Commit graph

53252 commits

Author SHA1 Message Date
Mateo Wang
b6bdc0dfc8 fix(bedrock): use native structured output on Haiku 5.5 and Sonnet 5.5, and fix Fable 5 and Sonnet 4.6 schema 400s (#46870)
* fix(bedrock): use native structured output on every Claude model whose Bedrock profiles accept it

Bedrock accepts outputConfig structured output on every reachable profile of
Claude Opus 4.7, 4.8, 5, 5.5, Sonnet 5, 5.5, Haiku 5.5, and Fable 5.1, so
json_schema now goes out natively there instead of as a forced tool, which kept
Bedrock from returning adaptive thinking. Fable 5 rejects outputConfig, so its
keys go back to the tool path. GovCloud keys keep their flags, and the drift
tests let GovCloud lag its commercial base but never lead it.

* fix(bedrock): clean Claude schemas for native structured output and keep recursive schemas on the old path

Bedrock native output rejects minimum/maximum, minItems, uniqueItems, oneOf,
additionalProperties true, and recursive or unresolved $ref, all of which the
forced tool accepted. Claude schemas on the Bedrock native paths now go through
the cleaner the Anthropic native path already uses, and a schema with a
recursive or unresolved $ref keeps the forced tool, inline, or drop path it had.

* fix(bedrock): keep older installs on the json tool with a Bedrock-only native output key

Older installs fetch the model map from main, so flipping supports_native_structured_output
moved them onto native output with their old schema handling. The new
supports_bedrock_native_structured_output key is ignored by older code. Schemas native output
cannot hold (recursive refs, open objects such as dict fields) keep the json tool on the new
key, models on the old key keep their current behavior for those shapes, and the Anthropic
API, Vertex AI, and Azure keep their current behavior.

* chore(cost-map): regenerate the model prices schema for the Bedrock native output key

* fix(bedrock): keep native output to profiles every account accepts and route schemas Bedrock rejects to the json tool

Only Opus 5.5, Sonnet 5.5, Haiku 5.5 and Fable 5.1 keep the Bedrock native output key, since
Opus 4.8 and Sonnet 5 answer 400 on some accounts and Opus 5 garbles escaped characters.
Lookaround patterns and string formats outside Bedrock's list now keep the json tool, objects
typed as a union with null are closed, and the cleaner keeps minItems of 0 or 1.

* fix(bedrock): keep Opus 5.5 and Fable 5.1 on the json tool, since native output stops their adaptive thinking

* fix(bedrock): keep shared $defs in native schemas and send refused schemas to the json tool on every native model

Native output now keeps $defs instead of inlining each use, so a submodel reused across a schema counts once against Bedrock's optional and union limits. Schemas native output cannot hold use the json tool on Converse and Invoke chat for both native keys, and minItems is only kept when it is an integer 0 or 1

* fix(bedrock): keep the json tool unless thinking is requested and bound native output schemas

The Bedrock native output key now applies only when the call asks for thinking, so calls without
thinking keep the json tool as before. Schemas past Bedrock's documented union and optional counts,
or past a grammar size estimate fitted to live probes, keep the json tool on both map keys. Open
objects keep the json tool on the Anthropic API, Vertex AI and Azure AI, where native output returned
them empty. The $ref check no longer expands $defs, so doubling definitions no longer stall the request.

* fix(bedrock): resolve root-path $defs refs so nested Pydantic models keep native output on Invoke

* test(bedrock): cover native structured output and json tool routing through the proxy

Integration cells drive the proxy against a scripted Bedrock peer that enforces Bedrock's structured output schema rules, covering thinking with native output on Converse, Converse streaming, Invoke, /v1/responses and /v1/messages, the deployment opt-in, Fable 5, schema cleaning on every native entry, open objects, and malformed schemas. Vertex, Anthropic and Azure AI cells cover open objects on native-capable models and the cleaner's minItems and nullable object handling. The unit table now covers reasoning_effort none and disabled thinking

* fix(bedrock): keep native output for numeric formats and wide old-key schemas, and keep thinking in json mode

Check string formats only on nodes that can be strings, apply the 64-field cap only to the Bedrock-only key, carry thinking blocks into the json mode message, and send malformed json tool schemas as object schemas.

* fix(anthropic): keep native output for dict fields unless the json tool can be forced

* fix(anthropic): keep native output for dict fields when the caller sends tools

* test(bedrock): close the dangerous-tool-use test schema, since Bedrock native output requires closed objects

* test(anthropic): cover forced and unforced open-object routing, json mode thinking, numeric formats and wide schemas

Add proxy cells for open-object schemas that keep native output when the json tool cannot be forced (a model that refuses forced tools on the Anthropic API and Vertex AI, thinking, caller tools), for the thinking block a json mode answer keeps on Vertex AI and Bedrock Invoke, and for integer and number formats and 65-field schemas on each Bedrock map key. Point the forced open-object cells at the Anthropic API and Azure AI only, since no Vertex AI model in the cost map takes both native output and a forced tool

* fix(bedrock): keep the native output opt-in on the deployment that sets it, and route effort none open objects to native output

* test(bedrock): cover deployment-scoped native output opt-in, register_model and effort none open objects
2026-10-20 17:44:50 -07:00
Joshua Valluru
11ab893c26 fix(deps): retain the configured three-day release window 2026-10-20 17:08:09 -07:00
Joshua Valluru
6780fd85db test(mcp): import the live limit regression HTTP client 2026-10-20 16:59:25 -07:00
Joshua Valluru
e44fc8a94a test(cost): isolate custom rerank pricing from the main map 2026-10-20 16:56:29 -07:00
Joshua Valluru
73fb5bed32 fix(deps): match main serde_with security version 2026-10-20 16:33:09 -07:00
Joshua Valluru
abbae7d12b fix(deps): match main Next.js and sharp security versions 2026-10-20 16:31:26 -07:00
Joshua Valluru
7868743c52 fix(deps): align vulnerable Python packages with main 2026-10-20 16:08:26 -07:00
Joshua Valluru
006d4ee977 test(mcp): preserve configured SSE limits and recovery 2026-10-20 16:06:55 -07:00
Joshua Valluru
e03bac8dc1 chore(cost-map): mirror main at 4aec80f5ee 2026-10-20 16:05:56 -07:00
Joshua Valluru
f53eebf2ae fix(mcp): validate backported SSE event limits 2026-10-20 16:05:56 -07:00
Joshua Valluru
1684ded5a1 Merge remote-tracking branch 'origin/stable/1.104.x' into litellm_backport_mcp_sse_1_104 2026-10-20 16:05:56 -07:00
joshua-berri
5346a40c37
Merge pull request #47454 from BerriAI/litellm_restore_token_count_input_precedence
Some checks are pending
LiteLLM Rust / rust-lint (push) Waiting to run
LiteLLM Rust / rust-test (push) Waiting to run
LiteLLM Rust / rust-wheel (push) Waiting to run
fix(proxy): restore legacy token count input precedence
2026-10-19 17:59:47 -07:00
Joshua Valluru
e109f8977d fix(proxy): restore legacy token count input precedence 2026-10-19 17:50:09 -07:00
Mateo Wang
a63df167da
Merge pull request #47446 from BerriAI/litellm_backport_google_count_shapes_stable_1_104_x
fix(count_tokens): restore Google request compatibility on stable/1.104.x
2026-10-20 00:29:29 +00:00
mateo-berri
5e71466355 fix(count_tokens): normalize Google system parts and null tools 2026-10-19 16:53:38 -07:00
yuneng-jiang
5e4d160423
Merge pull request #47389 from BerriAI/litellm_backport_count_tokens_stable_1_104_x
fix(count_tokens): backport October 19 fixes to stable/1.104.x
2026-10-19 15:42:02 -07:00
mateo-berri
3f3fb2ac60 fix(count_tokens): omit local PDF parsing from stable backport 2026-10-19 15:20:52 -07:00
mateo-berri
704a15720a test(count_tokens): type PDF budget regression cases 2026-10-19 15:14:43 -07:00
mateo-berri
c881f7cfc1 fix(count_tokens): bound PDF parsing and preserve legacy fallback 2026-10-19 15:10:51 -07:00
mateo-berri
498095ee1b fix(count_tokens): retain legacy translated content fallback 2026-10-19 14:38:30 -07:00
mateo-berri
e9e7e32142 test(count_tokens): make strict error envelope coverage explicit 2026-10-19 14:30:57 -07:00
mateo-berri
d522e2e09b fix(count_tokens): preserve legacy provider error fallback on stable
Defer the breaking error policy from #46121. Adapt the #46777 count settings flow to attempt a labelled local estimate for provider failures across providers and configured count models.

Keep provider errors when local counting is disabled or cannot estimate the request, and test the shared behavior across HTTP error statuses.
2026-10-19 14:28:17 -07:00
mateo-berri
e19f8fd173 ci(cost map): regenerate the schema for the stable line model map 2026-10-19 14:08:16 -07:00
mateo-berri
2ddc8946d0 fix(count_tokens): complete the selected count settings prerequisites
Preserve the prompt normalization required by #46777 from #46839 and classify its token_count_tokenizer key using the upstream schema definition.
2026-10-19 14:07:15 -07:00
mateo-berri
828ddf0c37 fix(count_tokens): narrow stable backport to the October 19 patch
Keep #43900, #45301, #46121, #46784, #46759, #46767, #46772, #46815, #47057 and #46777. Port only the Gemini prerequisites from #46430.

Remove the deferred Bedrock Mantle behavior and unrelated picks from the final tree. Restore the PDF test provider fixture and deployment, and retain required test imports.
2026-10-19 14:03:42 -07:00
mateo-berri
94828be8b0 test(integration): add the openai_wire helper the backported chat template wire test imports
Hand-ported verbatim to stable/1.104.x from main.
2026-10-19 12:35:34 -07:00
Mateo Wang
c3d07e70d6 fix(count_tokens)!: count tools the way DeepSeek V4 and Qwen3.5 chat templates render them (#47173)
Backport of #47173 to stable/1.104.x.
Cherry-picked from c82e8244a5 (main).
2026-10-19 12:35:33 -07:00
Mateo Wang
b6f1a033de feat(count_tokens)!: configurable token_count_framing and token_count_model, count source in the response (#46777)
Backport of #46777 to stable/1.104.x.
Cherry-picked from 75faaeab64 (main).
2026-10-19 12:35:33 -07:00
mateo-berri
52de3b0a29 test(integration): add the support helpers the backported count_tokens wire tests import
Hand-ported to stable/1.104.x from main (bedrock_runtime_peer, provider, _cache_control_marks_support verbatim, graceful_stop_seconds at the line's 30s stop wait).
2026-10-19 12:35:32 -07:00
Mateo Wang
748e03088d fix(count_tokens)!: answer 400 for Vertex partner models without a count API and OpenAI-shaped provider errors on /v1/responses/input_tokens (#47163)
Backport of #47163 to stable/1.104.x.
Cherry-picked from ef1d3959a0 (main).
2026-10-19 12:35:32 -07:00
Mateo Wang
da5a9ffd2e ci(cost map): remove the generated model prices JSON schema (#47064)
Backport of #47064 to stable/1.104.x.
Cherry-picked from ae6f7f6613 (main).

Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-10-19 12:35:32 -07:00
Mateo Wang
a4be1c1d13 ci(cost map): classify token_count_model as an object key in the schema generator (#47057)
Backport of #47057 to stable/1.104.x.
Cherry-picked from 6e191a58f1 (main).
2026-10-19 12:35:32 -07:00
Mateo Wang
b1037d2121 fix(proxy)!: count prompt-only token_counter requests with the provider when call_endpoint is set (#46839)
Backport of #46839 to stable/1.104.x.
Cherry-picked from 2dc6aafa6b (main).
2026-10-19 12:35:32 -07:00
Mateo Wang
4d5a6577c2 fix(proxy)!: count endpoints return errors in their source API shape instead of silent local counts (#46815)
Backport of #46815 to stable/1.104.x.
Cherry-picked from b21e659c24 (main).

Co-authored-by: Mrinal Chanshetty <mrinal@berri.ai>
2026-10-19 12:35:32 -07:00
Mateo Wang
13b4537c81 fix(count_tokens): return a 400 naming Anthropic blocks the OpenAI and Gemini count can't convert (#46772)
Backport of #46772 to stable/1.104.x.
Cherry-picked from 79ca820933 (main).

Co-authored-by: Mrinal Chanshetty <mrinal@berri.ai>
2026-10-19 12:35:32 -07:00
Mateo Wang
350def8ccf fix(count_tokens): count Anthropic-format requests for Gemini the way /v1/messages inference converts them (#46767)
Backport of #46767 to stable/1.104.x.
Cherry-picked from dba86bec28 (main).

Co-authored-by: Mrinal Chanshetty <mrinal@berri.ai>
2026-10-19 12:35:32 -07:00
Mateo Wang
531b66719a fix(count_tokens): count Anthropic-format requests for OpenAI the way /v1/messages inference converts them (#46759)
Backport of #46759 to stable/1.104.x.
Cherry-picked from d25b58f744 (main).

Co-authored-by: Mrinal Chanshetty <mrinal@berri.ai>
2026-10-19 12:35:32 -07:00
Mateo Wang
c99e4b86db ci(cost map): classify token_count_framing and token_count_fallback_model in the schema generator (#46784)
Backport of #46784 to stable/1.104.x.
Cherry-picked from c6a4437e98 (main).
2026-10-19 12:20:03 -07:00
mrinal-berri
fe9e6381a3 fix(count_tokens): count system and tools on Vertex Gemini, fall back to Mantle on Bedrock 404 (#46430)
Backport of #46430 to stable/1.104.x.
Cherry-picked from 08a24d391a (main).

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-10-19 12:20:03 -07:00
Krrish Dholakia
c703543b30 fix(proxy): raise on provider count_tokens errors instead of falling back to tiktoken (#46121)
Backport of #46121 to stable/1.104.x.
Cherry-picked from 719c44db3f (main).

Co-authored-by: Krrish Dholakia <krrishdholakia@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: yuneng-jiang <yuneng@berri.ai>
2026-10-19 12:20:03 -07:00
devin-ai-integration[bot]
095c4eaa41 fix(anthropic): count a leading system run through count_tokens' system parameter (#45463)
Backport of #45463 to stable/1.104.x.
Cherry-picked from d34ad35281 (main).

Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
2026-10-19 12:20:03 -07:00
devin-ai-integration[bot]
0b148c95ee fix(token_counter): price a base64 PDF document per page instead of as one image (#45301)
Backport of #45301 to stable/1.104.x.
Cherry-picked from 50ab9a3211 (main).

Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
2026-10-19 12:20:03 -07:00
devin-ai-integration[bot]
8222299f1d fix(bedrock): count tokens on bedrock-mantle when bedrock-runtime cannot count a Claude model (#45317)
Backport of #45317 to stable/1.104.x.
Cherry-picked from 577d74c1ce (main).
2026-10-19 12:20:03 -07:00
devin-ai-integration[bot]
3da404a7c1 fix(bedrock_mantle): add is_mantle_claude_model for the count_tokens picks
Hand-port of the is_mantle_claude_model helper from #43646 to stable/1.104.x, which #45317 imports.
Taken from 01b4ffe16b (main).
2026-10-19 11:36:30 -07:00
tin-berri
2f968f566b fix(anthropic): normalize images for provider token counting (#45185)
Backport of #45185 to stable/1.104.x.
Cherry-picked from 6befb9ad7d (main).
2026-10-19 11:36:30 -07:00
devin-ai-integration[bot]
a6a1956a17 fix(vertex_ai): forward system and tools to partner model count_tokens (#43900)
Backport of #43900 to stable/1.104.x.
Cherry-picked from 6d73fa6b49 (main).

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: jesus <jesus@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
2026-10-19 11:36:30 -07:00
devin-ai-integration[bot]
9d0cca7ed0 fix(token_counter): count Gemini function_declarations tools (#43417)
Backport of #43417 to stable/1.104.x.
Cherry-picked from 8e6d99d74a (main).

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-10-19 11:36:30 -07:00
Mateo Wang
b40bd7bf03
Merge pull request #47344 from BerriAI/litellm_backport_47067_stable_1_104_x
fix(proxy): backport #47067 team-scoped batch model access to stable/1.104.x for v1.104.4
2026-10-19 18:29:05 +00:00
mateo-berri
c47b760253 test(proxy): import Final in the line's batch endpoint tests
The tests picked from #47067 annotate module constants with Final, which main's copy imports and this line's did not, so the module failed to import on Python 3.12.
2026-10-19 09:10:10 -07:00
mateo-berri
d11f438e3d bump: version 1.104.4 2026-10-19 08:58:56 -07:00