feat(bedrock): add Anthropic Claude 4 support (fixes #5549)

[devin-ai-integration]

Summary

Fixes #5549. Adds Anthropic Claude 4 Bedrock cross-region inference profile model IDs to BEDROCK_MODELS (and the BedrockModels Literal type alias) in lib/crewai/src/crewai/llms/constants.py.

On AWS Bedrock, the Claude 4 family (Sonnet 4 / 4.5, Opus 4 / 4.1 / 4.5, Haiku 4.5) is only invocable through cross-region inference profile IDs (e.g. us.anthropic.claude-sonnet-4-5-20250929-v1:0, eu.anthropic.claude-sonnet-4-20250514-v1:0, global.anthropic.claude-haiku-4-5-20251001-v1:0), not the bare anthropic.claude-* IDs. Before this change those inference profile IDs were absent from BEDROCK_MODELS, so:

• LLM._infer_provider_from_model("us.anthropic.claude-sonnet-4-5-20250929-v1:0") returned the default "openai" instead of "bedrock".
• Users had to manually prefix the model with bedrock/ and/or specify provider="bedrock" to route the call correctly.

The fix lists the full Claude 4 inference-profile IDs across all Bedrock regional prefixes currently catalogued in LiteLLM's model_prices_and_context_window.json (us., us-gov., eu., apac., au., jp., global.).

Model IDs added (to both BedrockModels and BEDROCK_MODELS):

• us.anthropic.claude-sonnet-4-20250514-v1:0, us.anthropic.claude-sonnet-4-5-20250929-v1:0
• us.anthropic.claude-opus-4-20250514-v1:0, us.anthropic.claude-opus-4-1-20250805-v1:0, us.anthropic.claude-opus-4-5-20251101-v1:0
• us.anthropic.claude-haiku-4-5-20251001-v1:0
• eu.anthropic.claude-{sonnet,opus,haiku}-4-*-v1:0 variants
• apac.anthropic.claude-{sonnet,haiku}-4-*-v1:0, au.anthropic.claude-{sonnet,haiku}-4-5-*-v1:0, jp.anthropic.claude-{sonnet,haiku}-4-5-*-v1:0
• global.anthropic.claude-{sonnet,opus,haiku}-4-*-v1:0
• us-gov.anthropic.claude-sonnet-4-5-20250929-v1:0

Tests (lib/crewai/tests/test_llm.py):

• Extended test_validate_model_in_constants with Bedrock Claude 4 cross-region inference-profile assertions.
• Added test_bedrock_claude_4_models_in_constants verifying both direct IDs and inference profiles are present in BEDROCK_MODELS.
• Added test_infer_provider_for_bedrock_claude_4_models verifying _infer_provider_from_model returns "bedrock" for Claude 4 direct IDs and cross-region inference profiles.

All three tests pass locally (uv run pytest lib/crewai/tests/test_llm.py::test_validate_model_in_constants ::test_bedrock_claude_4_models_in_constants ::test_infer_provider_for_bedrock_claude_4_models). Ruff and mypy pass on the changed files.

Review & Testing Checklist for Human

• Confirm the set of Claude 4 inference-profile IDs matches what your AWS accounts actually expose (I sourced them from the current LiteLLM model_prices_and_context_window.json, which may lag real Bedrock availability by region). If any of these aren't yet generally available in your regions, they can still be listed here — they simply won't resolve at call-time — but flag any you'd rather defer.
• Intentional scope call: preview/future Claude 4.6 / 4.7 inference-profile IDs that appear in LiteLLM (e.g. us.anthropic.claude-opus-4-7, global.anthropic.claude-sonnet-4-6) are not included — they carry 2026 release dates and aren't generally available yet. If you'd like those added now, say the word.
• Verify with an actual Bedrock call that e.g. LLM(model="bedrock/us.anthropic.claude-sonnet-4-5-20250929-v1:0") completes successfully end-to-end. I only validated the static constants/provider-inference behavior, not a live Bedrock invocation (that requires AWS credentials and model access).

Notes

• No changes to _matches_provider_pattern for bedrock — its existing "." in model_lower fallback already accepts these IDs, so validation wasn't the blocker. The fix is specifically about _infer_provider_from_model and keeping the Literal type alias accurate.
• There is a pre-existing PR #5550 from the issue reporter targeting the same issue. It's based on a significantly stale main (diff deletes ~1,300 files), so this PR starts fresh against current main and adds the inference-profile IDs plus tests. Happy to close either side depending on your preference.

Link to Devin session: https://app.devin.ai/sessions/e0c7012d0e26488ab18765a172e71307

---

Note

Low Risk
Low risk constants/test update that only affects provider inference for specific Bedrock Claude 4 model ID strings, ensuring they no longer fall back to the default provider.

Overview
Fixes Bedrock provider inference for Anthropic Claude 4 by adding cross-region inference profile model IDs (e.g. us., eu., apac., global., etc.) to the BedrockModels literal and BEDROCK_MODELS list in constants.py.

Extends test_validate_model_in_constants and adds targeted tests to assert Claude 4 direct IDs + inference profiles are present in BEDROCK_MODELS and that LLM._infer_provider_from_model() returns "bedrock" for those IDs (preventing an "openai" fallback).

^{Reviewed by Cursor Bugbot for commit 04dcabc. Bugbot is set up for automated code reviews on this repo. Configure here.}

[devin-ai-integration]

Prompt hidden (unlisted session)

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring