What is an MCP Server?

An MCP (Model Context Protocol) Server acts as a bridge, allowing AI models to access and interact with external data sources and tools. It provides context to LLMs, enhancing their ability to deliver accurate and relevant responses.

How does an MCP Server work?

MCP Servers facilitate connections to various data sources, enabling LLMs to retrieve and process external information. This enhances the LLM’s understanding of the context surrounding a query, improving the accuracy and relevance of responses.

What types of data sources can an MCP Server connect to?

MCP Servers can connect to a wide array of data sources, including databases, APIs, real-time data streams, and even local files like PDFs and CSVs.

What are the benefits of using an MCP Server?

Key benefits include improved accuracy, access to real-time data, enhanced contextual awareness, and the ability for LLMs to perform actions beyond simple information retrieval.

Can I use multiple vector stores with an MCP Server?

Yes, the architecture supports multiple vector stores, enabling you to manage and search different document sets independently. Each vector store is stored in a separate directory, specified via the vector_store_dir parameter.

How do I reuse a vector store in another project?

Copy the vector store directory to the new project. Ensure langchain-chroma, langchain-openai, and other dependencies are installed. Use the same embedding model and set the appropriate environment variables.

What is UBOS and how does it relate to MCP Servers?

UBOS is a full-stack AI Agent development platform. It leverages MCP Servers to connect AI Agents to enterprise data, enabling the development of custom AI Agents and Multi-Agent Systems.

How do I upload documents to the RAG API?

Use the /upload endpoint with a multipart/form-data request, including the files to upload and optionally specifying the vector_store_dir.

How do I query documents using the RAG API?

Use the /query endpoint with a form-data request, including the search query and optionally specifying the vector_store_dir and the number of documents to retrieve (k).

Where can I find the API documentation for the RAG API?

The API documentation is available at http://localhost:8000/docs and http://localhost:8000/redoc when the server is running.