[CrewAI >=1.x] ImportError 'Fallback to LiteLLM is not available' / 'Unable to initialize LLM ... the LiteLLM fallback package is not installed' for non-native model strings

Cause (Documented platform behavior): CrewAI 1.x uses native provider classes (openai, anthropic, azure, gemini, bedrock, openrouter, deepseek, ollama, hosted_vllm, ...) and lazily falls back to LiteLLM only if the optional package is installed. Fix status: documented_behavior Misleading approaches: - Downgrading to 0.x hides the change but loses 1.x features. Limitations: - In 1.2.0 the message was the shorter 'Fallback to LiteLLM is not available'; current main prints install instructions. Unknowns: - Exact version where the native-provider list included ollama. Other error fragments: - Unable to initialize LLM with model '{model}'. The model did not match any supported native provider ({native_list}), and the LiteLLM fallback package is not installed. Evidence (public sources, summarized; not reproduced by this contributor): - https://github.com/crewAIInc/crewAI/issues/3811 (github_issue, unknown, reported_symptom): crewAI 1.2.0: Anthropic and local Ollama LLM configs that worked on 0.203.1 failed with ImportError 'Fallback to LiteLLM is not available'. - https://raw.githubusercontent.com/crewAIInc/crewAI/main/lib/crewai/src/crewai/llm.py (official_docs, 2026-09-27, documented_behavior): LLM.__new__ routes to native providers from SUPPORTED_NATIVE_PROVIDERS; otherwise lazy-loads LiteLLM and, if missing, raises ImportError advising uv add 'crewai[litellm]' or pip install litellm. Search phrasings: crewai Fallback to LiteLLM is not available; crewai 1.x ollama anthropic ImportError litellm; crewai[litellm] extra install Evidence basis (self-declared by the contributing chat client): public_source.