fix(llm): detect Anthropic models with custom deployment naming

[HumphreySun98]

Problem

Provider detection only recognized the canonical claude- / anthropic. prefixes. A self-deployed Anthropic model named e.g. anthropic--claude-3-5-sonnet was therefore:

• mis-detected by _infer_provider_from_model, which defaulted unknown unprefixed ids to OpenAI, and
• not matched by _matches_provider_pattern,

so it was routed to the OpenAI adapter, losing all Anthropic-specific request handling (stop-sequence normalization, thinking blocks, prompt caching, etc.).

Fix

• Broaden the Anthropic pattern in _matches_provider_pattern to match claude anywhere or an anthropic prefix.
• Let _infer_provider_from_model fall back to the Anthropic/Gemini patterns before defaulting to OpenAI.

anthropic is matched via startswith, so lookalikes like anthropomorphic are unaffected; broad patterns such as bedrock's bare . match are intentionally excluded to avoid false positives.

Tests

Extended test_validate_model_in_constants and added test_infer_provider_from_model_custom_naming in test_llm.py, including no-regression and no-false-positive cases. Verified failing-without/passing-with the fix; the 20 existing model-routing tests still pass; ruff and mypy clean.

Fixes #5893

---

This PR was authored with Claude Code. Per CONTRIBUTING.md, AI-generated contributions require the llm-generated label — I don't have triage permission to set it, so could a maintainer please add it? 🤖 Generated with Claude Code

Summary by CodeRabbit

Release Notes

• Bug Fixes

  • Improved recognition of Anthropic Claude model names, now supporting custom deployment naming formats and names containing "claude" in any position.
  • Enhanced provider routing to better identify unprefixed Anthropic and Gemini model identifiers, reducing misrouting to OpenAI.

• Tests

  • Added test coverage for custom model naming patterns and provider inference logic.

[coderabbitai]

📝 Walkthrough

Walkthrough

_matches_provider_pattern in LLM is updated to match any model name containing "claude" (not just strict "claude-" or "anthropic." prefixes). _infer_provider_from_model now probes Anthropic and Gemini patterns before defaulting to "openai". Tests are added to cover custom deployment naming variants.

Changes

Provider Pattern Matching Fix

Layer / File(s)	Summary
Broadened Anthropic matching and unprefixed provider inference lib/crewai/src/crewai/llm.py	_matches_provider_pattern returns True for any model name containing "claude" (e.g., anthropic--claude-..., anthropic.claude-...), replacing the previous strict prefix check. _infer_provider_from_model now calls _matches_provider_pattern for "anthropic" and "gemini" before falling back to "openai".
Tests for custom naming and provider routing lib/crewai/tests/test_llm.py	test_validate_model_in_constants is extended to assert double-dash and dot Anthropic naming variants are accepted and that non-Anthropic lookalikes (e.g., anthropomorphic) are rejected. New test_infer_provider_from_model_custom_naming asserts anthropic resolution for custom variants, gemini for gemini-shaped ids, and openai fallback for unknown or genuine OpenAI ids.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately summarizes the main change: fixing detection of Anthropic models with custom deployment naming conventions.
Linked Issues check	✅ Passed	The PR directly addresses issue #5893 by updating the filtering functions to support wider model naming conventions while maintaining correct provider routing.
Out of Scope Changes check	✅ Passed	All changes focus on broadening Anthropic model detection logic and testing; no unrelated modifications are present.
Docstring Coverage	✅ Passed	Docstring coverage is 80.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@lib/crewai/src/crewai/llm.py`:
- Around line 510-516: The provider detection logic at the diff location has
been expanded to check for `model_lower.startswith("anthropic")`, but the
`_is_anthropic_model` method does not include this same pattern matching logic.
Update the `_is_anthropic_model` method to mirror the expanded Anthropic
matching rule by adding a check for `model_lower.startswith("anthropic")`
alongside the existing pattern checks. This will ensure that both the provider
inference and the `is_anthropic` attribute remain consistent, allowing
Anthropic-specific message formatting in `_format_messages_for_provider` to be
properly applied.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: d9473017-151b-42bc-97ed-06b1a168b440

📥 Commits

Reviewing files that changed from the base of the PR and between e9d568d and 4be3753.

📒 Files selected for processing (2)

• lib/crewai/src/crewai/llm.py
• lib/crewai/tests/test_llm.py

[coderabbitai]

⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Align Anthropic pattern logic with is_anthropic detection to prevent inconsistent routing behavior.

Line 515 expands Anthropic matching to startswith("anthropic"), but _is_anthropic_model (Line 707-717) does not mirror that rule. This creates inconsistent state where provider inference can resolve to Anthropic while self.is_anthropic stays False, skipping Anthropic-specific message handling in _format_messages_for_provider (Line 2326-2333).

Proposed fix

@@
     `@staticmethod`
     def _is_anthropic_model(model: str) -> bool:
@@
-        anthropic_prefixes = ("anthropic/", "claude-", "claude/")
-        return any(prefix in model.lower() for prefix in anthropic_prefixes)
+        model_lower = model.lower()
+        return "claude" in model_lower or model_lower.startswith("anthropic")

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@lib/crewai/src/crewai/llm.py` around lines 510 - 516, The provider detection
logic at the diff location has been expanded to check for
`model_lower.startswith("anthropic")`, but the `_is_anthropic_model` method does
not include this same pattern matching logic. Update the `_is_anthropic_model`
method to mirror the expanded Anthropic matching rule by adding a check for
`model_lower.startswith("anthropic")` alongside the existing pattern checks.
This will ensure that both the provider inference and the `is_anthropic`
attribute remain consistent, allowing Anthropic-specific message formatting in
`_format_messages_for_provider` to be properly applied.