Luis Pater
dbb433fcf8
Merge branch 'main' into plus
2026-01-25 21:51:02 +08:00
hkfires
5743b78694
test(claude): update expectations for system message handling
2026-01-25 08:31:29 +08:00
Luis Pater
0d6ecb0191
Fixed : #1077
...
refactor(translator): improve tools handling by separating functionDeclarations and googleSearch nodes
2026-01-24 05:51:11 +08:00
Luis Pater
b9ff916494
Merge branch 'main' into plus
2026-01-24 05:02:32 +08:00
Luis Pater
873d41582f
Merge pull request #1125 from NightHammer1000/dev
...
Filter out Top_P when Temp is set on Claude
2026-01-24 02:03:33 +08:00
Luis Pater
6fb7d85558
Merge pull request #1137 from augustVino/fix/remove_empty_systemmsg
...
fix(translator): ensure system message is only added if it contains c…
2026-01-24 02:02:18 +08:00
Luis Pater
51611c25d7
Merge branch 'router-for-me:main' into main
2026-01-21 22:12:28 +08:00
hkfires
c8884f5e25
refactor(translator): enhance signature handling in Claude and Gemini requests, streamline cache usage and remove unnecessary tests
2026-01-21 20:21:49 +08:00
Luis Pater
d9c6317c84
refactor(cache, translator): refine signature caching logic and tests, replace session-based logic with model group handling
2026-01-21 18:30:05 +08:00
Vino
d29ec95526
fix(translator): ensure system message is only added if it contains content
2026-01-21 16:45:50 +08:00
Luis Pater
ef4508dbc8
refactor(cache, translator): remove session ID from signature caching and clean up logic
2026-01-21 13:37:10 +08:00
Luis Pater
f775e46fe2
refactor(translator): remove session ID logic from signature caching and associated tests
2026-01-21 12:45:07 +08:00
Luis Pater
65ad5c0c9d
refactor(cache): simplify signature caching by removing sessionID parameter
2026-01-21 12:38:05 +08:00
Luis Pater
88bf4e77ec
fix(translator): update HasValidSignature to require modelName parameter for improved validation
2026-01-21 11:31:37 +08:00
hkfires
9332316383
fix(translator): preserve thinking blocks by skipping signature
2026-01-21 10:49:20 +08:00
Luis Pater
2e14f787d4
feat(translator): enhance ConvertGeminiRequestToAntigravity with model name and refine reasoning block handling
2026-01-21 08:31:23 +08:00
Luis Pater
523b41ccd2
test(responses): add comprehensive tests for SSE event ordering and response transformations
2026-01-21 07:08:59 +08:00
N1GHT
09970dc7af
Accept Geminis Review Suggestion
2026-01-20 17:51:36 +01:00
N1GHT
d81abd401c
Returned the Code Comment I trashed
2026-01-20 17:36:27 +01:00
N1GHT
a6cba25bc1
Small fix to filter out Top_P when Temperature is set on Claude to make requests go through
2026-01-20 17:34:26 +01:00
hkfires
9b72ea9efa
fix(translator): enhance signature cache clearing logic and update test cases with model name
2026-01-20 20:02:29 +08:00
Luis Pater
e49a1c07bf
chore(translator): update cache functions to include model name parameter in tests
2026-01-20 18:36:51 +08:00
Luis Pater
c2bf600a39
Merge branch 'main' into plus
2026-01-20 13:50:41 +08:00
Luis Pater
8d9f4edf9b
feat(translator): unify model group references by introducing GetModelGroup helper function
2026-01-20 13:45:25 +08:00
Luis Pater
020e61d0da
feat(translator): improve signature handling by associating with model name in cache functions
2026-01-20 13:31:36 +08:00
Luis Pater
6184c43319
Fixed : #1109
...
feat(translator): enhance session ID derivation with user_id parsing in Claude
2026-01-20 12:35:40 +08:00
Luis Pater
2cbe4a790c
chore(translator): remove unnecessary whitespace in gemini_openai_response code
2026-01-20 11:47:33 +08:00
Luis Pater
68b3565d7b
Merge branch 'main' into dev (PR #961 )
2026-01-20 11:42:22 +08:00
Luis Pater
d4bb4e6624
refactor(antigravity): remove unused client signature handling in thinking objects
2026-01-20 10:17:55 +08:00
Luis Pater
a7ffc77e3d
Merge branch 'dev' into fix/cross-model-thinking-signature
2026-01-20 10:10:43 +08:00
Luis Pater
08779cc8a8
Merge branch 'router-for-me:main' into main
2026-01-19 21:00:58 +08:00
hkfires
52e46ced1b
fix(translator): avoid forcing RFC 8259 system prompt
2026-01-19 11:33:27 +08:00
hkfires
cf9daf470c
feat(translator): report cached token usage in Claude output
2026-01-19 11:23:44 +08:00
Luis Pater
2d9f6c104c
Merge branch 'main' into plus
2026-01-19 04:03:17 +08:00
Luis Pater
140d6211cc
feat(translator): add reasoning state tracking and improve reasoning summary handling
...
- Introduced `oaiToResponsesStateReasoning` to track reasoning data.
- Enhanced logic for emitting reasoning summary events and managing state transitions.
- Updated output generation to handle multiple reasoning entries consistently.
2026-01-19 03:58:28 +08:00
hkfires
d5ef4a6d15
refactor(translator): remove registry model lookups from thinking config conversions
2026-01-18 10:30:14 +08:00
Luis Pater
46433a25f8
fix(translator): add check for empty text to prevent invalid serialization in gemini and antigravity
2026-01-18 00:50:10 +08:00
Luis Pater
f8f3ad84fc
Fixed : #1064
...
feat(translator): improve system message handling and content indexing across translators
- Updated logic for processing system messages in `claude`, `gemini`, `gemini-cli`, and `antigravity` translators.
- Introduced indexing for `systemInstruction.parts` to ensure proper ordering and handling of multi-part content.
- Added safeguards for accurate content transformation and serialization.
2026-01-17 05:40:56 +08:00
Luis Pater
93d7883513
Merge pull request #110 from PancakeZik/fix/system-prompt-reinjection
...
fix: prevent system prompt re-injection on subsequent turns
2026-01-17 05:19:11 +08:00
Luis Pater
015a3e8a83
Merge branch 'router-for-me:main' into main
2026-01-17 05:17:38 +08:00
Joao
6b074653f2
fix: prevent system prompt re-injection on subsequent turns
...
When tool results are sent back to the model, the system prompt was being
re-injected into the user message content, causing the model to think the
user had pasted the system prompt again. This was especially noticeable
after multiple tool uses.
The fix checks if there is conversation history (len(history) > 0). If so,
it's a subsequent turn and we skip system prompt injection. The system
prompt is only injected on the first turn (len(history) == 0).
This ensures:
- First turn: system prompt is injected
- Tool result turns: system prompt is NOT re-injected
- New conversations: system prompt is injected fresh
2026-01-16 20:16:44 +00:00
Luis Pater
65b4e1ec6c
feat(codex): enable instruction toggling and update role terminology
...
- Added conditional logic for Codex instruction injection based on configuration.
- Updated role terminology from "user" to "developer" for better alignment with context.
2026-01-17 04:12:29 +08:00
Luis Pater
06afa29f2d
Merge branch 'router-for-me:main' into main
2026-01-16 20:01:35 +08:00
Luis Pater
6600d58ba2
feat(codex): enhance input transformation and remove unused safety_identifier field
...
- Added logic to transform `inputResults` into structured JSON for improved processing.
- Removed redundant `safety_identifier` field in executor payload to streamline requests.
2026-01-16 19:59:01 +08:00
Luis Pater
bca244df67
Merge branch 'router-for-me:main' into main
2026-01-16 11:37:33 +08:00
Luis Pater
cec4e251bd
feat(translator): preserve text field in serialized output during chat completions processing
2026-01-16 11:35:34 +08:00
Luis Pater
c29839d2ed
Merge remote-tracking branch 'origin/main' into pr-104
...
# Conflicts:
# config.example.yaml
# internal/config/config.go
# sdk/cliproxy/auth/model_name_mappings.go
2026-01-16 09:40:07 +08:00
hkfires
199cf480b0
refactor(thinking): remove support for non-standard thinking configurations
...
This change removes the translation logic for several non-standard, proprietary extensions used to configure thinking/reasoning. Specifically, support for `extra_body.google.thinking_config` and the Anthropic-style `thinking` object has been dropped from the OpenAI request translators.
This simplification streamlines the translators, focusing them on the standard `reasoning_effort` parameter. It also removes the need to look up model information from the registry within these components.
BREAKING CHANGE: Support for non-standard thinking configurations via `extra_body.google.thinking_config` and the Anthropic-style `thinking` object has been removed. Clients should now use the standard `reasoning_effort` parameter to control reasoning.
2026-01-15 19:32:12 +08:00
hkfires
5a77b7728e
refactor(thinking): improve budget clamping and logging with provider/model context
2026-01-15 13:06:41 +08:00
hkfires
ed8b0f25ee
fix(thinking): use LookupModelInfo for model data
2026-01-15 13:06:41 +08:00
