fix: support anthropic-- prefix in model name filtering

[C1-BA-B1-F3]

Problem

The _matches_provider_pattern method only recognizes claude- and anthropic. prefixes for Anthropic models. Self-hosted deployments using naming conventions like anthropic--claude-... are incorrectly rejected or routed to the wrong provider.

Fixes #5893

Changes

lib/crewai/src/crewai/llm.py (2 changes):

1. _matches_provider_pattern: Added "anthropic--" to the Anthropic prefix list so models like anthropic--claude-3-5-sonnet pass validation.

2. _infer_provider_from_model: Added a pattern-based fallback after the constants check, so unprefixed models with non-standard Anthropic naming conventions (e.g. anthropic--claude-3-5-sonnet) are correctly inferred as Anthropic instead of defaulting to OpenAI.

lib/crewai/tests/test_llm.py (3 new tests):

• test_anthropic_double_dash_prefix_uses_native_sdk — end-to-end routing test
• test_infer_provider_from_model_anthropic_prefixes — provider inference test
• Extended test_validate_model_in_constants with anthropic-- prefix assertions

Testing

All 3 new tests pass, plus all existing anthropic-related tests remain green.

Summary by CodeRabbit

• Bug Fixes

  • Improved model-provider detection for additional Anthropic model name formats, including double-dash prefixes.
  • Better fallback provider inference now recognizes more Anthropic-style model names instead of defaulting incorrectly.
  • Preserved the original model name while using the native Anthropic path when applicable.

• Tests

  • Added coverage for Anthropic double-dash model prefixes and provider inference behavior.
  • Expanded validation checks for supported Anthropic model naming patterns.

[corridor-security]

Summary: This PR updates LLM provider detection to recognize an additional Anthropic model-name prefix and adds tests for that behavior.

Risk: Low risk. The changes do not introduce new authentication, authorization, file handling, subprocess execution, public endpoints, or user-controlled network targets, and no exploitable security vulnerabilities were identified.

[coderabbitai]

📝 Walkthrough

Walkthrough

LLM now recognizes anthropic-- as an Anthropic prefix in provider matching and inference. Tests were added to cover native Anthropic routing and validation for anthropic-- model names.

Changes

Anthropic prefix handling

Layer / File(s)	Summary
Provider prefix handling and tests lib/crewai/src/crewai/llm.py, lib/crewai/tests/test_llm.py	LLM accepts anthropic-- in Anthropic prefix checks and fallback inference, and tests cover native Anthropic routing plus validation of anthropic-- model names.

🚥 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 clearly summarizes the main change: supporting the anthropic-- model prefix in model-name filtering.
Linked Issues check	✅ Passed	The changes address #5893 by broadening Anthropic prefix matching and provider inference, with tests covering the new naming pattern.
Out of Scope Changes check	✅ Passed	The PR stays focused on Anthropic prefix handling and related tests, with no obvious unrelated changes.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ 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.}

[coderabbitai]

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (1)

lib/crewai/src/crewai/llm.py (1)

701-712: 📐 Maintainability & Code Quality | 🟡 Minor

Extend _is_anthropic_model to match broadened Anthropic patterns.

_matches_provider_pattern and _infer_provider_from_model recognize anthropic-- and anthropic. as Anthropic identifiers, but _is_anthropic_model only checks ("anthropic/", "claude-", "claude/"). This causes models like anthropic--my-deploy (without claude) to incorrectly return False, blocking the litellm fallback's Anthropic-specific "first message must be user" handling.

Update lib/crewai/src/crewai/llm.py lines 701‑712:

✅ Suggested fix

-        anthropic_prefixes = ("anthropic/", "claude-", "claude/")
+        anthropic_prefixes = ("anthropic/", "anthropic--", "anthropic.", "claude-", "claude/")

🤖 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 701 - 712, The _is_anthropic_model
helper in crewai.llm currently misses the broader Anthropic naming variants
already handled elsewhere, so models like anthropic--my-deploy and
anthropic.my-deploy return false. Update the prefix check in _is_anthropic_model
to align with _matches_provider_pattern and _infer_provider_from_model by
recognizing the Anthropic-specific forms, while keeping the existing
claude-based matches. This will ensure litellm fallback logic in the LLM path
correctly applies Anthropic-only handling for first-message validation.

🤖 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.

Outside diff comments:
In `@lib/crewai/src/crewai/llm.py`:
- Around line 701-712: The _is_anthropic_model helper in crewai.llm currently
misses the broader Anthropic naming variants already handled elsewhere, so
models like anthropic--my-deploy and anthropic.my-deploy return false. Update
the prefix check in _is_anthropic_model to align with _matches_provider_pattern
and _infer_provider_from_model by recognizing the Anthropic-specific forms,
while keeping the existing claude-based matches. This will ensure litellm
fallback logic in the LLM path correctly applies Anthropic-only handling for
first-message validation.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: fdcfa8e5-9f55-4292-b2fb-0f547b2d6be0

📥 Commits

Reviewing files that changed from the base of the PR and between f364a7d and 7bad52a.

📒 Files selected for processing (2)

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