https://github.com/rajagopal17/mcpRAG
rag using Ollama as emebddings, gemini as LLM and MCP server for agentic use
https://github.com/rajagopal17/mcpRAG
Last synced: about 2 months ago
JSON representation
rag using Ollama as emebddings, gemini as LLM and MCP server for agentic use
- Host: GitHub
- URL: https://github.com/rajagopal17/mcpRAG
- Owner: rajagopal17
- Created: 2025-04-20T09:10:45.000Z (7 months ago)
- Default Branch: main
- Last Pushed: 2025-04-21T15:21:56.000Z (7 months ago)
- Last Synced: 2025-04-21T16:30:49.052Z (7 months ago)
- Language: Python
- Size: 3.17 MB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
- awesome-mcp-servers - **mcpRAG** - rag using Ollama as emebddings, gemini as LLM and MCP server for agentic use `python` `mcp` `server` `llm` `pip install git+https://github.com/rajagopal17/mcpRAG` (🤖 AI/ML)
- awesome-mcp-servers - **mcpRAG** - rag using Ollama as emebddings, gemini as LLM and MCP server for agentic use `python` `mcp` `server` `llm` `pip install git+https://github.com/rajagopal17/mcpRAG` (AI/ML)
README
# mcpRAG
RAG- using opensource embeddings, opensource vector database and Gemini LLM
______________________________________________________________________________
In this project i have created RAG using txt documents:
Embeddings : 'nomic embeddings' are used with ollama locally
LLM : gemini-2.0-flash
Vector Database : FAISS
All the txt files are chunked with file name, chunk id and chunk text in JSON format and stored locally.
Each chunk is converted into embeddings and collected in a list
This embedding list is indexed using FAISS and stored locally.
when query is embedding using nomic embeddings, these embeddings are searched in FAISS index and relevant indices(location of chunk) is retrieved. These indices are passed to JSON file to get the actual text.
THis text is passed to LLM with the query to formulate the answer.
Additional text is appended to the exiting index and queries are run on the updated index by loading the stored index and embedding file.