fix(bedrock_mantle): normalize Codex input item types Mantle rejects by mateo-berri · Pull Request #38227 · BerriAI/litellm

mateo-berri · 2026-08-25T17:05:01Z

TLDR

Problem this solves:

Codex multi-agent sessions through Bedrock Mantle die on the first sub-agent turn
Mantle 400s the Codex history item types agent_message, context_compaction, and local_shell_call
Every later turn (auto-compaction included) replays the same history and dies too

How it solves it:

Rewrites agent_message into an assistant output_text message, keeping encrypted_content slot payloads
Renames context_compaction to the compaction spelling Mantle accepts, same ciphertext
Rewrites local_shell_call into the function_call its recorded output already pairs with

User Flow

Before: a Codex CLI user on a gpt-5.6 model served via Bedrock Mantle spawns a sub-agent and the session hard-fails

They point Codex CLI at the gateway (base_url = "http://litellm-domain/v1", wire_api = "responses") with a gpt-5.6-sol deployment backed by Bedrock Mantle
They prompt Codex to spawn one sub-agent via the spawn_agent tool and the TUI shows "Started /root/arithmetic"
The sub-agent's first POST http://litellm-domain/v1/responses carries an agent_message history item and comes back 400 "invalid request body: Invalid 'input': value did not match any expected variant"
The parent turn dies with that 400 banner on screen, and every retry or follow-up turn fails the same way because the history still holds the item

After: the same session completes and the sub-agent's reply is relayed

They point Codex CLI at the gateway the same way, same gpt-5.6-sol deployment
The same prompt spawns the sub-agent and the TUI shows "Started /root/arithmetic"
Every POST http://litellm-domain/v1/responses returns 200: the sub-agent finishes and the parent relays its reply, "2 + 2 equals 4."
Follow-up turns over the same history keep working, and a replayed compaction summary comes back 200 with the pre-compaction context intact

Relevant issues

Fixes #36528. Addresses the input-item half of #36546 (its 400-masked-as-500 half is separate; in the runs below the 400 passes through to the client). Supersedes #36530, which rewrote agent_message into an assistant input_text message, a shape Mantle also rejects (verified live), and dropped encrypted_content payloads

Linear ticket

Resolves LIT-6093

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

I have added meaningful tests
The handful of test files covering my change pass locally, e.g. uv run pytest tests/test_litellm/<your_test_file>.py -v. Leave the suites (make test-unit-*, make test-unit) to CI: it finishes in ~15 minutes where a laptop takes an hour or more
My PR passes all required CI/CD checks (e.g., lint, schema.d.ts sync check, etc.)
My PR's scope is as isolated as possible; it only solves 1 specific problem
I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review (Greptile reviews automatically once the PR is opened; only comment @greptileai to re-request a review after pushing changes)

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

Screenshots / Proof of Fix

Commit fd1dca0 since the proven run is typing-only (annotations plus a pyright suppression, zero runtime change), so the ed28581 proof and /live-pr-risk pass below stand. Commit 636ae5d is a conflict-resolution merge of the base branch whose only manual resolution was uniting two import lines in the test file: the PR's diff against the base is unchanged and the transformation module is byte-identical, so the proof still stands

Shared setup, both legs identical except the commit the proxy was booted from. Proxy: .venv/bin/python litellm/proxy/proxy_cli.py --config lit6093.yaml --num_workers 2 (2 uvicorn workers; port 57013 before, 47608 after), config:

model_list:
  - model_name: gpt-5.6-sol
    litellm_params:
      model: bedrock_mantle/openai.gpt-5.6-sol
      aws_region_name: us-east-1
general_settings:
  master_key: sk-1234

Real Mantle in us-east-1, bearer token auth. Codex case: Codex CLI 0.149.1 TUI driven interactively, custom provider base_url = "http://localhost:<port>/v1", wire_api = "responses", model gpt-5.6-sol. The curl cases share the trailing user message {"type":"message","role":"user","content":[{"type":"input_text","text":"Reply with the single word pong."}]}, abbreviated <USER_MSG> below

Before (`bb27bfd`)

Codex CLI multi-agent session

TUI prompt: Spawn exactly one sub-agent with the spawn_agent tool. Its task: answer the question 'what is 2+2' in one sentence. Wait for its reply, then tell me the exact reply it sent you. Do not read or edit any files.
On screen, right after Started /root/arithmetic_reply and Waiting for agents:

■ {"error":{"message":"litellm.BadRequestError: BedrockException - {\"error\":{\"code\":\"validation_error\",\"message\":\"invalid request body: Invalid 'input': value did not match any expected
variant\",\"param\":null,\"type\":\"invalid_request_error\"}}. Received Model Group=gpt-5.6-sol\nAvailable Model Group Fallbacks=None","type":null,"param":null,"code":"400"}}

agent_message history item

curl -s -X POST http://localhost:57013/v1/responses -H 'Authorization: Bearer sk-1234' -H 'Content-Type: application/json' -d '{"model":"gpt-5.6-sol","max_output_tokens":20000,"input":[{"type":"agent_message","author":"worker_1","recipient":"user","content":[{"type":"input_text","text":"Message Type: MESSAGE\nTask name: worker_1\nSender: worker_1\nPayload:\nThe answer is 42."}]},<USER_MSG>]}'
HTTP 400: invalid request body: Invalid 'input': value did not match any expected variant

agent_message with encrypted_content slot

Same curl with content [{"type":"input_text","text":"Message Type: MESSAGE\n...Payload:\n"},{"type":"encrypted_content","encrypted_content":"gAAAAABopaque"}]
HTTP 400, same variant error

context_compaction carrying a genuine Mantle compaction summary

curl -s -X POST http://localhost:57013/v1/responses ... -d '{"model":"gpt-5.6-sol","max_output_tokens":20000,"input":[{"type":"context_compaction","encrypted_content":"smry_5ZVrif4J0bXIqmWdledj7QIFeOXasUEw18xiIaS6wkWRusNr..."},{"type":"message","role":"user","content":[{"type":"input_text","text":"What single word did I ask you to reply with earlier? Answer with just that word."}]}]}' (the ciphertext is one Mantle itself returned for a compaction_trigger over a "reply with pong" conversation)
HTTP 400, same variant error

local_shell_call history pair

curl -s -X POST http://localhost:57013/v1/responses ... -d '{"model":"gpt-5.6-sol","max_output_tokens":20000,"input":[{"type":"local_shell_call","call_id":"call_lsh1","status":"completed","action":{"type":"exec","command":["echo","hi"]}},{"type":"function_call_output","call_id":"call_lsh1","output":"hi\n"},<USER_MSG>]}'
HTTP 400, same variant error

/v1/chat/completions control (surface this PR does not touch)

curl -s -X POST http://localhost:57013/v1/chat/completions -H 'Authorization: Bearer sk-1234' -H 'Content-Type: application/json' -d '{"model":"gpt-5.6-sol","messages":[{"role":"user","content":"Reply with the single word pong."}]}'
HTTP 200, content pong

After (`ed28581`)

Codex CLI multi-agent session

Same TUI prompt against the port-47608 proxy
On screen: Started /root/arithmetic, then the parent's answer 2 + 2 equals 4. with no error banner
Follow-up prompt Did the sub-agent send you a reply message? Quote the exact payload text of the reply it sent you. (a second turn replaying the agent_message-laden history) answers "2 + 2 equals 4.", so the sub-agent's reply survived the rewrite and replayed histories keep working

agent_message history item

Same curl against port 47608
HTTP 200, status: completed, output_text pong

agent_message with encrypted_content slot

Same curl against port 47608
HTTP 200, status: completed, output_text pong

context_compaction carrying a genuine Mantle compaction summary

Same curl against port 47608
HTTP 200, output_text pong: the model recalls the pre-compaction instruction out of the summary, so the rewrite preserves it byte for byte

local_shell_call history pair

Same curl against port 47608
HTTP 200, status: completed, output_text pong

/v1/chat/completions control (surface this PR does not touch)

Same curl against port 47608
HTTP 200, content pong

Observations from the QA run:

Foreign/garbage compaction ciphertext still 400s Mantle-side (decrypt error); pre-existing, unfixable gateway-side
Mantle rejects assistant messages with input_text content; the rewrite must use output_text
agent_message encrypted_content slots carry verbatim plaintext through Mantle; this PR preserves them
function_call + encrypted_function_args and tool_search_call/output returned 200 live, unlike [Bug]: bedrock_mantle Responses: several Codex input-item types rejected wholesale, and the 400 is masked into an opaque 500 #36546's table

Type

🐛 Bug Fix

Caveats (if any)

context_compaction and local_shell_call appear only in migrated or resumed Codex histories
Codex sends agent_message on every multi-agent v2 session; that path is the customer hit

Final Attestation

The tests check the right things, including the edge cases, and regressions in the respective real-world customer use-cases are not possible after this PR
ed28581 passes /live-pr-risk

Mantle 400s ("Invalid 'input': value did not match any expected variant") on the Codex history item types agent_message, context_compaction, and local_shell_call, killing every Codex multi-agent session on the first sub-agent turn. Rewrite agent_message into an assistant output_text message (preserving encrypted_content slot payloads, which carry the plaintext task through Mantle), context_compaction into Mantle's supported compaction spelling, and local_shell_call into the function_call its recorded function_call_output already pairs with.

greptile-apps · 2026-08-25T17:07:57Z

Greptile Summary

This PR normalizes Codex history item types before forwarding Responses API requests to Bedrock Mantle.

Rewrites agent messages while preserving their textual payloads.
Renames compaction items and converts local shell calls into paired function calls.
Adds focused transformation tests for rewritten, dropped, passthrough, and logged cases.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains.

Important Files Changed

Filename	Overview
litellm/llms/bedrock_mantle/responses/transformation.py	Adds isolated Bedrock Mantle request-item normalization and resolves the previously reported `Any` return annotation.
tests/test_litellm/llms/bedrock_mantle/test_bedrock_mantle_responses_transformation.py	Adds regression coverage for all three normalized Codex item types, malformed-item dropping, passthrough behavior, and warning logs.

_{Reviews (2): Last reviewed commit: "fix(bedrock_mantle): type the codex item..." | Re-trigger Greptile}

greptile-apps · 2026-08-25T17:08:00Z

+        return rewritten
+
+    @classmethod
+    def _normalize_codex_input_item(cls, item: object) -> "tuple[Any, str | None]":


Normalization return type uses Any

The new tuple[Any, str | None] annotation prevents static verification of normalized items, weakening regression detection in this request transformation.

Context Used: CLAUDE.md (source)

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

veria-ai · 2026-08-25T17:10:02Z

+        if not isinstance(content, list):
+            return ""
+        return "".join(
+            str(block.get("text") or block.get("encrypted_content") or "")


Low: Input guardrail bypass

Proxy guardrails run before this provider transformation, and their Responses API text walkers inspect text fields but ignore encrypted_content. A caller can therefore place blocked plaintext in agent_message.content[].encrypted_content; it passes pre-call inspection here and is then converted into provider-visible output_text. Ensure this plaintext is inspected and, where applicable, rewritten by guardrails before normalization, or reject this shape when it cannot be inspected safely.

The shared text walker inspects only text parts, skipping function_call arguments; that gap is guardrail-layer, not this rewrite's. Rejecting the shape breaks the Codex flow.

Thanks for the detail. I can't automatically confirm this is safe to dismiss, so I'm leaving the thread open for a maintainer to make the call.

veria-ai · 2026-08-25T17:10:08Z

PR overview

This pull request updates the Bedrock Mantle response transformation to normalize Codex input item types that Mantle otherwise rejects. It converts supported content shapes into provider-compatible response items.

One security issue remains open: callers can place plaintext in an encrypted_content field that is skipped by pre-call text inspection but later converted into provider-visible text. This permits bypassing configured input guardrails for affected Bedrock Mantle requests, though the impact is limited to policy enforcement on this transformation path.

Open issues (1)

Low: Input guardrail bypass — litellm/llms/bedrock_mantle/responses/transformation.py:251

Fixed/addressed: 0 · PR risk: 5/10

codecov · 2026-08-25T17:15:44Z

Codecov Report

❌ Patch coverage is 97.22222% with 2 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...lm/llms/bedrock_mantle/responses/transformation.py	97.22%	2 Missing ⚠️

📢 Thoughts on this report? Let us know!

mateo-berri · 2026-08-25T17:20:50Z

@greptileai

mateo-berri · 2026-08-25T17:56:01Z

bugbot run

cursor

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit fd1dca0. Configure here.}

codspeed · 2026-08-25T18:05:32Z

Merging this PR will not alter performance

✅ 31 untouched benchmarks

_{Comparing litellm_mantle_codex_input_items (636ae5d) with litellm_internal_staging (bb27bfd)¹}

No successful run was found on litellm_internal_staging (636ae5d) during the generation of this report, so bb27bfd was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

tin-berri

Rewrite logic for agent_message/context_compaction/local_shell_call is well-scoped: each rewrite has a clear drop-if-empty guard (no call_id, no text, no ciphertext), encrypted_content payload text is preserved rather than discarded, and local_shell_call keeps the same call_id so it still pairs with its existing function_call_output. Supported item types pass through untouched (tested). Good coverage across plaintext/encrypted/drop/pass-through cases plus the warning log naming the rewritten types.

…itellm_mantle_codex_input_items

…00.0) (#124) This PR contains the following updates: | Package | Update | Change | |---|---|---| | [ghcr.io/berriai/litellm](https://images.chainguard.dev/directory/image/wolfi-base/overview) ([source](https://github.com/BerriAI/litellm)) | minor | `v1.99.1` → `v1.100.0` | --- ### Release Notes <details> <summary>BerriAI/litellm (ghcr.io/berriai/litellm)</summary> ### [`v1.100.0`](https://github.com/BerriAI/litellm/releases/tag/v1.100.0) [Compare Source](https://github.com/BerriAI/litellm/compare/v1.99.1...v1.100.0) #### Verify Docker Image Signature All LiteLLM Docker images are signed with [cosign](https://docs.sigstore.dev/cosign/overview/). Every release is signed with the same key introduced in [commit `0112e53`](https://github.com/BerriAI/litellm/commit/0112e53046018d726492c814b3644b7d376029d0). **Verify using the pinned commit hash (recommended):** A commit hash is cryptographically immutable, so this is the strongest way to ensure you are using the original signing key: ```bash cosign verify \ --key https://raw.githubusercontent.com/BerriAI/litellm/0112e53046018d726492c814b3644b7d376029d0/cosign.pub \ ghcr.io/berriai/litellm:v1.100.0 ``` **Verify using the release tag (convenience):** Tags are protected in this repository and resolve to the same key. This option is easier to read but relies on tag protection rules: ```bash cosign verify \ --key https://raw.githubusercontent.com/BerriAI/litellm/v1.100.0/cosign.pub \ ghcr.io/berriai/litellm:v1.100.0 ``` Expected output: ``` The following checks were performed on each of these signatures: - The cosign claims were validated - The signatures were verified against the specified public key ``` *** #### What's Changed - fix(responses): keep the conversation when chaining previous\_response\_id on the bridge by [@mateo-berri](https://github.com/mateo-berri) in [#37956](https://github.com/BerriAI/litellm/pull/37956) - feat(newrelic): per-team New Relic trace routing via team callbacks by [@yucheng-berri](https://github.com/yucheng-berri) in [#37603](https://github.com/BerriAI/litellm/pull/37603) - perf(ci): cache uv dependencies in the lint job by [@yuneng-berri](https://github.com/yuneng-berri) in [#37783](https://github.com/BerriAI/litellm/pull/37783) - perf(ci): fan the budget checkers out across cores by [@yuneng-berri](https://github.com/yuneng-berri) in [#37784](https://github.com/BerriAI/litellm/pull/37784) - ci: port the Postgres suites off CircleCI onto service containers by [@yuneng-berri](https://github.com/yuneng-berri) in [#37785](https://github.com/BerriAI/litellm/pull/37785) - feat(ci): gate patching of SDK internals in tests as TQ008 by [@yuneng-berri](https://github.com/yuneng-berri) in [#37787](https://github.com/BerriAI/litellm/pull/37787) - ci: measure enterprise/ coverage by [@yuneng-berri](https://github.com/yuneng-berri) in [#37788](https://github.com/BerriAI/litellm/pull/37788) - ci: run the keyless caching tests that ran in no job by [@yuneng-berri](https://github.com/yuneng-berri) in [#37790](https://github.com/BerriAI/litellm/pull/37790) - fix(ci): run the migration DDL guard, and stop it reading comments as SQL by [@yuneng-berri](https://github.com/yuneng-berri) in [#37791](https://github.com/BerriAI/litellm/pull/37791) - ci: run the enterprise package suite in GitHub Actions by [@yuneng-berri](https://github.com/yuneng-berri) in [#37798](https://github.com/BerriAI/litellm/pull/37798) - perf(ci): give the two longest unit shards the runner's spare cores by [@yuneng-berri](https://github.com/yuneng-berri) in [#37804](https://github.com/BerriAI/litellm/pull/37804) - test(exception-mapping): pin the status and error-shape table every provider maps to by [@yuneng-berri](https://github.com/yuneng-berri) in [#37807](https://github.com/BerriAI/litellm/pull/37807) - fix(terraform): add soft\_budget, tags, and soft\_budget\_alerting\_emails to litellm\_team by [@yuneng-berri](https://github.com/yuneng-berri) in [#37918](https://github.com/BerriAI/litellm/pull/37918) - fix(ui): theme the created-key box so it follows dark mode by [@yuneng-berri](https://github.com/yuneng-berri) in [#37985](https://github.com/BerriAI/litellm/pull/37985) - fix(ui): restore the public model name tooltip layout in the add model flow by [@yuneng-berri](https://github.com/yuneng-berri) in [#37986](https://github.com/BerriAI/litellm/pull/37986) - fix(ui): render team and org tpm/rpm limits of 0 as 0 instead of Unlimited by [@yuneng-berri](https://github.com/yuneng-berri) in [#37916](https://github.com/BerriAI/litellm/pull/37916) - fix(ui): repoint the key detail URL to the rotated hash after regenerating by [@yuneng-berri](https://github.com/yuneng-berri) in [#37968](https://github.com/BerriAI/litellm/pull/37968) - fix(ui): make playground chat bubbles theme-aware by [@yuneng-berri](https://github.com/yuneng-berri) in [#37978](https://github.com/BerriAI/litellm/pull/37978) - fix(UI): correct skill install command and marketplace setup UX by [@ozolam](https://github.com/ozolam) in [#33514](https://github.com/BerriAI/litellm/pull/33514) - fix(proxy): skip health checks for strategy routers by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#37966](https://github.com/BerriAI/litellm/pull/37966) - fix(databricks): bill cached tokens at cache rates and add missing Claude pricing by [@mateo-berri](https://github.com/mateo-berri) in [#37975](https://github.com/BerriAI/litellm/pull/37975) - fix(anthropic): round-trip thinking blocks to OpenAI backends on /v1/messages by [@mateo-berri](https://github.com/mateo-berri) in [#37953](https://github.com/BerriAI/litellm/pull/37953) - fix(a2a): normalize agent card protocolBinding casing before transport match by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#37917](https://github.com/BerriAI/litellm/pull/37917) - fix(interactions): track cost and spend for Google Interactions API requests by [@mateo-berri](https://github.com/mateo-berri) in [#33310](https://github.com/BerriAI/litellm/pull/33310) - fix(bedrock): stop emitting an empty assistant delta after the finish\_reason chunk by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#36806](https://github.com/BerriAI/litellm/pull/36806) - fix(anthropic): reconcile enum with declared type in output\_format schema by [@dkindlund](https://github.com/dkindlund) in [#37882](https://github.com/BerriAI/litellm/pull/37882) - feat(azure\_ai): support entra id / oauth auth on every azure ai foundry route by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#35415](https://github.com/BerriAI/litellm/pull/35415) - fix(ui): boot the UI image as an arbitrary uid by anchoring nginx writes under /tmp by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#37982](https://github.com/BerriAI/litellm/pull/37982) - fix(proxy): parse form-encoded video edit/extension bodies after auth by [@Souravrajvi0](https://github.com/Souravrajvi0) in [#36513](https://github.com/BerriAI/litellm/pull/36513) - fix(anthropic): keep legacy thinking budget\_tokens on Claude 4.6 models on /v1/messages by [@mateo-berri](https://github.com/mateo-berri) in [#38108](https://github.com/BerriAI/litellm/pull/38108) - fix(utils): make prompt\_token\_calculator count claude models again by [@ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#38130](https://github.com/BerriAI/litellm/pull/38130) - fix(proxy): keep every value of a repeated form key, and gate the tests that hid it by [@ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#37908](https://github.com/BerriAI/litellm/pull/37908) - fix(health): apply model\_info.health\_check\_params to health check probes by [@mateo-berri](https://github.com/mateo-berri) in [#38101](https://github.com/BerriAI/litellm/pull/38101) - fix(runwayml): route every generation endpoint and fix video cost tracking by [@mateo-berri](https://github.com/mateo-berri) in [#38115](https://github.com/BerriAI/litellm/pull/38115) - fix(passthrough): attribute spend and release budget reservation on router-model /vllm and /azure routes by [@mateo-berri](https://github.com/mateo-berri) in [#38111](https://github.com/BerriAI/litellm/pull/38111) - fix: match OpenAI SDK wire format on image/video routes by [@mateo-berri](https://github.com/mateo-berri) in [#38104](https://github.com/BerriAI/litellm/pull/38104) - fix(ci): give three unit shards a job deadline that outlasts their pytest budget by [@tin-berri](https://github.com/tin-berri) in [#38139](https://github.com/BerriAI/litellm/pull/38139) - feat(ui): add Gemini Family auto-router preset by [@tin-berri](https://github.com/tin-berri) in [#38138](https://github.com/BerriAI/litellm/pull/38138) - fix(logging\_worker): carry queued tasks across event-loop change instead of dropping them by [@mateo-berri](https://github.com/mateo-berri) in [#38144](https://github.com/BerriAI/litellm/pull/38144) - feat(proxy): enforce vector-store upload security controls on /v1/rag/ingest by [@mateo-berri](https://github.com/mateo-berri) in [#38135](https://github.com/BerriAI/litellm/pull/38135) - test(e2e): pin require\_managed\_files enforcement behind a marker-gated stack phase by [@mateo-berri](https://github.com/mateo-berri) in [#38117](https://github.com/BerriAI/litellm/pull/38117) - refactor(ui): move the dashboard onto class-variance-authority by [@yuneng-berri](https://github.com/yuneng-berri) in [#38125](https://github.com/BerriAI/litellm/pull/38125) - refactor(utils)!: delete prompt\_token\_calculator by [@ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#38132](https://github.com/BerriAI/litellm/pull/38132) - fix(auto-router): list configured auto-routers in the usage picker before they have traffic by [@tin-berri](https://github.com/tin-berri) in [#38129](https://github.com/BerriAI/litellm/pull/38129) - refactor(ui): install the shadcn field primitive by [@yuneng-berri](https://github.com/yuneng-berri) in [#38126](https://github.com/BerriAI/litellm/pull/38126) - fix(complexity\_router): keep both ends of a clipped classifier context turn by [@tin-berri](https://github.com/tin-berri) in [#38141](https://github.com/BerriAI/litellm/pull/38141) - fix(ci): ignore-list recursive form-field flatteners in recursive\_detector by [@mateo-berri](https://github.com/mateo-berri) in [#38149](https://github.com/BerriAI/litellm/pull/38149) - fix(passthrough): stop leaking the caller's virtual key on credential-less Vertex passthrough by [@mateo-berri](https://github.com/mateo-berri) in [#38114](https://github.com/BerriAI/litellm/pull/38114) - fix(router): stop copying forwarded credentials into retry breadcrumbs by [@mateo-berri](https://github.com/mateo-berri) in [#38133](https://github.com/BerriAI/litellm/pull/38133) - feat(e2e): record and replay streamed provider responses chunk-for-chunk by [@mateo-berri](https://github.com/mateo-berri) in [#38136](https://github.com/BerriAI/litellm/pull/38136) - fix: tolerate stream chunks without a choices key in stream\_chunk\_builder by [@AkshaySasi](https://github.com/AkshaySasi) in [#34382](https://github.com/BerriAI/litellm/pull/34382) - fix(files): decode x-litellm-model encoded file\_id in chat + responses by [@hclsys](https://github.com/hclsys) in [#29832](https://github.com/BerriAI/litellm/pull/29832) - fix(videos): forward uploaded source file on /v1/videos/edits to the provider by [@mateo-berri](https://github.com/mateo-berri) in [#38155](https://github.com/BerriAI/litellm/pull/38155) - fix(s3\_v2): percent-encode object keys once so signed and sent URLs match by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38005](https://github.com/BerriAI/litellm/pull/38005) - feat(ui): add error-code drilldown for failed requests on caching page by [@ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#38156](https://github.com/BerriAI/litellm/pull/38156) - feat(search): add Grounding with Bing Search (bing\_grounding) as a search provider by [@mateo-berri](https://github.com/mateo-berri) in [#38119](https://github.com/BerriAI/litellm/pull/38119) - ci: ban row-rewriting DML from prisma migrations by [@yucheng-berri](https://github.com/yucheng-berri) in [#37899](https://github.com/BerriAI/litellm/pull/37899) - fix(langsmith): keep root-run ids self-consistent so batch ingest stops rejecting header-tagged requests by [@yucheng-berri](https://github.com/yucheng-berri) in [#38116](https://github.com/BerriAI/litellm/pull/38116) - ci(e2e): record the e2e suite weekly and replay it on weekdays with zero egress by [@mateo-berri](https://github.com/mateo-berri) in [#38163](https://github.com/BerriAI/litellm/pull/38163) - chore(codeowners): unown ui container plumbing and generated files by [@ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#38124](https://github.com/BerriAI/litellm/pull/38124) - fix(logging): skip parsing redacted tool call arguments by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38169](https://github.com/BerriAI/litellm/pull/38169) - feat(complexity\_router): bound the classifier context block, not each turn in it by [@tin-berri](https://github.com/tin-berri) in [#38145](https://github.com/BerriAI/litellm/pull/38145) - fix(http\_handler): dispose aiohttp session when AsyncHTTPHandler is finalized without a running loop by [@anmolg1997](https://github.com/anmolg1997) in [#36670](https://github.com/BerriAI/litellm/pull/36670) - fix(proxy): reset a stuck team member's budget by [@yassin-berriai](https://github.com/yassin-berriai) in [#37971](https://github.com/BerriAI/litellm/pull/37971) - fix(anthropic/bedrock): request summarized adaptive thinking for reasoning\_effort and use provider thinking token counts by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#37979](https://github.com/BerriAI/litellm/pull/37979) - fix(completion\_extras): forward reasoning\_effort=max through the Responses API bridge by [@mateo-berri](https://github.com/mateo-berri) in [#38222](https://github.com/BerriAI/litellm/pull/38222) - feat(vertex\_ai): add native Vertex AI Interactions API support by [@mateo-berri](https://github.com/mateo-berri) in [#38229](https://github.com/BerriAI/litellm/pull/38229) - test(mcp): drain the logging worker after each test so queued callbacks cannot leak into the next test by [@ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#38228](https://github.com/BerriAI/litellm/pull/38228) - fix(ui): forward OAuth issuer/authorization/token/registration URLs from the MCP server edit form by [@yassin-berriai](https://github.com/yassin-berriai) in [#38154](https://github.com/BerriAI/litellm/pull/38154) - fix(together\_ai): default endpoints to api.together.ai instead of api.together.xyz by [@mateo-berri](https://github.com/mateo-berri) in [#38233](https://github.com/BerriAI/litellm/pull/38233) - fix(bedrock\_mantle): register a Bedrock runtime passthrough config so /bedrock/model/<deployment>/invoke works by [@mateo-berri](https://github.com/mateo-berri) in [#38231](https://github.com/BerriAI/litellm/pull/38231) - fix(router): resolve provider from api\_base in deployment validation and acompletion by [@mateo-berri](https://github.com/mateo-berri) in [#38235](https://github.com/BerriAI/litellm/pull/38235) - fix(model\_prices): raise bedrock\_mantle gpt-5.6 max\_input\_tokens to Mantle's enforced [`1050000`](https://github.com/BerriAI/litellm/commit/1050000) by [@mateo-berri](https://github.com/mateo-berri) in [#38225](https://github.com/BerriAI/litellm/pull/38225) - fix(bedrock\_mantle): normalize Codex input item types Mantle rejects by [@mateo-berri](https://github.com/mateo-berri) in [#38227](https://github.com/BerriAI/litellm/pull/38227) - fix(proxy): store the actual selected model in spend logs for Azure Model Router by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#37770](https://github.com/BerriAI/litellm/pull/37770) - feat(router): per-group supported reasoning efforts with the max level by [@tin-berri](https://github.com/tin-berri) in [#37897](https://github.com/BerriAI/litellm/pull/37897) - fix(proxy): stop expected 4xx responses from saturating worker CPU on failure logging by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38102](https://github.com/BerriAI/litellm/pull/38102) - fix(caching): use upstream RedisCluster on redis-py with per-connection recovery by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38171](https://github.com/BerriAI/litellm/pull/38171) - perf(auth): drop guaranteed-miss internal-cache Redis read from team object lookup by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38073](https://github.com/BerriAI/litellm/pull/38073) - fix(together\_ai): route chat completions through a dedicated TogetherAIChatConfig by [@mateo-berri](https://github.com/mateo-berri) in [#38248](https://github.com/BerriAI/litellm/pull/38248) - fix(ui): read reasoning tokens from Responses API output\_tokens\_details by [@yassin-berriai](https://github.com/yassin-berriai) in [#37952](https://github.com/BerriAI/litellm/pull/37952) - fix(dashboard): don't show a stale provider prompt-cache chip on a response-cache hit by [@yassin-berriai](https://github.com/yassin-berriai) in [#37951](https://github.com/BerriAI/litellm/pull/37951) - fix(ui): render tag-based guardrail mode instead of crashing the guardrails page by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#37493](https://github.com/BerriAI/litellm/pull/37493) - fix(scim): return user\_id as Group members\[].value on transformed group responses by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38161](https://github.com/BerriAI/litellm/pull/38161) - fix(scim): preserve existing team memberships when POST /Users adoption carries no groups by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38166](https://github.com/BerriAI/litellm/pull/38166) - fix(router): support mid-stream fallback for anthropic\_messages route type by [@yassin-berriai](https://github.com/yassin-berriai) in [#38153](https://github.com/BerriAI/litellm/pull/38153) - fix(auth): support wildcard prefixes in jwt team\_allowed\_routes by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#37756](https://github.com/BerriAI/litellm/pull/37756) - feat(models): add missing Together AI serverless models to the cost map by [@mateo-berri](https://github.com/mateo-berri) in [#38230](https://github.com/BerriAI/litellm/pull/38230) - fix(cerebras)!: add max\_retries and extra\_headers to get\_supported\_openai\_params by [@deepanshululla](https://github.com/deepanshululla) in [#36601](https://github.com/BerriAI/litellm/pull/36601) - fix(anthropic): translate tool\_result document blocks in the /v1/messages bridge by [@mateo-berri](https://github.com/mateo-berri) in [#38251](https://github.com/BerriAI/litellm/pull/38251) - fix(team): serialize member\_add, member\_delete, and delete under the team's advisory lock by [@yassin-berriai](https://github.com/yassin-berriai) in [#37969](https://github.com/BerriAI/litellm/pull/37969) - docs(pr-template): split Caveats bullets into severity tiers and call for plain engineering language by [@mateo-berri](https://github.com/mateo-berri) in [#38252](https://github.com/BerriAI/litellm/pull/38252) - fix(anthropic): carry tool\_result document blocks through the /v1/messages responses bridge by [@mateo-berri](https://github.com/mateo-berri) in [#38261](https://github.com/BerriAI/litellm/pull/38261) - fix(together\_ai): pass tools through for models missing from the registry by [@mateo-berri](https://github.com/mateo-berri) in [#38265](https://github.com/BerriAI/litellm/pull/38265) - fix(anthropic): carry user-content document blocks through the /v1/messages responses bridge by [@mateo-berri](https://github.com/mateo-berri) in [#38267](https://github.com/BerriAI/litellm/pull/38267) - fix(rerank): emit latency and cost headers on /rerank by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#35419](https://github.com/BerriAI/litellm/pull/35419) - perf(streaming): add shared JSONFragmentAccumulator for Vertex and Anthropic by [@deepanshululla](https://github.com/deepanshululla) in [#36610](https://github.com/BerriAI/litellm/pull/36610) - fix(together\_ai): strip internal thinking fields from outbound messages, keep reasoning\_content by [@mateo-berri](https://github.com/mateo-berri) in [#38275](https://github.com/BerriAI/litellm/pull/38275) - fix(router): persist attempted\_fallbacks and original\_model\_group into spend logs metadata by [@yucheng-berri](https://github.com/yucheng-berri) in [#38107](https://github.com/BerriAI/litellm/pull/38107) - fix(logging): redact tool call arguments to valid JSON and preserve null content by [@yucheng-berri](https://github.com/yucheng-berri) in [#38182](https://github.com/BerriAI/litellm/pull/38182) - fix(ui): stack policy flow builder below the popup layer so guardrail options render by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38273](https://github.com/BerriAI/litellm/pull/38273) - test: gate the test tree on B003 so a test cannot swap os.environ for a plain dict by [@ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#38274](https://github.com/BerriAI/litellm/pull/38274) - refactor(repositories): type prisma table access with one generic protocol by [@mateo-berri](https://github.com/mateo-berri) in [#38205](https://github.com/BerriAI/litellm/pull/38205) - fix(anthropic): buffer streamed responses carrying server-fulfilled tools so retrieval tool calls never reach the client by [@mateo-berri](https://github.com/mateo-berri) in [#36245](https://github.com/BerriAI/litellm/pull/36245) - test(together\_ai): regression suite across chat, responses, and messages surfaces by [@mateo-berri](https://github.com/mateo-berri) in [#38283](https://github.com/BerriAI/litellm/pull/38283) - feat(logging): add async\_post\_call\_failure\_deployment\_hook by [@deepanshululla](https://github.com/deepanshululla) in [#36657](https://github.com/BerriAI/litellm/pull/36657) - chore: bump litellm-enterprise 0.1.59 -> 0.1.60, litellm 1.99.0 -> 1.100.0 by [@yuneng-berri](https://github.com/yuneng-berri) in [#38243](https://github.com/BerriAI/litellm/pull/38243) - test(e2e): cover Together AI reasoning, tool calls, template kwargs, and cost through a live proxy by [@mateo-berri](https://github.com/mateo-berri) in [#38286](https://github.com/BerriAI/litellm/pull/38286) - fix(logging): keep tracebacks for provider-originated 4xx errors by [@mateo-berri](https://github.com/mateo-berri) in [#38296](https://github.com/BerriAI/litellm/pull/38296) - chore(ci): promote internal staging to main by [@yuneng-berri](https://github.com/yuneng-berri) in [#38293](https://github.com/BerriAI/litellm/pull/38293) - refactor(ui): install the shadcn alert primitive by [@yuneng-berri](https://github.com/yuneng-berri) in [#38300](https://github.com/BerriAI/litellm/pull/38300) - refactor(ui): re-pull label, textarea, separator and skeleton from the registry by [@yuneng-berri](https://github.com/yuneng-berri) in [#38302](https://github.com/BerriAI/litellm/pull/38302) - feat(prometheus): configure deployment caller identity by [@mphilippnv](https://github.com/mphilippnv) in [#38221](https://github.com/BerriAI/litellm/pull/38221) - test(e2e): let the Together replayed-reasoning case survive a single provider miss by [@mateo-berri](https://github.com/mateo-berri) in [#38314](https://github.com/BerriAI/litellm/pull/38314) - fix(otel): map /v1/messages provider errors before failure logging by [@mateo-berri](https://github.com/mateo-berri) in [#38310](https://github.com/BerriAI/litellm/pull/38310) - fix(exceptions): map upstream status codes for providers with no exception\_type branch by [@mateo-berri](https://github.com/mateo-berri) in [#38318](https://github.com/BerriAI/litellm/pull/38318) - fix(passthrough): record ownership of streamed responses under managed ids by [@mateo-berri](https://github.com/mateo-berri) in [#38320](https://github.com/BerriAI/litellm/pull/38320) - fix(proxy): encrypt streamed responses ids on /openai/v1/responses and /responses aliases by [@mateo-berri](https://github.com/mateo-berri) in [#38325](https://github.com/BerriAI/litellm/pull/38325) - test(cost-calc): pin the rate fallbacks inside a tiered-pricing tier by [@yuneng-berri](https://github.com/yuneng-berri) in [#38327](https://github.com/BerriAI/litellm/pull/38327) - test(e2e): cover the Bedrock provider-feature cells customers run by [@mateo-berri](https://github.com/mateo-berri) in [#38232](https://github.com/BerriAI/litellm/pull/38232) - fix(together\_ai): fail open on response\_format instead of dropping it for unregistered models by [@mateo-berri](https://github.com/mateo-berri) in [#38269](https://github.com/BerriAI/litellm/pull/38269) - fix(proxy): honor DATABASE\_DISABLE\_PREPARED\_STATEMENTS in componentized entrypoints by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38363](https://github.com/BerriAI/litellm/pull/38363) - fix(anthropic-responses): preserve structured output strictness by [@eugene-yao-zocdoc](https://github.com/eugene-yao-zocdoc) in [#38211](https://github.com/BerriAI/litellm/pull/38211) - chore(typing): roll up the daily tech debt cleanups from Aug 20 to Aug 26 by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#37922](https://github.com/BerriAI/litellm/pull/37922) - fix(router): resolve hidden aliases for explicit lookup by [@daniel-meismer-zocdoc](https://github.com/daniel-meismer-zocdoc) in [#38272](https://github.com/BerriAI/litellm/pull/38272) - fix(ui): keep focus in the add model public name input while typing by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38366](https://github.com/BerriAI/litellm/pull/38366) - fix(model\_prices): price 1-hour cache writes on claude-3-haiku and claude-3-opus at 2x input by [@mateo-berri](https://github.com/mateo-berri) in [#38371](https://github.com/BerriAI/litellm/pull/38371) - fix(proxy): keep the caller's Google token on credential-less Vertex passthrough under custom auth by [@mateo-berri](https://github.com/mateo-berri) in [#38299](https://github.com/BerriAI/litellm/pull/38299) - fix(mcp): preserve provider access token lifetime by [@daniel-meismer-zocdoc](https://github.com/daniel-meismer-zocdoc) in [#38271](https://github.com/BerriAI/litellm/pull/38271) - chore(ui): remove stale "New" badges from the dashboard by [@yuneng-berri](https://github.com/yuneng-berri) in [#38374](https://github.com/BerriAI/litellm/pull/38374) - test(cost-estimate): pin the prices and period totals /cost/estimate returns by [@yuneng-berri](https://github.com/yuneng-berri) in [#38315](https://github.com/BerriAI/litellm/pull/38315) - fix(ci): let the mutation workflow find covered lines so it generates mutants by [@yuneng-berri](https://github.com/yuneng-berri) in [#38305](https://github.com/BerriAI/litellm/pull/38305) - fix(model\_prices): raise bedrock\_mantle gpt-5.5 and gpt-5.4 max\_input\_tokens to Mantle's enforced [`1050000`](https://github.com/BerriAI/litellm/commit/1050000) by [@mateo-berri](https://github.com/mateo-berri) in [#38368](https://github.com/BerriAI/litellm/pull/38368) - fix(azure/realtime): authenticate realtime websocket with Azure AD token when no api-key by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#34658](https://github.com/BerriAI/litellm/pull/34658) - fix(bedrock): map reasoning\_effort to reasoning.effort for OpenAI GPT-5.x on Converse by [@6matt](https://github.com/6matt) in [#38279](https://github.com/BerriAI/litellm/pull/38279) - test(prometheus): cover caller-identity config failure cases by [@yucheng-berri](https://github.com/yucheng-berri) in [#38380](https://github.com/BerriAI/litellm/pull/38380) - fix(redis): support credential providers across clients by [@eugene-yao-zocdoc](https://github.com/eugene-yao-zocdoc) in [#38094](https://github.com/BerriAI/litellm/pull/38094) - fix(health): support `mode: image_edit` in health checks by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38291](https://github.com/BerriAI/litellm/pull/38291) - fix(cost-map): add US data residency uplift to claude-sonnet-4-6 and mythos entries by [@mateo-berri](https://github.com/mateo-berri) in [#38369](https://github.com/BerriAI/litellm/pull/38369) - fix(anthropic): raise missing-credential error on /v1/messages passthrough by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38240](https://github.com/BerriAI/litellm/pull/38240) - fix(mcp): complete DCR bridge OAuth challenges by [@daniel-meismer-zocdoc](https://github.com/daniel-meismer-zocdoc) in [#37384](https://github.com/BerriAI/litellm/pull/37384) - test(proxy): pin the request-validation contracts in proxy/\_types.py by [@yuneng-berri](https://github.com/yuneng-berri) in [#38307](https://github.com/BerriAI/litellm/pull/38307) - docs(CLAUDE.md): add pull-before-work rule by [@mateo-berri](https://github.com/mateo-berri) in [#38386](https://github.com/BerriAI/litellm/pull/38386) - fix(anthropic): scale cache costs by fast mode and trust served speed by [@mateo-berri](https://github.com/mateo-berri) in [#38378](https://github.com/BerriAI/litellm/pull/38378) - fix(pricing): add azure gpt-5.6 cache write rates and correct data zone priority by [@mateo-berri](https://github.com/mateo-berri) in [#38370](https://github.com/BerriAI/litellm/pull/38370) - docs: tighten the pull-before-work rule in CLAUDE.md by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38389](https://github.com/BerriAI/litellm/pull/38389) - fix(health): strip credential fields from GET /health output by [@Siraj637909](https://github.com/Siraj637909) in [#37090](https://github.com/BerriAI/litellm/pull/37090) - fix(minimax): attach MINIMAX\_API\_KEY on anthropic messages requests by [@mateo-berri](https://github.com/mateo-berri) in [#38393](https://github.com/BerriAI/litellm/pull/38393) - refactor(ui): replace hand-picked z-index values with one named scale and lint it by [@ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#38282](https://github.com/BerriAI/litellm/pull/38282) - fix(health): probe Azure GA realtime path for transcription-only models by [@mateo-berri](https://github.com/mateo-berri) in [#38390](https://github.com/BerriAI/litellm/pull/38390) - fix(bedrock): parse cacheDetails for Converse 1h/5m cache write cost split by [@danielva-monday](https://github.com/danielva-monday) in [#36762](https://github.com/BerriAI/litellm/pull/36762) - fix(caching): flush async cache writes cancelled at event loop shutdown by [@mateo-berri](https://github.com/mateo-berri) in [#38385](https://github.com/BerriAI/litellm/pull/38385) - fix(router): resolve model\_group\_alias before pre-routing strategy dispatch by [@tin-berri](https://github.com/tin-berri) in [#38382](https://github.com/BerriAI/litellm/pull/38382) - feat(proxy): enforce rpm/tpm on model add + fix validation error title in UI by [@kunal2002](https://github.com/kunal2002) in [#36518](https://github.com/BerriAI/litellm/pull/36518) - refactor(ui): move every page header onto the shared PageHeader by [@yuneng-berri](https://github.com/yuneng-berri) in [#38306](https://github.com/BerriAI/litellm/pull/38306) - fix(proxy): stop cache eviction errors from failing /key/update by [@yuneng-berri](https://github.com/yuneng-berri) in [#38308](https://github.com/BerriAI/litellm/pull/38308) - fix(aiohttp): honor global ssl\_verify on the aiohttp\_openai handler path by [@mateo-berri](https://github.com/mateo-berri) in [#38400](https://github.com/BerriAI/litellm/pull/38400) - fix(logging\_worker): rescue dequeued logging tasks lost at event loop close by [@mateo-berri](https://github.com/mateo-berri) in [#38394](https://github.com/BerriAI/litellm/pull/38394) - fix(caching): require the namespace delimiter when checking already-namespaced redis keys by [@mateo-berri](https://github.com/mateo-berri) in [#38403](https://github.com/BerriAI/litellm/pull/38403) - fix(prompts): reject keyed prompt\_data with prompt\_id and populate prompt version by [@mateo-berri](https://github.com/mateo-berri) in [#38404](https://github.com/BerriAI/litellm/pull/38404) - fix(cost\_calculator): resolve real cost key when model\_name alias contains '/' by [@ksk2023](https://github.com/ksk2023) in [#38344](https://github.com/BerriAI/litellm/pull/38344) - fix(cost-map): correct prompt\_cache\_min\_tokens for Claude Fable 5 and backfill Anthropic re-export entries by [@mateo-berri](https://github.com/mateo-berri) in [#38405](https://github.com/BerriAI/litellm/pull/38405) - test(azure-ai): pin the 422 retry that drops the field the provider rejected by [@yuneng-berri](https://github.com/yuneng-berri) in [#38309](https://github.com/BerriAI/litellm/pull/38309) - refactor(ui): read the auto-router tier set through one row list by [@tin-berri](https://github.com/tin-berri) in [#38408](https://github.com/BerriAI/litellm/pull/38408) - fix(proxy): stop empty DB router\_settings lists from clobbering yaml fallbacks by [@mateo-berri](https://github.com/mateo-berri) in [#38406](https://github.com/BerriAI/litellm/pull/38406) - fix(team): allow no-reset default budgets for team members by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#37708](https://github.com/BerriAI/litellm/pull/37708) - fix: forward image content lists to DeepSeek vision models by [@yucheng-berri](https://github.com/yucheng-berri) in [#38397](https://github.com/BerriAI/litellm/pull/38397) - fix(fireworks\_ai): stop using the trace id as the session affinity key by [@Hamjaster](https://github.com/Hamjaster) in [#35754](https://github.com/BerriAI/litellm/pull/35754) - fix(gemini-realtime): keep the client's voice on Vertex AI native-audio Live by [@yucheng-berri](https://github.com/yucheng-berri) in [#38395](https://github.com/BerriAI/litellm/pull/38395) - fix(vertex\_ai): bill Gemini grounding per unique web search query by [@ousamabenyounes](https://github.com/ousamabenyounes) in [#36397](https://github.com/BerriAI/litellm/pull/36397) - fix(health): make the image\_edit health probe moderation-safe by [@mateo-berri](https://github.com/mateo-berri) in [#38417](https://github.com/BerriAI/litellm/pull/38417) - test: gate the test tree on fifteen assertion and handler rules it already satisfies by [@ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#38361](https://github.com/BerriAI/litellm/pull/38361) - fix(proxy): key lazy openapi stubs off registered features, not sys.modules by [@mateo-berri](https://github.com/mateo-berri) in [#38416](https://github.com/BerriAI/litellm/pull/38416) - fix(proxy): derive auto-router health from its underlying models by [@tin-berri](https://github.com/tin-berri) in [#38174](https://github.com/BerriAI/litellm/pull/38174) - fix(responses): let cache-control injection reach the system prompt from instructions by [@tin-berri](https://github.com/tin-berri) in [#38120](https://github.com/BerriAI/litellm/pull/38120) - fix(gemini): bill Google Maps grounding as its own SKU by [@mateo-berri](https://github.com/mateo-berri) in [#38418](https://github.com/BerriAI/litellm/pull/38418) - fix(speech): keep proxy metadata and completion cost through the TTS completion bridge by [@mateo-berri](https://github.com/mateo-berri) in [#38414](https://github.com/BerriAI/litellm/pull/38414) - fix: map Gemini ON\_DEMAND\_FLEX traffic type to flex service tier by [@bisma-nawaz](https://github.com/bisma-nawaz) in [#37724](https://github.com/BerriAI/litellm/pull/37724) - feat(langfuse): support langfuse\_environment as a per-key dynamic callback param by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38264](https://github.com/BerriAI/litellm/pull/38264) - feat(proxy): hide unhealthy models from model listings, opt-in by [@yuneng-berri](https://github.com/yuneng-berri) in [#38313](https://github.com/BerriAI/litellm/pull/38313) - fix(mcp): honor admin-entered OAuth URLs on authorize after issuer yield by [@mateo-berri](https://github.com/mateo-berri) in [#38379](https://github.com/BerriAI/litellm/pull/38379) - fix(model\_prices): correct gemini-3.5-flash-lite flex cache-read pricing by [@mateo-berri](https://github.com/mateo-berri) in [#38422](https://github.com/BerriAI/litellm/pull/38422) - fix(cost): price gemini-live-2.5-flash-native-audio realtime sessions by [@mateo-berri](https://github.com/mateo-berri) in [#38419](https://github.com/BerriAI/litellm/pull/38419) - fix(cost-map): correct Gemini TTS and native-audio rates by [@mateo-berri](https://github.com/mateo-berri) in [#38412](https://github.com/BerriAI/litellm/pull/38412) - fix(prompts): propagate PATCHed prompt templates to every worker and pod by [@mateo-berri](https://github.com/mateo-berri) in [#38411](https://github.com/BerriAI/litellm/pull/38411) - fix(model\_prices): bill gemini -latest/preview alias cache reads at 10% of input by [@mateo-berri](https://github.com/mateo-berri) in [#38423](https://github.com/BerriAI/litellm/pull/38423) - fix(proxy): sync search tools into the router on management writes by [@yuneng-berri](https://github.com/yuneng-berri) in [#38392](https://github.com/BerriAI/litellm/pull/38392) - feat(guardrails): track Azure Prompt Shield usage and cost with spend isolation by [@yucheng-berri](https://github.com/yucheng-berri) in [#38387](https://github.com/BerriAI/litellm/pull/38387) - fix(prompts): apply prompt templates before routing on /v1/responses and honor ignore\_prompt\_manager\_model by [@mateo-berri](https://github.com/mateo-berri) in [#38407](https://github.com/BerriAI/litellm/pull/38407) - fix(cost): make cost-breakdown headers respect service tier by [@mateo-berri](https://github.com/mateo-berri) in [#38424](https://github.com/BerriAI/litellm/pull/38424) - fix(mcp): add litellm\[mcp] extra and actionable error when streamable\_http\_client is missing by [@mateo-berri](https://github.com/mateo-berri) in [#38399](https://github.com/BerriAI/litellm/pull/38399) - revert(proxy): remove router\_model\_name from auto-routed response bodies by [@tin-berri](https://github.com/tin-berri) in [#38429](https://github.com/BerriAI/litellm/pull/38429) - fix(google\_genai): price streamed generateContent with the provider that served it by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#36055](https://github.com/BerriAI/litellm/pull/36055) - fix(logging): stop billing and logging response reads as LLM calls by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#36890](https://github.com/BerriAI/litellm/pull/36890) - fix(budget): serialize model\_max\_budget before the /budget/update write by [@yuneng-berri](https://github.com/yuneng-berri) in [#38430](https://github.com/BerriAI/litellm/pull/38430) - fix(ui): block the auto-router submit on a missing classifier model and an orphaned keyword rule by [@tin-berri](https://github.com/tin-berri) in [#38427](https://github.com/BerriAI/litellm/pull/38427) - feat(complexity\_router): heuristic-first classifier chaining by [@tin-berri](https://github.com/tin-berri) in [#38428](https://github.com/BerriAI/litellm/pull/38428) - test(e2e): un-skip the per-model budget update case by [@yuneng-berri](https://github.com/yuneng-berri) in [#38437](https://github.com/BerriAI/litellm/pull/38437) - fix(cost): stop double-billing cached tokens that overlap a modality by [@Srivatsa03](https://github.com/Srivatsa03) in [#37407](https://github.com/BerriAI/litellm/pull/37407) - feat(ui): add Teams list CSV export with budgets, model grants, and rate limits by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38436](https://github.com/BerriAI/litellm/pull/38436) - fix(mcp): accept raw x-litellm-api-key on streamable HTTP admission by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38364](https://github.com/BerriAI/litellm/pull/38364) - fix: bound row count on GET /spend/logs to stop unbounded LiteLLM\_SpendLogs scans by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38420](https://github.com/BerriAI/litellm/pull/38420) - fix(scim): apply default\_team\_params (incl. models) to SCIM-created teams by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38433](https://github.com/BerriAI/litellm/pull/38433) - fix(prompts): propagate prompt deletes to every worker and pod by [@mateo-berri](https://github.com/mateo-berri) in [#38434](https://github.com/BerriAI/litellm/pull/38434) - fix(anthropic\_adapter): carry web search cost into /v1/messages breakdown headers by [@mateo-berri](https://github.com/mateo-berri) in [#38439](https://github.com/BerriAI/litellm/pull/38439) - fix(ui): show custom technical keywords on every router whose scorer runs by [@tin-berri](https://github.com/tin-berri) in [#38451](https://github.com/BerriAI/litellm/pull/38451) - fix(e2e): move the vertex realtime suite off the retired Live preview model by [@yuneng-berri](https://github.com/yuneng-berri) in [#38454](https://github.com/BerriAI/litellm/pull/38454) - feat(newrelic): per-team cost and usage metrics via team callbacks by [@yucheng-berri](https://github.com/yucheng-berri) in [#37610](https://github.com/BerriAI/litellm/pull/37610) - fix(ui): carry a preset's per-tier litellm\_params through the prefill by [@tin-berri](https://github.com/tin-berri) in [#38453](https://github.com/BerriAI/litellm/pull/38453) - fix(e2e): size the mid-conversation-system cache prefix above the minimum deterministically by [@yuneng-berri](https://github.com/yuneng-berri) in [#38468](https://github.com/BerriAI/litellm/pull/38468) - fix(e2e): disable thinking on the gemini chat cost test instead of racing its budget by [@yuneng-berri](https://github.com/yuneng-berri) in [#38469](https://github.com/BerriAI/litellm/pull/38469) - feat(ui): put the auto-router savings hero on a spend rail and a four-tile row by [@tin-berri](https://github.com/tin-berri) in [#38470](https://github.com/BerriAI/litellm/pull/38470) - feat(ui): toggle internal health check visibility in request logs by [@mateo-berri](https://github.com/mateo-berri) in [#38391](https://github.com/BerriAI/litellm/pull/38391) - fix(mcp): canonicalize bearer scheme on bridge egress by [@daniel-meismer-zocdoc](https://github.com/daniel-meismer-zocdoc) in [#38398](https://github.com/BerriAI/litellm/pull/38398) - refactor: clean up fresh tech debt from 2026-08-27 window by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38484](https://github.com/BerriAI/litellm/pull/38484) - fix(exception\_mapping\_utils): map unmapped exceptions when model and provider are unset by [@mateo-berri](https://github.com/mateo-berri) in [#38496](https://github.com/BerriAI/litellm/pull/38496) - fix(ui\_sso): resolve highest privilege Entra app role, not first in claim by [@imranismail](https://github.com/imranismail) in [#36728](https://github.com/BerriAI/litellm/pull/36728) - fix(proxy): regenerate lazy OpenAPI snapshot and guard it in CI by [@mateo-berri](https://github.com/mateo-berri) in [#38410](https://github.com/BerriAI/litellm/pull/38410) - feat(ui): add cache hit/miss filter to Request Logs by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38432](https://github.com/BerriAI/litellm/pull/38432) - fix(bedrock): sign rerank requests with the shared header-filtered SigV4 helper (internal copy of [#36462](https://github.com/BerriAI/litellm/issues/36462)) by [@mateo-berri](https://github.com/mateo-berri) in [#38093](https://github.com/BerriAI/litellm/pull/38093) - fix(bedrock): sign rerank requests with the shared, header-filtered SigV4 helper by [@noahnistler](https://github.com/noahnistler) in [#36462](https://github.com/BerriAI/litellm/pull/36462) - fix(ui): order the auto-routers table newest first so a new router lands on page one by [@tin-berri](https://github.com/tin-berri) in [#38545](https://github.com/BerriAI/litellm/pull/38545) - feat(ui): run the Anthropic Family preset's reasoning tier on Opus 5 at high thinking by [@tin-berri](https://github.com/tin-berri) in [#38490](https://github.com/BerriAI/litellm/pull/38490) - build(ui): bump nginx to 1.31-alpine by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38541](https://github.com/BerriAI/litellm/pull/38541) - feat(otel): support per-team/per-key service.name for OTel v2 destinations by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38532](https://github.com/BerriAI/litellm/pull/38532) - test(e2e): de-flake the cost-header cache read and the router fallback control by [@yuneng-berri](https://github.com/yuneng-berri) in [#38435](https://github.com/BerriAI/litellm/pull/38435) - feat(gemini): day-0 support for gemini-3.5-transcribe and transcribe-live by [@mateo-berri](https://github.com/mateo-berri) in [#38540](https://github.com/BerriAI/litellm/pull/38540) - feat(health): opt-in model-group allowlist for background health checks and health-check routing by [@mateo-berri](https://github.com/mateo-berri) in [#38539](https://github.com/BerriAI/litellm/pull/38539) - fix(mcp): keep upstream OAuth Authorization when jwt signer hook injects one on tools/call by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38555](https://github.com/BerriAI/litellm/pull/38555) - fix: suppress misleading register\_model unresolved-cost warnings for entries without custom pricing by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38542](https://github.com/BerriAI/litellm/pull/38542) - feat(proxy): opt-in budget rollover carrying overage into the next window by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38514](https://github.com/BerriAI/litellm/pull/38514) - fix(auth): skip guaranteed-miss team lookup for the litellm-dashboard sentinel by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38471](https://github.com/BerriAI/litellm/pull/38471) - fix(key\_management): allow /key/update to keep or shrink MCP server grants the key already holds by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38463](https://github.com/BerriAI/litellm/pull/38463) - fix: keep schema reconciliation from fighting a partitioned LiteLLM\_SpendLogs by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38452](https://github.com/BerriAI/litellm/pull/38452) - fix(ui): open select popups below the trigger instead of over it by [@tin-berri](https://github.com/tin-berri) in [#38554](https://github.com/BerriAI/litellm/pull/38554) - fix(realtime): bill trailing audio when a Gemini transcribe Live session closes by [@mateo-berri](https://github.com/mateo-berri) in [#38563](https://github.com/BerriAI/litellm/pull/38563) - test(e2e): cover key generate and update on the Admin UI path by [@yuneng-berri](https://github.com/yuneng-berri) in [#38448](https://github.com/BerriAI/litellm/pull/38448) - chore: bump litellm-enterprise 0.1.60 -> 0.1.61, litellm-proxy-extras 0.4.89 -> 0.4.90 by [@yuneng-berri](https://github.com/yuneng-berri) in [#38566](https://github.com/BerriAI/litellm/pull/38566) - fix(ui): let the paginated search select keep what the user types by [@yuneng-berri](https://github.com/yuneng-berri) in [#38475](https://github.com/BerriAI/litellm/pull/38475) - fix(otel): anchor MCP tool-call spans to the gateway's own trace, link the client's context by [@yucheng-berri](https://github.com/yucheng-berri) in [#38317](https://github.com/BerriAI/litellm/pull/38317) - fix: roll up the open deflake fixes for the MCP logging queue, PTU rollup, license gate, and pricing test isolation by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#37833](https://github.com/BerriAI/litellm/pull/37833) - fix(model\_prices): rolling registry audit - verified models and rates for Novita, DeepInfra, W\&B, Bedrock Sol, Gemini, Fireworks, Azure gpt-5.6, Mistral, Together by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38207](https://github.com/BerriAI/litellm/pull/38207) - test(together\_ai): assert fail-open supported params for models missing from the registry by [@mateo-berri](https://github.com/mateo-berri) in [#38487](https://github.com/BerriAI/litellm/pull/38487) - test(e2e): let the together tool tests accept parallel calls by [@yuneng-berri](https://github.com/yuneng-berri) in [#38567](https://github.com/BerriAI/litellm/pull/38567) - feat(mcp): let a resolved OAuth token target a custom upstream header by [@tin-berri](https://github.com/tin-berri) in [#38456](https://github.com/BerriAI/litellm/pull/38456) - feat(together\_ai): map reasoning\_effort per model class by [@mateo-berri](https://github.com/mateo-berri) in [#38263](https://github.com/BerriAI/litellm/pull/38263) - feat(dashscope): support qwen-image-3.0 and qwen-image-3.0-pro image generation by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38449](https://github.com/BerriAI/litellm/pull/38449) - fix(cost): apply Together AI cache read pricing and per-model registry rates by [@mateo-berri](https://github.com/mateo-berri) in [#38280](https://github.com/BerriAI/litellm/pull/38280) - fix(guardrails): forward aws\_external\_id when the bedrock guardrail assumes a role by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38376](https://github.com/BerriAI/litellm/pull/38376) - fix(transcription): synthesize srt/vtt output for adapters without native subtitle formats by [@mateo-berri](https://github.com/mateo-berri) in [#38561](https://github.com/BerriAI/litellm/pull/38561) - fix(streaming): preserve provider service-tier metadata so Vertex flex streams bill at flex rates by [@mateo-berri](https://github.com/mateo-berri) in [#38458](https://github.com/BerriAI/litellm/pull/38458) - fix(realtime): bill Gemini Live native-audio output tokens at the audio rate by [@mateo-berri](https://github.com/mateo-berri) in [#38457](https://github.com/BerriAI/litellm/pull/38457) - fix(anthropic): carry tool\_reference tool results through the guardrail translation round trip by [@mateo-berri](https://github.com/mateo-berri) in [#38465](https://github.com/BerriAI/litellm/pull/38465) - fix(anthropic-adapter): pass provider-native and OpenAI-format tools through on /v1/messages by [@mateo-berri](https://github.com/mateo-berri) in [#38431](https://github.com/BerriAI/litellm/pull/38431) - test(e2e): serve the vision image from our own fixture by [@yuneng-berri](https://github.com/yuneng-berri) in [#38575](https://github.com/BerriAI/litellm/pull/38575) - feat(together\_ai): add zai-org/GLM-5.3-Flash to the model registry by [@mateo-berri](https://github.com/mateo-berri) in [#38486](https://github.com/BerriAI/litellm/pull/38486) - fix(ui): stop server-searched comboboxes from clobbering picks and queries by [@ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#38574](https://github.com/BerriAI/litellm/pull/38574) - feat(model\_prices): let a map entry declare its exact reasoning\_effort levels by [@tin-berri](https://github.com/tin-berri) in [#38481](https://github.com/BerriAI/litellm/pull/38481) - fix(anthropic): carry the adaptive effort tier to every bridged Claude target by [@tin-berri](https://github.com/tin-berri) in [#38533](https://github.com/BerriAI/litellm/pull/38533) - feat(alerting): add native Microsoft Teams alerting destination by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38367](https://github.com/BerriAI/litellm/pull/38367) - chore(proxy): resync the generated API artifacts with the current models by [@tin-berri](https://github.com/tin-berri) in [#38587](https://github.com/BerriAI/litellm/pull/38587) - fix(router): reject complexity-router settings written outside complexity\_router\_config by [@tin-berri](https://github.com/tin-berri) in [#38570](https://github.com/BerriAI/litellm/pull/38570) - feat(ui): session-level cache observability in request logs by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38442](https://github.com/BerriAI/litellm/pull/38442) - fix(ui): link Virtual Keys hint through the migrated /ui route by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38596](https://github.com/BerriAI/litellm/pull/38596) - fix(anthropic): carry the effort tier only where the target declares reasoning\_effort by [@tin-berri](https://github.com/tin-berri) in [#38592](https://github.com/BerriAI/litellm/pull/38592) - fix(presidio): chunk oversized text before /analyze so large content blocks do not fail by [@yucheng-berri](https://github.com/yucheng-berri) in [#38483](https://github.com/BerriAI/litellm/pull/38483) - fix(logging): stop stream-based log collectors classifying INFO logs as errors by [@yucheng-berri](https://github.com/yucheng-berri) in [#38476](https://github.com/BerriAI/litellm/pull/38476) - feat(ui): dry-run an auto-router config against the backend before saving it by [@tin-berri](https://github.com/tin-berri) in [#38595](https://github.com/BerriAI/litellm/pull/38595) - fix(guardrails): add fail-open mode to CrowdStrike AIDR guardrail by [@yucheng-berri](https://github.com/yucheng-berri) in [#38568](https://github.com/BerriAI/litellm/pull/38568) - fix(router): copy instead of mutating caller metadata when scrubbing fallback stamp keys by [@yucheng-berri](https://github.com/yucheng-berri) in [#38586](https://github.com/BerriAI/litellm/pull/38586) - feat(proxy): opt-in enforce\_fallback\_model\_access authorizes router fallbacks against the calling key by [@ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#38572](https://github.com/BerriAI/litellm/pull/38572) - fix(langfuse): warn and drop invalid LANGFUSE\_TRACING\_ENVIRONMENT instead of failing requests by [@yucheng-berri](https://github.com/yucheng-berri) in [#38582](https://github.com/BerriAI/litellm/pull/38582) - fix(tencent): route thinking through extra\_body in chat completions by [@FelipeRodriguesGare](https://github.com/FelipeRodriguesGare) in [#38100](https://github.com/BerriAI/litellm/pull/38100) - test-check-commits by [@nickhac](https://github.com/nickhac) in [#36344](https://github.com/BerriAI/litellm/pull/36344) - feat(proxy): dry-run a real request body on /auto\_router/test\_routing by [@tin-berri](https://github.com/tin-berri) in [#38590](https://github.com/BerriAI/litellm/pull/38590) - fix(shadow\_eval): refuse a judge model that also serves one of the arms it grades by [@tin-berri](https://github.com/tin-berri) in [#38589](https://github.com/BerriAI/litellm/pull/38589) - fix(anthropic): resolve /v1/messages effort tiers through the capability owner by [@tin-berri](https://github.com/tin-berri) in [#38492](https://github.com/BerriAI/litellm/pull/38492) - fix(router): fall over on raised mid-stream errors in /v1/messages streams by [@mateo-berri](https://github.com/mateo-berri) in [#38606](https://github.com/BerriAI/litellm/pull/38606) - feat(models): add daily Together AI model registry sync script and workflow by [@mateo-berri](https://github.com/mateo-berri) in [#38257](https://github.com/BerriAI/litellm/pull/38257) - feat(ui): the model and wire layer for operator-defined auto-router tier sets by [@tin-berri](https://github.com/tin-berri) in [#38602](https://github.com/BerriAI/litellm/pull/38602) - fix(moonshot, together\_ai): send the reasoning effort Kimi K3 accepts by [@tin-berri](https://github.com/tin-berri) in [#38611](https://github.com/BerriAI/litellm/pull/38611) - fix(ui): one-click theme toggle and matching Docs/Blog styling in the top bar by [@ryan-crabbe-berri](https://github.com/ryan-crabbe-berri) in [#38601](https://github.com/BerriAI/litellm/pull/38601) - feat(proxy): opt-in flags to require rpm/tpm on model and project create by [@ansh-agrawal](https://github.com/ansh-agrawal) in [#36514](https://github.com/BerriAI/litellm/pull/36514) - fix(exceptions): keep a refused connection an APIConnectionError by [@yuneng-berri](https://github.com/yuneng-berri) in [#38624](https://github.com/BerriAI/litellm/pull/38624) - chore(ci): promote internal staging to main by [@yuneng-berri](https://github.com/yuneng-berri) in [#38616](https://github.com/BerriAI/litellm/pull/38616) - fix(anthropic): drop and self-heal empty thinking blocks on /v1/messages by [@tin-berri](https://github.com/tin-berri) in [#38625](https://github.com/BerriAI/litellm/pull/38625) - fix(anthropic): handle per-level reasoning\_effort flags without supports\_reasoning by [@tin-berri](https://github.com/tin-berri) in [#38618](https://github.com/BerriAI/litellm/pull/38618) - fix(complexity\_router): route client housekeeping calls to the cheapest tier by [@tin-berri](https://github.com/tin-berri) in [#38598](https://github.com/BerriAI/litellm/pull/38598) - test: fix staging CI regressions from [#38182](https://github.com/BerriAI/litellm/issues/38182), [#38144](https://github.com/BerriAI/litellm/issues/38144), [#38265](https://github.com/BerriAI/litellm/issues/38265), [#37962](https://github.com/BerriAI/litellm/issues/37962), and [#37969](https://github.com/BerriAI/litellm/issues/37969) by [@mateo-berri](https://github.com/mateo-berri) in [#38304](https://github.com/BerriAI/litellm/pull/38304) - feat(spend): report prompt caching savings as total and gateway-attributed by [@tin-berri](https://github.com/tin-berri) in [#38134](https://github.com/BerriAI/litellm/pull/38134) - fix(proxy): let llm\_api virtual keys read /model\_group/info by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38662](https://github.com/BerriAI/litellm/pull/38662) - feat(ui): edit the auto-router tier set with custom classifier-defined tiers by [@tin-berri](https://github.com/tin-berri) in [#38603](https://github.com/BerriAI/litellm/pull/38603) - fix(proxy): count tools, system, and Anthropic image and document blocks in the count\_tokens fallback (internal copy of [#36671](https://github.com/BerriAI/litellm/issues/36671)) by [@mateo-berri](https://github.com/mateo-berri) in [#38657](https://github.com/BerriAI/litellm/pull/38657) - test: refresh the suites that drifted from langfuse and OpenAI's retired Assistants API by [@yuneng-berri](https://github.com/yuneng-berri) in [#38637](https://github.com/BerriAI/litellm/pull/38637) - test(e2e): unskip four tests whose blockers no longer hold by [@yuneng-berri](https://github.com/yuneng-berri) in [#38640](https://github.com/BerriAI/litellm/pull/38640) - feat(proxy): add paginated GET /public/v1/model\_hub by [@yuneng-berri](https://github.com/yuneng-berri) in [#38636](https://github.com/BerriAI/litellm/pull/38636) - refactor(ui): type search tool params from the generated schema by [@yuneng-berri](https://github.com/yuneng-berri) in [#38633](https://github.com/BerriAI/litellm/pull/38633) - feat(a2a): semantic search over the agent registry via GET /v1/agents?query and an agent\_search MCP tool by [@mateo-berri](https://github.com/mateo-berri) in [#38609](https://github.com/BerriAI/litellm/pull/38609) - fix(model\_prices): add bedrock\_mantle gpt-5.5/5.4 272K tiers, align sol with AWS invoice by [@mateo-berri](https://github.com/mateo-berri) in [#38615](https://github.com/BerriAI/litellm/pull/38615) - fix(ui): keep the usage filter visible when the caller's scope is empty by [@yuneng-berri](https://github.com/yuneng-berri) in [#38581](https://github.com/BerriAI/litellm/pull/38581) - fix(registry): add Gemini Omni 1.1 Flash, xAI grok-imagine image models, Mistral cache-read pricing, GLM 5.3 Flash + Kimi K2.7 Code entries by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38560](https://github.com/BerriAI/litellm/pull/38560) - fix(logging): preserve null end user in callbacks by [@yucheng-berri](https://github.com/yucheng-berri) in [#38642](https://github.com/BerriAI/litellm/pull/38642) - test: close mutation-testing gaps in container, skills and openai-like config factories by [@yuneng-berri](https://github.com/yuneng-berri) in [#38677](https://github.com/BerriAI/litellm/pull/38677) - fix: enforce MCP toolsets attached to a team, org, or internal user by [@yucheng-berri](https://github.com/yucheng-berri) in [#38488](https://github.com/BerriAI/litellm/pull/38488) - fix(tests): drain the global logging worker in RAG aquery billing tests by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38653](https://github.com/BerriAI/litellm/pull/38653) - chore(techdebt): type new signatures and drop slop comments from the last 24h by [@devin-ai-integration](https://github.com/devin-ai-integration)\[bot] in [#38644](https://github.com/Be…

greptile-apps Bot reviewed Aug 25, 2026

View reviewed changes

veria-ai Bot reviewed Aug 25, 2026

View reviewed changes

fix(bedrock_mantle): type the codex item dispatcher without Any

fd1dca0

cursor Bot reviewed Aug 25, 2026

View reviewed changes

mateo-berri enabled auto-merge August 25, 2026 18:02

tin-berri approved these changes Aug 25, 2026

View reviewed changes

Merge remote-tracking branch 'origin/litellm_internal_staging' into l…

636ae5d

…itellm_mantle_codex_input_items

mateo-berri merged commit a9f42f0 into litellm_internal_staging Aug 25, 2026
78 checks passed

mateo-berri deleted the litellm_mantle_codex_input_items branch August 25, 2026 19:04

mateo-berri mentioned this pull request Aug 25, 2026

fix(bedrock_mantle): normalize Codex agent_message inputs #36530

Closed

4 of 5 tasks

devin-ai-integration Bot mentioned this pull request Sep 22, 2026

test(integration): regression tests for August provider translation and streaming bugs #42621

Merged

3 of 5 tasks

Navigation Menu

Uh oh!

fix(bedrock_mantle): normalize Codex input item types Mantle rejects - #38227

fix(bedrock_mantle): normalize Codex input item types Mantle rejects#38227
mateo-berri merged 3 commits into
litellm_internal_stagingfrom
litellm_mantle_codex_input_items

mateo-berri commented Aug 25, 2026 •

edited

Loading

Uh oh!

greptile-apps Bot commented Aug 25, 2026 •

edited

Loading

Important Files Changed

Uh oh!

greptile-apps Bot Aug 25, 2026

Uh oh!

veria-ai Bot Aug 25, 2026

Uh oh!

mateo-berri Aug 25, 2026

Uh oh!

veria-ai Bot Aug 25, 2026

Uh oh!

veria-ai Bot commented Aug 25, 2026

Uh oh!

codecov Bot commented Aug 25, 2026 •

edited

Loading

Uh oh!

mateo-berri commented Aug 25, 2026

Uh oh!

mateo-berri commented Aug 25, 2026

Uh oh!

cursor Bot left a comment

Uh oh!

codspeed Bot commented Aug 25, 2026 •

edited

Loading

Uh oh!

tin-berri left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

mateo-berri commented Aug 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

TLDR

User Flow

Relevant issues

Linear ticket

Pre-Submission checklist

Delays in PR merge?

Screenshots / Proof of Fix

Before (bb27bfd)

Codex CLI multi-agent session

agent_message history item

agent_message with encrypted_content slot

context_compaction carrying a genuine Mantle compaction summary

local_shell_call history pair

/v1/chat/completions control (surface this PR does not touch)

After (ed28581)

Codex CLI multi-agent session

agent_message history item

agent_message with encrypted_content slot

context_compaction carrying a genuine Mantle compaction summary

local_shell_call history pair

/v1/chat/completions control (surface this PR does not touch)

Type

Caveats (if any)

Final Attestation

Uh oh!

greptile-apps Bot commented Aug 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 5/5

Important Files Changed

Uh oh!

greptile-apps Bot Aug 25, 2026

Choose a reason for hiding this comment

Uh oh!

veria-ai Bot Aug 25, 2026

Choose a reason for hiding this comment

Uh oh!

mateo-berri Aug 25, 2026

Choose a reason for hiding this comment

Uh oh!

veria-ai Bot Aug 25, 2026

Choose a reason for hiding this comment

Uh oh!

veria-ai Bot commented Aug 25, 2026

PR overview

Open issues (1)

Uh oh!

codecov Bot commented Aug 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

mateo-berri commented Aug 25, 2026

Uh oh!

mateo-berri commented Aug 25, 2026

Uh oh!

cursor Bot left a comment

Choose a reason for hiding this comment

Uh oh!

codspeed Bot commented Aug 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merging this PR will not alter performance

Footnotes

Uh oh!

tin-berri left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

mateo-berri commented Aug 25, 2026 •

edited

Loading

Before (`bb27bfd`)

After (`ed28581`)

greptile-apps Bot commented Aug 25, 2026 •

edited

Loading

codecov Bot commented Aug 25, 2026 •

edited

Loading

codspeed Bot commented Aug 25, 2026 •

edited

Loading