Overview
An intelligent assistant for power trading. It connects multiple large language models, uses Function Calling and MCP to reach local business data and external sources, and improves answer accuracy with a vectorized knowledge base and role-based conversations.
Highlights
Model Access and Local Tools
- Connect multiple large language models and switch capabilities by business scenario.
- Wrap local data APIs as tools with Function Calling so the model can query business data on demand.
MCP and Knowledge Base Expansion
- Introduce MCP so the model can call external data sources beyond local APIs.
- Process business documents with embedding models to expand the knowledge base for retrieval-augmented answers.
Role-based Collaboration and Cross-validation
- Assign LLM roles in a conversation to handle specific types of data tasks.
- Have roles cross-check each other to improve the accuracy of power-trading data.