Implement CrewAI LLM adapter for mellea backends

[psschwei]

Summary

Create a CrewAI-compatible LLM adapter, potentially leveraging the LangChain integration since CrewAI often uses LangChain underneath.

Problem / Opportunity

CrewAI is a popular framework for multi-agent workflows. CrewAI users want to use mellea's optimized backends for their agent crews. Since CrewAI often uses LangChain under the hood, this may be achievable with minimal additional work.

User Story / Use Case

As a CrewAI user, I want to use mellea backends for my agent crews so that I can benefit from mellea's optimizations in my multi-agent workflows.

Proposed Solution (High-level)

Create mellea/integrations/crewai/llm.py containing either:

• Direct adapter if CrewAI has its own LLM interface, OR
• Documentation and helpers for using the LangChain adapter with CrewAI
• Configuration helpers for CrewAI-specific settings

Scope

In scope:

• Adapter or documentation for using mellea with CrewAI agents
• Configuration helpers for CrewAI-specific settings
• Integration tests with basic CrewAI workflows

Out of scope:

• CrewAI-specific agent logic
• Custom CrewAI tools (use tool converters separately)
• CrewAI memory management

Acceptance Criteria

• Works as llm parameter in CrewAI Agent
• Supports CrewAI's expected callback interface
• Documentation for CrewAI-specific configuration

[akihikokuroda]

I'll look into this.