoning effort levels for the model.
// Returns nil if the model has no thinking support or no levels defined.
func GetModelThinkingLevels(model string) []string {
if model == "" {
return nil
}
info := registry.GetGlobalRegistry().GetModelInfo(model)
if info == nil || info.Thinking == nil {
return nil
}
return info.Thinking.Levels
}
// ModelUsesThinkingLevels reports whether the model uses discrete reasoning
// effort levels instead of numeric budgets.
func ModelUsesThinkingLevels(model string) bool {
levels := GetModelThinkingLevels(model)
return len(levels) > 0
}
// NormalizeReasoningEffortLevel validates and normalizes a reasoning effort
// level for the given model. Returns false when the level is not supported.
func NormalizeReasoningEffortLevel(model, effort string) (string, bool) {
levels := GetModelThinkingLevels(model)
if len(levels) == 0 {
return "", false
}
loweredEffort := strings.ToLower(strings.TrimSpace(effort))
for _, lvl := range levels {
if strings.ToLower(lvl) == loweredEffort {
return lvl, true
}
}
return "", false
}

313
internal/util/thinking_suffix.go Normal file
View File

@@ -0,0 +1,313 @@
package util
import (
"encoding/json"
"strconv"
"strings"
)
const (
ThinkingBudgetMetadataKey = "thinking_budget"
ThinkingIncludeThoughtsMetadataKey = "thinking_include_thoughts"
ReasoningEffortMetadataKey = "reasoning_effort"
ThinkingOriginalModelMetadataKey = "thinking_original_model"
)
// NormalizeThinkingModel parses dynamic thinking suffixes on model names and returns
// the normalized base model with extracted metadata. Supported pattern:
// - "(<value>)" where value can be:
// - A numeric budget (e.g., "(8192)", "(16384)")
// - A reasoning effort level (e.g., "(high)", "(medium)", "(low)")
//
// Examples:
// - "claude-sonnet-4-5-20250929(16384)" → budget=16384
// - "gpt-5.1(high)" → reasoning_effort="high"
// - "gemini-2.5-pro(32768)" → budget=32768
//
// Note: Empty parentheses "()" are not supported and will be ignored.
func NormalizeThinkingModel(modelName string) (string, map[string]any) {
if modelName == "" {
return modelName, nil
}
baseModel := modelName
var (
budgetOverride *int
reasoningEffort *string
matched bool
)
// Match "(<value>)" pattern at the end of the model name
if idx := strings.LastIndex(modelName, "("); idx != -1 {
if !strings.HasSuffix(modelName, ")") {
// Incomplete parenthesis, ignore
return baseModel, nil
}
value := modelName[idx+1 : len(modelName)-1] // Extract content between ( and )
if value == "" {
// Empty parentheses not supported
return baseModel, nil
}
candidateBase := modelName[:idx]
// Auto-detect: pure numeric → budget, string → reasoning effort level
if parsed, ok := parseIntPrefix(value); ok {
// Numeric value: treat as thinking budget
baseModel = candidateBase
budgetOverride = &parsed
matched = true
} else {
// String value: treat as reasoning effort level
baseModel = candidateBase
raw := strings.ToLower(strings.TrimSpace(value))
if raw != "" {
reasoningEffort = &raw
matched = true
}
}
}
if !matched {
return baseModel, nil
}
metadata := map[string]any{
ThinkingOriginalModelMetadataKey: modelName,
}
if budgetOverride != nil {
metadata[ThinkingBudgetMetadataKey] = *budgetOverride
}
if reasoningEffort != nil {
metadata[ReasoningEffortMetadataKey] = *reasoningEffort
}
return baseModel, metadata
}
// ThinkingFromMetadata extracts thinking overrides from metadata produced by NormalizeThinkingModel.
// It accepts both the new generic keys and legacy Gemini-specific keys.
func ThinkingFromMetadata(metadata map[string]any) (*int, *bool, *string, bool) {
if len(metadata) == 0 {
return nil, nil, nil, false
}
var (
budgetPtr *int
includePtr *bool
effortPtr *string
matched bool
)
readBudget := func(key string) {
if budgetPtr != nil {
return
}
if raw, ok := metadata[key]; ok {
if v, okNumber := parseNumberToInt(raw); okNumber {
budget := v
budgetPtr = &budget
matched = true
}
}
}
readInclude := func(key string) {
if includePtr != nil {
return
}
if raw, ok := metadata[key]; ok {
switch v := raw.(type) {
case bool:
val := v
includePtr = &val
matched = true
case *bool:
if v != nil {
val := *v
includePtr = &val
matched = true
}
}
}
}
readEffort := func(key string) {
if effortPtr != nil {
return
}
if raw, ok := metadata[key]; ok {
if val, okStr := raw.(string); okStr && strings.TrimSpace(val) != "" {
normalized := strings.ToLower(strings.TrimSpace(val))
effortPtr = &normalized
matched = true
}
}
}
readBudget(ThinkingBudgetMetadataKey)
readBudget(GeminiThinkingBudgetMetadataKey)
readInclude(ThinkingIncludeThoughtsMetadataKey)
readInclude(GeminiIncludeThoughtsMetadataKey)
readEffort(ReasoningEffortMetadataKey)
readEffort("reasoning.effort")
return budgetPtr, includePtr, effortPtr, matched
}
// ResolveThinkingConfigFromMetadata derives thinking budget/include overrides,
// converting reasoning effort strings into budgets when possible.
func ResolveThinkingConfigFromMetadata(model string, metadata map[string]any) (*int, *bool, bool) {
budget, include, effort, matched := ThinkingFromMetadata(metadata)
if !matched {
return nil, nil, false
}
if budget == nil && effort != nil {
if derived, ok := ThinkingEffortToBudget(model, *effort); ok {
budget = &derived
}
}
return budget, include, budget != nil || include != nil || effort != nil
}
// ReasoningEffortFromMetadata resolves a reasoning effort string from metadata,
// inferring "auto" and "none" when budgets request dynamic or disabled thinking.
func ReasoningEffortFromMetadata(metadata map[string]any) (string, bool) {
budget, include, effort, matched := ThinkingFromMetadata(metadata)
if !matched {
return "", false
}
if effort != nil && *effort != "" {
return strings.ToLower(strings.TrimSpace(*effort)), true
}
if budget != nil {
switch *budget {
case -1:
return "auto", true
case 0:
return "none", true
}
}
if include != nil && !*include {
return "none", true
}
return "", true
}
// ThinkingEffortToBudget maps reasoning effort levels to approximate budgets,
// clamping the result to the model's supported range.
func ThinkingEffortToBudget(model, effort string) (int, bool) {
if effort == "" {
return 0, false
}
normalized, ok := NormalizeReasoningEffortLevel(model, effort)
if !ok {
normalized = strings.ToLower(strings.TrimSpace(effort))
}
switch normalized {
case "none":
return 0, true
case "auto":
return NormalizeThinkingBudget(model, -1), true
case "minimal":
return NormalizeThinkingBudget(model, 512), true
case "low":
return NormalizeThinkingBudget(model, 1024), true
case "medium":
return NormalizeThinkingBudget(model, 8192), true
case "high":
return NormalizeThinkingBudget(model, 24576), true
case "xhigh":
return NormalizeThinkingBudget(model, 32768), true
default:
return 0, false
}
}
// ResolveOriginalModel returns the original model name stored in metadata (if present),
// otherwise falls back to the provided model.
func ResolveOriginalModel(model string, metadata map[string]any) string {
normalize := func(name string) string {
if name == "" {
return ""
}
if base, _ := NormalizeThinkingModel(name); base != "" {
return base
}
return strings.TrimSpace(name)
}
if metadata != nil {
if v, ok := metadata[ThinkingOriginalModelMetadataKey]; ok {
if s, okStr := v.(string); okStr && strings.TrimSpace(s) != "" {
if base := normalize(s); base != "" {
return base
}
}
}
if v, ok := metadata[GeminiOriginalModelMetadataKey]; ok {
if s, okStr := v.(string); okStr && strings.TrimSpace(s) != "" {
if base := normalize(s); base != "" {
return base
}
}
}
}
// Fallback: try to re-normalize the model name when metadata was dropped.
if base := normalize(model); base != "" {
return base
}
return model
}
func parseIntPrefix(value string) (int, bool) {
if value == "" {
return 0, false
}
digits := strings.TrimLeft(value, "-")
if digits == "" {
return 0, false
}
end := len(digits)
for i := 0; i < len(digits); i++ {
if digits[i] < '0' || digits[i] > '9' {
end = i
break
}
}
if end == 0 {
return 0, false
}
val, err := strconv.Atoi(digits[:end])
if err != nil {
return 0, false
}
return val, true
}
func parseNumberToInt(raw any) (int, bool) {
switch v := raw.(type) {
case int:
return v, true
case int32:
return int(v), true
case int64:
return int(v), true
case float64:
return int(v), true
case json.Number:
if val, err := v.Int64(); err == nil {
return int(val), true
}
case string:
if strings.TrimSpace(v) == "" {
return 0, false
}
if parsed, err := strconv.Atoi(strings.TrimSpace(v)); err == nil {
return parsed, true
}
}
return 0, false
}

6
sdk/api/handlers/claude/code_handlers.go
View File

@@ -271,6 +271,11 @@ func (h *ClaudeCodeAPIHandler) forwardClaudeStream(c *gin.Context, flusher http.
continue continue
} }
if errMsg != nil { if errMsg != nil {
status := http.StatusInternalServerError
if errMsg.StatusCode > 0 {
status = errMsg.StatusCode
}
c.Status(status)
// An error occurred: emit as a proper SSE error event // An error occurred: emit as a proper SSE error event
errorBytes, _ := json.Marshal(h.toClaudeError(errMsg)) errorBytes, _ := json.Marshal(h.toClaudeError(errMsg))
_, _ = writer.WriteString("event: error\n") _, _ = writer.WriteString("event: error\n")
@@ -278,6 +283,7 @@ func (h *ClaudeCodeAPIHandler) forwardClaudeStream(c *gin.Context, flusher http.
_, _ = writer.Write(errorBytes) _, _ = writer.Write(errorBytes)
_, _ = writer.WriteString("\n\n") _, _ = writer.WriteString("\n\n")
_ = writer.Flush() _ = writer.Flush()
flusher.Flush()
} }
var execErr error var execErr error
if errMsg != nil { if errMsg != nil {

28
sdk/api/handlers/handlers.go
View File

@@ -343,18 +343,32 @@ func (h *BaseAPIHandler) getRequestDetails(modelName string) (providers []string
providerName, extractedModelName, isDynamic := h.parseDynamicModel(resolvedModelName) providerName, extractedModelName, isDynamic := h.parseDynamicModel(resolvedModelName)
// First, normalize the model name to handle suffixes like "-thinking-128" targetModelName := resolvedModelName
// This needs to happen before determining the provider for non-dynamic models. if isDynamic {
normalizedModel, metadata = normalizeModelMetadata(resolvedModelName) targetModelName = extractedModelName
}
// Normalize the model name to handle dynamic thinking suffixes before determining the provider.
normalizedModel, metadata = normalizeModelMetadata(targetModelName)
if isDynamic { if isDynamic {
providers = []string{providerName} providers = []string{providerName}
// For dynamic models, the extractedModelName is already normalized by parseDynamicModel
// so we use it as the final normalizedModel.
normalizedModel = extractedModelName
} else { } else {
// For non-dynamic models, use the normalizedModel to get the provider name. // For non-dynamic models, use the normalizedModel to get the provider name.
providers = util.GetProviderName(normalizedModel) providers = util.GetProviderName(normalizedModel)
if len(providers) == 0 && metadata != nil {
if originalRaw, ok := metadata[util.ThinkingOriginalModelMetadataKey]; ok {
if originalModel, okStr := originalRaw.(string); okStr {
originalModel = strings.TrimSpace(originalModel)
if originalModel != "" && !strings.EqualFold(originalModel, normalizedModel) {
if altProviders := util.GetProviderName(originalModel); len(altProviders) > 0 {
providers = altProviders
normalizedModel = originalModel
}
}
}
}
}
} }
if len(providers) == 0 { if len(providers) == 0 {
@@ -402,7 +416,7 @@ func cloneBytes(src []byte) []byte {
} }
func normalizeModelMetadata(modelName string) (string, map[string]any) { func normalizeModelMetadata(modelName string) (string, map[string]any) {
return util.NormalizeGeminiThinkingModel(modelName) return util.NormalizeThinkingModel(modelName)
} }
func cloneMetadata(src map[string]any) map[string]any { func cloneMetadata(src map[string]any) map[string]any {