fix(bedrock/converse): drop toolSpec.strict for Opus 4.7/4.8 (#31582) by mateo-berri · Pull Request #31923 · BerriAI/litellm

mateo-berri · 2026-07-02T00:15:26Z

Relevant issues

Internal copy of #31585 by @ly-wang19 (original authorship preserved on the commits) so CircleCI can run on it. All credit to the original author

Linear ticket

Pre-Submission checklist

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

I have added meaningful tests
My PR passes all CI/CD checks (e.g., lint, format, unit tests)
My PR's scope is as isolated as possible; it only solves 1 specific problem
I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

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

From the original PR: sending a tool with strict: true to bedrock/us.anthropic.claude-opus-4-7 or claude-opus-4-8 on Bedrock Converse fails with 400 tools.0.custom.strict: Extra inputs are not permitted. With this change, strict is dropped from toolSpec for the Opus 4.7/4.8 family only, so the call succeeds; Sonnet 4.5/4.6, Opus <=4.6 and other Anthropic models keep forwarding strict as before (#29814 behavior preserved), and non-Anthropic families (Nova, Llama, GPT-OSS) are unchanged

Regression tests: tests/test_litellm/litellm_core_utils/prompt_templates/test_bedrock_converse_strict_tools_opus_47_48.py (18 passed locally); existing test_bedrock_tools_pt_strict_parameter still passes

Latest e2e QA run: 2026-07-02 02:32 UTC against commit 6bd1f9a, on a live proxy hitting real Bedrock (us-west-2, no mocks). strict: true tool calls to us.anthropic.claude-opus-4-8 and us.anthropic.claude-opus-4-7 returned 200 with real tool calls, us.anthropic.claude-sonnet-4-5-20250929-v1:0 still forwards "strict": true in its raw toolSpec payload (verified in --detailed_debug logs, so the gate is model-scoped rather than a blanket drop), and the same Opus 4.8 request against a pre-fix proxy reproduced the exact 400 tools.0.custom.strict: Extra inputs are not permitted

Type

🐛 Bug Fix

Changes

PR #29814 started forwarding strict into the Bedrock Converse toolSpec, gated on get_bedrock_base_model(model).startswith("anthropic"). That gate is too broad: Bedrock routes Opus 4.7/4.8 through an Anthropic-compatible validator that maps toolSpec to the native tool shape (the custom variant) and rejects the extra strict key. This replaces the gate in litellm_core_utils/prompt_templates/factory.py with bedrock_converse_supports_strict_tools in litellm/llms/bedrock/common_utils.py, which returns False for non-Anthropic Bedrock models and for the Opus 4.7/4.8 family, and True for every other Anthropic model on Bedrock

Following review feedback, the Opus 4.7/4.8 exclusion is not hardcoded in the source: the affected Bedrock entries in model_prices_and_context_window.json carry a bedrock_converse_supports_strict_tools: false flag, resolved through get_model_info with a local cost map fallback (the same approach as bedrock_output_config_effort_ceiling), so future models with the same restriction only need a JSON update

> [!NOTE]
> Low Risk
> Targeted Bedrock tool-payload gating with JSON flags and regression tests; behavior change is limited to omitting strict for flagged Opus 4.7/4.8 models.
>
> Overview
> Fixes Bedrock Converse 400 errors when tools include strict: true on Claude Opus 4.7/4.8, where the Converse validator rejects toolSpec.strict even though other Anthropic Bedrock models accept it.
>
> _bedrock_tools_pt no longer treats every Anthropic Bedrock model as strict-capable; it uses bedrock_converse_supports_strict_tools(model), which keeps strict for Sonnet and Opus ≤4.6 and omits it for Opus 4.7/4.8 and non-Anthropic families (unchanged).
>
> The new helper reads bedrock_converse_supports_strict_tools from the model cost map (with version-suffix normalization and local fallback), matching the pattern used for bedrock_output_config_effort_ceiling. Opus 4.7/4.8 regional entries are flagged false in model_prices_and_context_window.json, and the flag is wired through ModelInfo / get_model_info. Regression tests cover Opus 4.7/4.8 drop vs other models keep/drop behavior.
>
> ^{Reviewed by Cursor Bugbot for commit 6bd1f9a. Bugbot is set up for automated code reviews on this repo. Configure here.}

Bedrock Converse routes Claude Opus 4.7/4.8 through an Anthropic-compatible validator that maps toolSpec to the native tool shape and rejects the extra `strict` key with `tools.N.custom.strict: Extra inputs are not permitted`, even though Anthropic's native API accepts `strict` as a top-level tool field for the same models. Sonnet 4.5/4.6 and Opus <=4.6 accept `toolSpec.strict` unchanged. The existing gate `get_bedrock_base_model(model).startswith("anthropic")` (introduced in #29814 to forward `strict` for Claude on Bedrock Converse) is too broad and regressed Opus 4.7/4.8 callers — see #31582. Replace the inline check with a small `bedrock_converse_supports_strict_tools` helper that excludes the Opus 4.7/4.8 family from strict forwarding. All other Anthropic models on Bedrock keep the existing behavior. Closes #31582.

The original regression test was added to test_litellm_core_utils_prompt_templates_factory.py, which has pre-existing ruff-format violations throughout (multi-line asserts that fit on one line). The lint workflow runs `ruff format --check` on changed files only, so touching that file surfaces those pre-existing violations and fails CI for unrelated reasons. Move the #31582 regression coverage into a new dedicated test file so the format check stays green. Also collapses the helper's `not any(...)` onto a single line to satisfy ruff format. Covers: #31582

CLAassistant · 2026-07-02T00:15:33Z

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ mateo-berri
❌ ly-wang19
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

mateo-berri · 2026-07-02T00:15:34Z

@greptileai

Generated by Claude Code

greptile-apps · 2026-07-02T00:18:15Z

Greptile Summary

This PR fixes a 400 error on Bedrock Converse when calling Claude Opus 4.7/4.8 with strict: true in a tool definition, caused by Bedrock's validator rejecting toolSpec.strict for those models even though other Anthropic models accept it.

Replaces the broad get_bedrock_base_model(model).startswith("anthropic") gate in factory.py with a new bedrock_converse_supports_strict_tools helper in bedrock/common_utils.py; the helper reads a bedrock_converse_supports_strict_tools: false flag from model_prices_and_context_window.json, following the same JSON-driven pattern already used by bedrock_output_config_effort_ceiling.
Adds the flag to all Opus 4.7/4.8 regional entries (anthropic, global, us, eu, au, jp) in both the primary and backup JSON files, and plumbs the field through ProviderSpecificModelInfo and _get_model_info_helper.
Includes 18 new regression tests covering the Opus 4.7/4.8 exclusion, other-Anthropic pass-through, and non-Anthropic exclusion paths — all using local data, no network calls.

Confidence Score: 5/5

Safe to merge — the change is narrowly scoped, follows an established pattern in the codebase, and is backed by a purpose-built regression test suite.

The fix correctly replaces a too-broad provider check with a JSON-driven capability flag, using the same two-phase lookup (model_info dict → local cost map via key) already in production for bedrock_output_config_effort_ceiling. All affected Opus 4.7/4.8 regional entries in both JSON files carry the new flag, and the new helper defaults to true for models not explicitly listed — preserving existing behavior for every other Anthropic model on Bedrock. Tests cover the exclusion, inclusion, and non-Anthropic paths without any network calls.

No files require special attention.

Important Files Changed

Filename	Overview
litellm/llms/bedrock/common_utils.py	Adds `bedrock_converse_supports_strict_tools` and its private helper; implementation mirrors the existing `_get_bedrock_output_config_effort_ceiling` pattern and is JSON-driven.
litellm/litellm_core_utils/prompt_templates/factory.py	Replaces the `get_bedrock_base_model(model).startswith("anthropic")` gate with the new `bedrock_converse_supports_strict_tools` helper; call site is clean.
model_prices_and_context_window.json	Adds `bedrock_converse_supports_strict_tools: false` to all Opus 4.7/4.8 regional entries; correct coverage across anthropic, global, us, eu, au, jp prefixes.
litellm/model_prices_and_context_window_backup.json	Mirror of the main JSON changes; backup kept in sync.
litellm/types/utils.py	Adds `bedrock_converse_supports_strict_tools: Optional[bool]` to `ProviderSpecificModelInfo` TypedDict; correct typing.
litellm/utils.py	Plumbs `bedrock_converse_supports_strict_tools` from the JSON cost map into the `_get_model_info_helper` return dict; follows the same pattern as other capability flags.
tests/test_litellm/litellm_core_utils/prompt_templates/test_bedrock_converse_strict_tools_opus_47_48.py	New regression suite; covers Opus 4.7/4.8 exclusion, other-Anthropic inclusion, and non-Anthropic exclusion; all tests use local data without network calls.
tests/test_litellm/test_utils.py	Adds `bedrock_converse_supports_strict_tools` to the JSON schema validator; no change to existing assertions.

_{Reviews (4): Last reviewed commit: "chore: revert unrelated credential_migra..." | Re-trigger Greptile}

codecov · 2026-07-02T00:21:21Z

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

yucheng-berri · 2026-07-02T00:35:45Z

is the greptile comment legit?

Replace the hardcoded Opus 4.7/4.8 pattern list with a bedrock_converse_supports_strict_tools flag on the affected entries in model_prices_and_context_window.json, resolved via get_model_info with a local cost map fallback, so future models with the same restriction only need a JSON update

mateo-berri · 2026-07-02T00:46:25Z

@yucheng-berri yes, the P1 is legit. The repo already stores this kind of Bedrock capability info in model_prices_and_context_window.json (see bedrock_output_config_effort_ceiling, read via get_model_info in the same file), and there is a team rule in Greptile's context prohibiting hardcoded model flags. Addressed in 3652798: the Opus 4.7/4.8 entries now carry bedrock_converse_supports_strict_tools: false in the cost map and the helper resolves it through get_model_info with a local cost map fallback, so future models with the same restriction only need a JSON update

Generated by Claude Code

mateo-berri · 2026-07-02T00:46:30Z

@greptileai

Generated by Claude Code

mateo-berri · 2026-07-02T00:55:28Z

@greptileai

Generated by Claude Code

mateo-berri · 2026-07-02T01:01:35Z

bugbot run

Generated by Claude Code

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 6bd1f9a. Configure here.}

…1.91.0) litellm 1.90.x forwards the OpenAI `strict` tool field into the Bedrock Converse toolSpec for Anthropic models (PR BerriAI/litellm#29814, gated on get_bedrock_base_model(model).startswith("anthropic")). Converse rejects it for Opus 4.7/4.8: `tools.0.custom.strict: Extra inputs are not permitted` -> every tool-calling scan 400s at warm-up. The openai-agents SDK stamps strict on every FunctionTool (ensure_strict_json_schema), so it fires universally. UPSTREAM ALREADY FIXED THIS: BerriAI/litellm#31923 + #31943 (merged 2026-07-02, closing litellm#31582 which is this exact symptom) replace the broad startswith("anthropic") gate with bedrock_converse_supports_strict_tools. The fix ships in litellm 1.91.0. We pin 1.90.2 today, which predates it, so this monkeypatch (AmazonConverseConfig._transform_request_helper strips `strict`) bridges until we bump litellm. REMOVE THIS FILE + its apply() call in models.py once litellm>=1.91.0 is the floor. Idempotent, fail-open. Not an upstream candidate (upstream fixed it). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…#31582) (BerriAI#31923) * fix(bedrock/converse): drop toolSpec.strict for Opus 4.7/4.8 Bedrock Converse routes Claude Opus 4.7/4.8 through an Anthropic-compatible validator that maps toolSpec to the native tool shape and rejects the extra `strict` key with `tools.N.custom.strict: Extra inputs are not permitted`, even though Anthropic's native API accepts `strict` as a top-level tool field for the same models. Sonnet 4.5/4.6 and Opus <=4.6 accept `toolSpec.strict` unchanged. The existing gate `get_bedrock_base_model(model).startswith("anthropic")` (introduced in BerriAI#29814 to forward `strict` for Claude on Bedrock Converse) is too broad and regressed Opus 4.7/4.8 callers — see BerriAI#31582. Replace the inline check with a small `bedrock_converse_supports_strict_tools` helper that excludes the Opus 4.7/4.8 family from strict forwarding. All other Anthropic models on Bedrock keep the existing behavior. Closes BerriAI#31582. * fix(bedrock/converse): move strict-tools regression to a clean test file The original regression test was added to test_litellm_core_utils_prompt_templates_factory.py, which has pre-existing ruff-format violations throughout (multi-line asserts that fit on one line). The lint workflow runs `ruff format --check` on changed files only, so touching that file surfaces those pre-existing violations and fails CI for unrelated reasons. Move the BerriAI#31582 regression coverage into a new dedicated test file so the format check stays green. Also collapses the helper's `not any(...)` onto a single line to satisfy ruff format. Covers: BerriAI#31582 * refactor(bedrock/converse): drive strict-tools gate from model cost map Replace the hardcoded Opus 4.7/4.8 pattern list with a bedrock_converse_supports_strict_tools flag on the affected entries in model_prices_and_context_window.json, resolved via get_model_info with a local cost map fallback, so future models with the same restriction only need a JSON update * chore: revert unrelated credential_migration.py reformat --------- Co-authored-by: ly-wang19 <ly-wang19@users.noreply.github.com>

chore(release): backport #31923, #31929, #31393 to stable/1.90.x and cut 1.90.3

chore(release): backport #31912/#31920/#31921 (+#31923/#31929 parity, #31635 prereq) onto patch-1.91.0rc1

….3) (#1400) 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)) | patch | `v1.90.2` → `v1.90.3` | --- ### Release Notes <details> <summary>BerriAI/litellm (ghcr.io/berriai/litellm)</summary> ### [`v1.90.3`](https://github.com/BerriAI/litellm/releases/tag/v1.90.3) [Compare Source](BerriAI/litellm@v1.90.3...v1.90.3) ##### 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`](BerriAI/litellm@0112e53). **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:v1.90.3 ``` **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/v1.90.3/cosign.pub \ ghcr.io/berriai/litellm:v1.90.3 ``` 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 - chore(release): backport [#31923](BerriAI/litellm#31923), [#31929](BerriAI/litellm#31929), [#31393](BerriAI/litellm#31393) to stable/1.90.x and cut 1.90.3 by [@mateo-berri](https://github.com/mateo-berri) in [#32025](BerriAI/litellm#32025) **Full Changelog**: <BerriAI/litellm@v1.90.2...v1.90.3> ### [`v1.90.3`](https://github.com/BerriAI/litellm/releases/tag/v1.90.3) [Compare Source](BerriAI/litellm@v1.90.2...v1.90.3) ##### 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`](BerriAI/litellm@0112e53). **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:v1.90.3 ``` **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/v1.90.3/cosign.pub \ ghcr.io/berriai/litellm:v1.90.3 ``` 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 - chore(release): backport [#31923](BerriAI/litellm#31923), [#31929](BerriAI/litellm#31929), [#31393](BerriAI/litellm#31393) to stable/1.90.x and cut 1.90.3 by [@mateo-berri](https://github.com/mateo-berri) in [#32025](BerriAI/litellm#32025) **Full Changelog**: <BerriAI/litellm@v1.90.2...v1.90.3> </details> --- ### Configuration 📅 **Schedule**: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined). 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about these updates again. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR has been generated by [Renovate Bot](https://github.com/renovatebot/renovate).  Reviewed-on: https://git.erwanleboucher.dev/eleboucher/homelab/pulls/1400

This PR contains the following updates: | Package | Type | Update | Change | |---|---|---|---| | [ghcr.io/berriai/litellm](https://images.chainguard.dev/directory/image/wolfi-base/overview) ([source](https://github.com/BerriAI/litellm)) | final | patch | `v1.90.2` → `v1.90.3` | --- ### Release Notes <details> <summary>BerriAI/litellm (ghcr.io/berriai/litellm)</summary> ### [`v1.90.3`](https://github.com/BerriAI/litellm/releases/tag/v1.90.3) [Compare Source](BerriAI/litellm@v1.90.3...v1.90.3) ##### 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`](BerriAI/litellm@0112e53). **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:v1.90.3 ``` **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/v1.90.3/cosign.pub \ ghcr.io/berriai/litellm:v1.90.3 ``` 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 - chore(release): backport [#31923](BerriAI/litellm#31923), [#31929](BerriAI/litellm#31929), [#31393](BerriAI/litellm#31393) to stable/1.90.x and cut 1.90.3 by [@mateo-berri](https://github.com/mateo-berri) in [#32025](BerriAI/litellm#32025) **Full Changelog**: <BerriAI/litellm@v1.90.2...v1.90.3> ### [`v1.90.3`](https://github.com/BerriAI/litellm/releases/tag/v1.90.3) [Compare Source](BerriAI/litellm@v1.90.2...v1.90.3) ##### 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`](BerriAI/litellm@0112e53). **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:v1.90.3 ``` **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/v1.90.3/cosign.pub \ ghcr.io/berriai/litellm:v1.90.3 ``` 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 - chore(release): backport [#31923](BerriAI/litellm#31923), [#31929](BerriAI/litellm#31929), [#31393](BerriAI/litellm#31393) to stable/1.90.x and cut 1.90.3 by [@mateo-berri](https://github.com/mateo-berri) in [#32025](BerriAI/litellm#32025) **Full Changelog**: <BerriAI/litellm@v1.90.2...v1.90.3> </details> --- ### Configuration 📅 **Schedule**: (UTC) - Branch creation - At any time (no schedule defined) - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about this update again. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR has been generated by [Mend Renovate](https://github.com/renovatebot/renovate).  Co-authored-by: Renovate Bot <renovate@bhamm-lab.com> Reviewed-on: https://codeberg.org/blake-hamm/bhamm-lab/pulls/258

…to v1.90.3 (#257) This PR contains the following updates: | Package | Update | Change | |---|---|---| | [https://github.com/BerriAI/litellm.git](https://github.com/BerriAI/litellm) | patch | `v1.90.0` → `v1.90.3` | --- ### Release Notes <details> <summary>BerriAI/litellm (https://github.com/BerriAI/litellm.git)</summary> ### [`v1.90.3`](https://github.com/BerriAI/litellm/releases/tag/v1.90.3) [Compare Source](BerriAI/litellm@v1.90.2...v1.90.3) #### 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`](BerriAI/litellm@0112e53). **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:v1.90.3 ``` **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/v1.90.3/cosign.pub \ ghcr.io/berriai/litellm:v1.90.3 ``` 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 - chore(release): backport [#31923](BerriAI/litellm#31923), [#31929](BerriAI/litellm#31929), [#31393](BerriAI/litellm#31393) to stable/1.90.x and cut 1.90.3 by [@mateo-berri](https://github.com/mateo-berri) in [#32025](BerriAI/litellm#32025) **Full Changelog**: <BerriAI/litellm@v1.90.2...v1.90.3> ### [`v1.90.2`](https://github.com/BerriAI/litellm/releases/tag/v1.90.2) [Compare Source](BerriAI/litellm@v1.90.1...v1.90.2) #### 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`](BerriAI/litellm@0112e53). **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:v1.90.2 ``` **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/v1.90.2/cosign.pub \ ghcr.io/berriai/litellm:v1.90.2 ``` 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 - chore(release): backport [#31519](BerriAI/litellm#31519), [#31733](BerriAI/litellm#31733) to stable/1.90.x and cut 1.90.2 by [@yuneng-berri](https://github.com/yuneng-berri) in [#31782](BerriAI/litellm#31782) **Full Changelog**: <BerriAI/litellm@v1.90.1...v1.90.2> ### [`v1.90.1`](https://github.com/BerriAI/litellm/releases/tag/v1.90.1) [Compare Source](BerriAI/litellm@v1.90.0-rc.1...v1.90.1) #### 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`](BerriAI/litellm@0112e53). **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:v1.90.1 ``` **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/v1.90.1/cosign.pub \ ghcr.io/berriai/litellm:v1.90.1 ``` 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 - chore(release): backport [#31036](BerriAI/litellm#31036), [#31342](BerriAI/litellm#31342), [#31653](BerriAI/litellm#31653) to stable/1.90.x and cut 1.90.1 (litellm-enterprise 0.1.43.post1) by [@yuneng-berri](https://github.com/yuneng-berri) in [#31667](BerriAI/litellm#31667) **Full Changelog**: <BerriAI/litellm@v1.90.0...v1.90.1> </details> --- ### Configuration 📅 **Schedule**: (UTC) - Branch creation - At any time (no schedule defined) - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about this update again. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR has been generated by [Mend Renovate](https://github.com/renovatebot/renovate).  Co-authored-by: Renovate Bot <renovate@bhamm-lab.com> Reviewed-on: https://codeberg.org/blake-hamm/bhamm-lab/pulls/257

…#31582) (BerriAI#31923) * fix(bedrock/converse): drop toolSpec.strict for Opus 4.7/4.8 Bedrock Converse routes Claude Opus 4.7/4.8 through an Anthropic-compatible validator that maps toolSpec to the native tool shape and rejects the extra `strict` key with `tools.N.custom.strict: Extra inputs are not permitted`, even though Anthropic's native API accepts `strict` as a top-level tool field for the same models. Sonnet 4.5/4.6 and Opus <=4.6 accept `toolSpec.strict` unchanged. The existing gate `get_bedrock_base_model(model).startswith("anthropic")` (introduced in BerriAI#29814 to forward `strict` for Claude on Bedrock Converse) is too broad and regressed Opus 4.7/4.8 callers — see BerriAI#31582. Replace the inline check with a small `bedrock_converse_supports_strict_tools` helper that excludes the Opus 4.7/4.8 family from strict forwarding. All other Anthropic models on Bedrock keep the existing behavior. Closes BerriAI#31582. * fix(bedrock/converse): move strict-tools regression to a clean test file The original regression test was added to test_litellm_core_utils_prompt_templates_factory.py, which has pre-existing ruff-format violations throughout (multi-line asserts that fit on one line). The lint workflow runs `ruff format --check` on changed files only, so touching that file surfaces those pre-existing violations and fails CI for unrelated reasons. Move the BerriAI#31582 regression coverage into a new dedicated test file so the format check stays green. Also collapses the helper's `not any(...)` onto a single line to satisfy ruff format. Covers: BerriAI#31582 * refactor(bedrock/converse): drive strict-tools gate from model cost map Replace the hardcoded Opus 4.7/4.8 pattern list with a bedrock_converse_supports_strict_tools flag on the affected entries in model_prices_and_context_window.json, resolved via get_model_info with a local cost map fallback, so future models with the same restriction only need a JSON update * chore: revert unrelated credential_migration.py reformat --------- Co-authored-by: ly-wang19 <ly-wang19@users.noreply.github.com> (cherry picked from commit 85f9241)

…et 4.6 and Haiku 4.5 Bedrock Converse rejects toolSpec.strict for Claude Sonnet 4.6 and Claude Haiku 4.5, causing a grammar-size error (>300MB compiled grammar) that results in a 504 Gateway Timeout for callers. The same fix was already applied for Opus 4.7/4.8 (BerriAI#31923) and Sonnet 4 (BerriAI#31943). This extends it to the missing model variants by setting bedrock_converse_supports_strict_tools: false on all regional Bedrock Converse entries for these two models. Fixes BerriAI#34388 Co-authored-by: Hermes Agent <hermes-agent@nousresearch.com>

ly-wang19 added 2 commits July 2, 2026 00:06

greptile-apps Bot reviewed Jul 2, 2026

View reviewed changes

Comment thread litellm/llms/bedrock/common_utils.py Outdated

mateo-berri marked this pull request as ready for review July 2, 2026 00:36

chore: revert unrelated credential_migration.py reformat

6bd1f9a

cursor Bot reviewed Jul 2, 2026

View reviewed changes

yucheng-berri approved these changes Jul 2, 2026

View reviewed changes

mateo-berri merged commit 85f9241 into litellm_internal_staging Jul 2, 2026
127 checks passed

mateo-berri deleted the litellm_fix_bedrock_opus_47_strict_tools branch July 2, 2026 02:08

mateo-berri mentioned this pull request Jul 3, 2026

chore(release): backport #31923, #31929, #31393 to stable/1.90.x and cut 1.90.3 #32025

Merged

2 tasks

yuneng-berri added a commit that referenced this pull request Jul 3, 2026

Merge pull request #32025 from BerriAI/litellm_cherrypick_1_90_x

2d28d8f

chore(release): backport #31923, #31929, #31393 to stable/1.90.x and cut 1.90.3

yuneng-berri mentioned this pull request Jul 4, 2026

chore(release): backport #31912/#31920/#31921 (+#31923/#31929 parity, #31635 prereq) onto patch-1.91.0rc1 #32091

Merged

yuneng-berri added a commit that referenced this pull request Jul 4, 2026

Merge pull request #32091 from BerriAI/litellm_backport_191_rc1

465a7b2

chore(release): backport #31912/#31920/#31921 (+#31923/#31929 parity, #31635 prereq) onto patch-1.91.0rc1

seanturner83 mentioned this pull request Jul 8, 2026

fix(bedrock): drop toolSpec.strict for Claude Sonnet 5 and Fable 5 #32455

Open

3 tasks

seanturner83 mentioned this pull request Jul 21, 2026

[Bug]: Bedrock Converse fails with parallel_tool_calls on Claude 4.5+ #22637

Open

1 task

luiscastejon-aily mentioned this pull request Jul 23, 2026

LiteLLM forwards unsupported strict tool fields to Bedrock Claude models, causing 400 errors and hanging requests #34388

Open

1 task

baiyuxi930826 mentioned this pull request Jul 25, 2026

fix(bedrock): drop toolSpec.strict for Haiku 4.5 and Sonnet 4.6 on Converse #34609

Open

3 tasks

kr0k mentioned this pull request Jul 31, 2026

fix(bedrock): drop toolSpec.strict for Claude Sonnet 5 on Converse #33196

Merged

4 tasks

nickhac mentioned this pull request Aug 3, 2026

fix(bedrock): drop strict/additionalProperties from toolSpec for Sonnet 4.6 and Haiku 4.5 #35617

Open

5 tasks

This was referenced Aug 4, 2026

fix(proxy): retry Anthropic tools without strict when Bedrock rejects it farion1231/cc-switch#6118

Open

fix(proxy): retry Anthropic tools without strict farion1231/cc-switch#6121

Open

Navigation Menu

Uh oh!

Uh oh!

fix(bedrock/converse): drop toolSpec.strict for Opus 4.7/4.8 (#31582) - #31923

fix(bedrock/converse): drop toolSpec.strict for Opus 4.7/4.8 (#31582)#31923
mateo-berri merged 4 commits into
litellm_internal_stagingfrom
litellm_fix_bedrock_opus_47_strict_tools

mateo-berri commented Jul 2, 2026 •

edited

Loading

Uh oh!

CLAassistant commented Jul 2, 2026 •

edited

Loading

Uh oh!

mateo-berri commented Jul 2, 2026

Uh oh!

greptile-apps Bot commented Jul 2, 2026 •

edited

Loading

Important Files Changed

Uh oh!

Uh oh!

codecov Bot commented Jul 2, 2026 •

edited

Loading

Uh oh!

yucheng-berri commented Jul 2, 2026

Uh oh!

mateo-berri commented Jul 2, 2026

Uh oh!

mateo-berri commented Jul 2, 2026

Uh oh!

mateo-berri commented Jul 2, 2026

Uh oh!

mateo-berri commented Jul 2, 2026

Uh oh!

cursor Bot left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Uh oh!

Uh oh!

Conversation

mateo-berri commented Jul 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Relevant issues

Linear ticket

Pre-Submission checklist

Delays in PR merge?

Screenshots / Proof of Fix

Type

Changes

Uh oh!

CLAassistant commented Jul 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mateo-berri commented Jul 2, 2026

Uh oh!

greptile-apps Bot commented Jul 2, 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!

Uh oh!

codecov Bot commented Jul 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

yucheng-berri commented Jul 2, 2026

Uh oh!

mateo-berri commented Jul 2, 2026

Uh oh!

mateo-berri commented Jul 2, 2026

Uh oh!

mateo-berri commented Jul 2, 2026

Uh oh!

mateo-berri commented Jul 2, 2026

Uh oh!

cursor Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

mateo-berri commented Jul 2, 2026 •

edited

Loading

CLAassistant commented Jul 2, 2026 •

edited

Loading

greptile-apps Bot commented Jul 2, 2026 •

edited

Loading

codecov Bot commented Jul 2, 2026 •

edited

Loading