scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-09-17 23:51:30 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 67
49820 commits 2515 branches 1410 tags 3 GiB
Commit graph

49820 commits

This branch
This branch
All branches
Author SHA1 Message Date
kerry
954dfa6ba7 test(utils): allow modality batch cost fields in cost map schema test 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 02:38:53 +00:00
kerry
7761d04450 test(vertex): cover malformed batch usage details 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 02:37:51 +00:00
Devin AI
123b76e2d8 fix(health): annotate credential resolution casts 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 02:34:49 +00:00
mateo-berri
8b80971586 Merge remote-tracking branch 'origin/main' into litellm_anthropic_guardrail_system_and_tool_use 2026-09-14 19:27:58 -07:00
kerry
168d0d3d99 fix(cost): drop remaining unnecessary cast in batch cost calculator 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 02:26:19 +00:00
mateo-berri
0a0a9509dc test(bedrock): type the session policy test helpers with the policy TypedDict 2026-09-14 19:21:23 -07:00
mateo-berri
22b377fe2a fix(proxy): log the provider usage on deferred /v1/messages calls and price cache writes without a creation rate 
With a post-call guardrail the proxy defers async success logging, and every nested wrapper on a
/v1/messages call bridged to the Responses API overwrote the stored closure, so the spend log was
built from the outermost Anthropic-shaped reply under Responses semantics and recorded the prompt
tokens without the cache hit. The first wrapper to exit now keeps the slot, which is the innermost
provider response, the same one the non-deferred path logs.

The flat cost path also billed cache-creation tokens at 0 when the model had no
cache_creation_input_token_cost. It now falls back to the input rate, and the 1h rate to the
creation rate, matching the tiered path and the custom pricing helper.
 2026-09-14 19:21:03 -07:00
Devin AI
7255a201a3 fix(health): resolve stored credentials for realtime checks 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 02:19:15 +00:00
kerry
a5f00b9189 fix(cost): drop unnecessary cast in batch rate lookup 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 02:15:22 +00:00
yassin
b9dd397746 fix(prometheus): count 401 auth failures in litellm_proxy_failed_requests_metric 
Invalid or unknown virtual keys were filtered out of the proxy failed and total
request counters entirely. Count them with hashed_api_key unset so caller-chosen
key strings cannot create unbounded label series, and normalize the request
route on the auth failure path so dynamic path ids do not leak into the route
label either.

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 02:09:00 +00:00
kerry
bf1bdb3045 fix(ci): keep cost map schema generated by the base branch generator 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 02:07:44 +00:00
mateo-berri
8d5d5d5824 fix(bedrock): grant rerank, retrieve, agent, and agentcore actions in the web identity session policy 2026-09-14 19:03:55 -07:00
kerry
a5fc880c90 refactor(vertex): move batch usage modality parsing under llms 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 02:03:32 +00:00
kerry
ca7364fb05 fix(batches): avoid strict lint violation in usage parser 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 02:00:48 +00:00
kerry
c0c5044c45 fix(batches): keep modality token details in raw vertex batch usage 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 01:57:50 +00:00
kerry
2b32f586c0 refactor(cost): extract batch modality rate lookup 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 01:51:41 +00:00
kerry
ce83fac351 fix(cost): bill batch embeddings per modality token rate 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 01:50:21 +00:00
mateo-berri
16c326537f fix(guardrails): define UnappliableRequestRewrite in the shared guardrail translation utils 
The three guardrail translation handlers imported the exception from the proxy policy engine through a function-local import, which CodeQL flagged as a cyclic import. The exception and its helper now live next to the handlers in the shared guardrail translation utils, and the tests import it from there.

The Prompt Security modify-mode helper is also restructured into early-return TypedDict displays so the LIT002 budget stays at its limit
 2026-09-14 18:39:54 -07:00
Mateo Wang
d2859e18d7
Merge pull request #40988 from BerriAI/litellm_responses_stream_error_exception_mapping 
fix(responses): route mid-stream error events through exception_type so content_policy_fallbacks fire
 2026-09-14 18:36:20 -07:00
kerry
a28ea22ec1 fix(cost): move gemini-embedding-2-preview to per-token rates 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 01:34:22 +00:00
kerry
27a486e4d3 test(cost): cover modality guards and image detection fallbacks 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 01:34:17 +00:00
Mateo Wang
1ba97665b2
Merge pull request #41046 from BerriAI/devin/1789368982-cli-strenum-py310 
fix(cli): drop enum.StrEnum so the CLI imports on Python 3.10
 2026-09-14 18:32:30 -07:00
mateo-berri
931bdb8c0b fix(compression): protect part-level cache_control rows in compress() too 
compress() scores text-only copies of the rows, so a content-part cache_control marker was gone by the time get_protected_indices ran and the pinned row could still be stubbed. Read protection from the original rows, which are index-aligned with the normalized copies, and add a regression test that fails without the change.
 2026-09-14 18:19:51 -07:00
mateo-berri
6574d83eae Merge remote-tracking branch 'origin/main' into HEAD 
# Conflicts:
#	litellm/proxy/guardrails/guardrail_hooks/prompt_security/prompt_security.py
 2026-09-14 18:19:38 -07:00
kerry
4a8ec7b9d8 fix(cost): bill gemini-embedding-2-preview per token like the GA entries 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 01:18:50 +00:00
kerry
6a18105275 fix(vertex): only bill image rate without modality details when every input is an image 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 01:16:00 +00:00
jesus
e390dfbb64 fix(headroom): format protected index assignment 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 01:13:17 +00:00
kerry
6cbed7b4c0 fix(vertex): drop Final image_tokens redeclaration flagged by basedpyright 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 01:10:27 +00:00
jesus
9c59feee7c fix(headroom): protect the cached prefix through the last cache_control breakpoint 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 01:07:41 +00:00
kerry
ac8e1a355c test(vertex): load local pricing in embedding billing tests 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 01:00:32 +00:00
Mateo Wang
dad6e9be4a
Merge branch 'main' into fix/headroom-protect-cache-control-rows 2026-09-14 17:58:17 -07:00
yucheng-berri
8481bc27f9
Merge pull request #41132 from BerriAI/litellm_bedrock_grounding_from_plain_messages 
fix(bedrock guardrails): derive contextual grounding source and query from plain messages
 2026-09-14 17:56:58 -07:00
kerry
e31c64d2e0 fix(schema): sync model price schema with cost map 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 00:54:58 +00:00
kerry
e5845c17ff fix(vertex): bill image inputs at the image rate when usage lacks modality details 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 00:51:49 +00:00
yassin
1943667fef fix(router): run sync pre-call checks when a model_name collides with a deployment id 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 00:51:13 +00:00
kerry
6c9fe65608 style(cost): apply ruff formatting to modality guards 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 00:47:30 +00:00
kerry
0c91d9157c refactor(vertex): drop unused resolved_files from embed response parsing 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 00:45:08 +00:00
kerry
e26a4970dd fix(test): complete synthetic model metadata 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 00:42:21 +00:00
yassin
bdc63d590e fix(router): keep weighted routing when a deployment id equals a model_name 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 00:42:18 +00:00
kerry
d4f2119b03 fix(cost): bill gemini-embedding-2 per token and stop double charging audio 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 00:41:46 +00:00
yucheng
0d03ea3154 Merge remote-tracking branch 'origin/main' into litellm_agent365_mcp_guardrail
Some checks failed
LiteLLM Rust / rust-lint (push) Has been cancelled
Details
ai-gateway image / ai-gateway release image (push) Has been cancelled
Details
LiteLLM Rust / rust-test (push) Has been cancelled
Details
2026-09-15 00:30:12 +00:00
yucheng
4bfaf74edc refactor(guardrails): keep Agent 365 PR to the guardrail, move MCP changes to stacked PRs 
The listed-tool metadata and per-caller catalog cache move to a follow-up PR
stacked on this one, and the RFC 9728 sign-in challenge and discovery metadata
move to a second PR stacked on that. On its own the guardrail evaluates the
tool name and arguments and the caller presents its Entra token

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 00:29:58 +00:00
Yuneng Jiang
37bde0bdbe
test: keep provider request snapshots immutable 2026-09-14 17:27:18 -07:00
yucheng
b030164f73 Merge remote-tracking branch 'origin/main' into litellm_agent365_mcp_guardrail 2026-09-15 00:26:03 +00:00
yucheng-berri
b67137b67f
Merge pull request #40702 from BerriAI/litellm_model_armor_logging_only_streaming 
feat(model_armor): logging_only mode scans completed streams after delivery
 2026-09-14 17:23:32 -07:00
Yuneng Jiang
51db940514
test: relocate strict replay harness coverage 2026-09-14 17:20:58 -07:00
Yuneng Jiang
b9c194076a
test: preserve strict replay numeric spelling 2026-09-14 17:11:56 -07:00
ryan
98ea14758f fix(proxy): only widen an id-only bulk member delete to its email when the user lists the team 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 00:09:57 +00:00
yucheng
5ceec4c21e fix(guardrails): cite message scoping only when an unscoped pass finds content 
The not_run reason now says after message scoping only when the same messages carry text or tool calls without the skip flags applied. A request that is empty to begin with, whatever the flags, records no scannable content

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-15 00:03:02 +00:00
yucheng
52da64a45b fix(guardrails): only cite message scoping in the not_run reason when scoping is on 
A request whose messages carry no scannable content at all, with no skip flag set, now records the neutral reason no scannable content instead of blaming configuration

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-14 23:56:32 +00:00