-
Notifications
You must be signed in to change notification settings - Fork 30
Open
Description
Add Modular Support for Multiple Agent Frameworks and LLM Providers
Description
Implement a modular architecture to support multiple agent frameworks (LangChain, CrewAI, AutoGen) and LLM providers (OpenAI, Anthropic, Gemini) through a plugin-based system. This allows users to build agents using their preferred framework and LLM while maintaining interoperability through NEST's protocol layer.
Implementation Approach
- Create
AgentFrameworkabstract interface for framework adapters - Create
LLMProviderabstract interface for model providers - Implement framework adapters:
LangChainAdapter,CrewAIAdapter,AutoGenAdapter - Implement LLM provider adapters:
OpenAIProvider,AnthropicProvider,GeminiProvider - Config-driven framework and provider selection
- Framework adapters bridge to NEST's protocol router (SLIM/A2A)
Acceptance Criteria
-
AgentFrameworkabstract base class withinitialize(),process_message(),get_capabilities()methods -
LLMProviderabstract base class withgenerate(),stream(),get_models()methods - LangChain, CrewAI, AutoGen framework adapters implemented
- OpenAI, Anthropic, Gemini provider adapters implemented
- Config allows specifying framework and provider:
agent.framework,agent.llm_provider,agent.model - Framework adapters integrate with protocol router for inter-agent communication
- Agents built with different frameworks can communicate via common protocols
- Support for framework-specific features (LangChain tools, CrewAI roles, AutoGen conversations)
- Documentation for adding custom framework/provider adapters
- Tests for each framework and provider adapter
Metadata
Metadata
Assignees
Labels
No labels