phoenix/litellm
1
0
Fork 0
forked from phoenix/litellm-mirror
Code Pull requests Activity
15937 commits 1603 branches 645 tags 320 MiB
Commit graph

15937 commits

This branch
This branch
All branches
Author SHA1 Message Date
Krish Dholakia
5f13d2ee64
Merge pull request #5029 from BerriAI/litellm_azure_ui_fix 
fix(utils.py): Fix adding azure models on ui
 2024-08-02 22:12:19 -07:00
Ishaan Jaff
057bb37a39 test vertex ft jobs 2024-08-02 22:03:52 -07:00
Joe Cheng
1fbfc09b44 Add unit test 2024-08-02 20:51:08 -07:00
Ishaan Jaff
0463e73ee3 fix typing 2024-08-02 18:49:16 -07:00
Ishaan Jaff
f194aa3a93 fix typing 2024-08-02 18:46:43 -07:00
Ishaan Jaff
4fc27e87c5 add vertex ai ft on proxy 2024-08-02 18:26:36 -07:00
Ishaan Jaff
40430dde10 add support for sync vertex ft 2024-08-02 18:16:14 -07:00
Ishaan Jaff
db3b2a3ee0 fix translating response 2024-08-02 18:13:59 -07:00
Ishaan Jaff
0e70aca776 convert response obj from vertex ai 2024-08-02 18:13:25 -07:00
Ishaan Jaff
960de8a3d0 test ft response vertex ai 2024-08-02 18:02:34 -07:00
Ishaan Jaff
ac6224c2b1 translate response from vertex to openai 2024-08-02 18:02:24 -07:00
Krrish Dholakia
5d96ff6694 fix(utils.py): handle scenario where model="azure/*" and custom_llm_provider="azure" 
Fixes https://github.com/BerriAI/litellm/issues/4912
 2024-08-02 17:48:53 -07:00
Ishaan Jaff
7ec1f241fc
Merge pull request #5026 from BerriAI/litellm_fix_whisper_caching 
[Fix] Whisper Caching - Use correct cache keys for checking request in cache
 2024-08-02 17:26:28 -07:00
Ishaan Jaff
d7ab440280 example vertex ai.jsonl 2024-08-02 17:25:32 -07:00
Ishaan Jaff
cb8a592ccf test vertex ft 2024-08-02 17:25:18 -07:00
Ishaan Jaff
693ab3759c add vertex ft support 2024-08-02 17:25:03 -07:00
Ishaan Jaff
d364d76bd0 add vertex FT spec 2024-08-02 17:24:25 -07:00
Ishaan Jaff
653cf55850 add fine tuning for vertex 2024-08-02 17:23:58 -07:00
Krish Dholakia
3ee8ae231c
Merge pull request #5027 from BerriAI/litellm_admin_viewer_team_tab 
build(ui): allow admin_viewer to view teams tab
 2024-08-02 17:18:48 -07:00
Ishaan Jaff
cff7050147 add vertex_credentials in router param 2024-08-02 16:58:17 -07:00
Krrish Dholakia
186e5f55e8 fix(user_api_key_auth.py): fix linting errors 2024-08-02 16:34:32 -07:00
Krish Dholakia
3d4ee61358
Merge pull request #5019 from BerriAI/litellm_deepseek_caching 
fix(types/utils.py): Support deepseek prompt caching
 2024-08-02 16:26:49 -07:00
Krrish Dholakia
1080c7014e build(ui): allow admin_viewer to view teams tab 
Allows admin viewe role to see available teams on proxy ui
 2024-08-02 16:19:40 -07:00
Ishaan Jaff
7fa25c443a docs caching 2024-08-02 15:19:27 -07:00
Ishaan Jaff
aadc29aab8 docs call types 2024-08-02 15:17:09 -07:00
Ishaan Jaff
2fd523a0f9 docs supported call types 2024-08-02 15:14:38 -07:00
Ishaan Jaff
87a15b408f fix test 2024-08-02 15:08:44 -07:00
Ishaan Jaff
d6c3560ff1 test whisper 2024-08-02 15:07:38 -07:00
Ishaan Jaff
ec3b0d0d0b return cache hit True on cache hits 2024-08-02 15:07:05 -07:00
Ishaan Jaff
b6b19dc128 use file name when getting cache key 2024-08-02 14:52:08 -07:00
Ishaan Jaff
1b3bc32090 log correct file name on langfuse 2024-08-02 14:49:25 -07:00
Joe Cheng
33f4411f17 Fix tool call coalescing 
The previous code seemed to assume that the tool call index property
started at 0, but Anthropic sometimes returns them starting at 1.
This was causing an extra null-ish tool call to be materialized.
 2024-08-02 13:05:23 -07:00
Joe Cheng
90dd60fa71 fix(main.py): Handle bedrock tool calling in stream_chunk_builder 
Fixes #5022.

The streaming chunks from Anthropic seem to violate an assumption
that is implicit in the stream_chunk_builder implementation: that
only tool_calls OR function_calls OR content will appear in a
streamed response. The repro in #5022 shows that you can get
content followed by tool calls.

These changes properly handle these combinations by building
separate lists of each type of chunk (note that in theory a chunk
could appear in multiple lists, e.g. both delta.tool_calls and
delta.content being present on one chunk).
 2024-08-02 12:41:13 -07:00
Ishaan Jaff
7f1d32a75b use regular ci/cd pipeline 2024-08-02 12:13:15 -07:00
Ishaan Jaff
fc8a87efec ci/cd run again 2024-08-02 11:46:45 -07:00
Ishaan Jaff
ffa7818339 fix config.yaml 2024-08-02 11:41:53 -07:00
Ishaan Jaff
459964decf fix config 2024-08-02 11:39:54 -07:00
Ishaan Jaff
53d57509ab bump: version 1.42.10 → 1.42.11 2024-08-02 11:29:56 -07:00
Ishaan Jaff
d42549d6e4 temp testing ci/cd 2024-08-02 11:29:41 -07:00
Ishaan Jaff
b3196df0b8 queue stable release testing after new GH release 2024-08-02 10:57:56 -07:00
Krrish Dholakia
0ce6c038ee fix(anthropic.py): fix linting error 2024-08-02 10:37:53 -07:00
Krish Dholakia
70c45767c2
Merge pull request #5020 from BerriAI/litellm_openai_raw_response_headers 
refactor(openai/azure.py): move to returning openai/azure response headers by default
 2024-08-02 10:36:14 -07:00
Krrish Dholakia
fe2aa706e8 refactor(openai/azure.py): move to returning openai/azure response headers by default 
Allows token tracking to work more reliably across multiple azure/openai deployments
 2024-08-02 09:42:08 -07:00
Krrish Dholakia
0a30ba9674 fix(types/utils.py): support passing prompt cache usage stats in usage object 
Passes deepseek prompt caching values through to end user
 2024-08-02 09:30:50 -07:00
Krrish Dholakia
cd073d5ad3 test: handle anthropic rate limit error 2024-08-02 08:57:09 -07:00
Ishaan Jaff
48419ca8bc add step to ghcr deploy 2024-08-02 08:06:09 -07:00
Krrish Dholakia
fe7e68adc8 fix(utils.py): fix codestral streaming 2024-08-02 07:38:06 -07:00
Ishaan Jaff
75fba18c9f fix langfuse hardcoded public key 2024-08-02 07:21:02 -07:00
Krrish Dholakia
4c2ef8ea64 fix(bedrock_httpx.py): fix ai21 streaming 2024-08-01 22:03:24 -07:00
Krrish Dholakia
57e3044974 bump: version 1.42.9 → 1.42.10 2024-08-01 21:29:36 -07:00