Research & summarize
Index and Query Local Documents
Open-source RAG server for local document search and indexing, with ChatGPT and Claude integration for fully private AI-powered insights.
Without it
Stand up the Ollama server yourself — bot tokens or OAuth, a process to host and keep alive, updates to chase.
With it
Install once and it is wired straight into your agent.
What you get
- Index documents from various file types recursively.
- Perform vector search on indexed documents using Qdrant.
- Query local documents and receive answers via chatbot interfaces.
- Integrate with local or cloud-based LLMs for enhanced generation.
Connect it yourself
Use hosted
Use this with MCP Gate
Connect this and other MCP tools to your agents through one endpoint, with usage billed in one place.
Open source RAG (Retrieval Augmented Generation) server that provides document search and indexing capabilities on local servers, with support for ChatGPT integration and fully local operation.
Installation
Docker - Fully Local
docker compose -f docker-compose-ollama.yml --env-file .env up --build
Docker - With ChatGPT Support
docker compose -f docker-compose-chatgpt.yml --env-file .env up --build
Docker - MCP Integration
docker compose -f docker-compose-mcp.yml --env-file .env up --build
Smithery
npx -y @smithery/cli install minima --client claude
Shell Script
./run.sh
Configuration
Claude Desktop
{
"mcpServers": {
"minima": {
"command": "uv",
"args": [
"--directory",
"/path_to_cloned_minima_project/mcp-server",
"run",
"minima"
]
}
}
}
GitHub Copilot
{
"servers": {
"minima": {
"type": "stdio",
"command": "path_to_cloned_minima_project/run_in_copilot.sh",
"args": [
"path_to_cloned_minima_project"
]
}
}
}
Features
- Isolated installation - fully local with containers, no external dependencies
- Custom GPT integration - query local documents through ChatGPT with custom GPTs
- Anthropic Claude integration - use the Claude app to query local documents
- Support for multiple file types: .pdf, .xls, .docx, .txt, .md, .csv
- Recursive document indexing in folders and subfolders
- Vector search using Qdrant storage
- Support for reranker models for improved search relevance
- Electron app for fully local use
Environment Variables
Required
LOCAL_FILES_PATH- Root folder for document indexing (recursive process)EMBEDDING_MODEL_ID- Sentence Transformer model for embeddingsEMBEDDING_SIZE- Embedding dimensionality for Qdrant vector storage configuration
Optional
OLLAMA_MODEL- Ollama LLM model ID for local inferenceRERANKER_MODEL- Reranker model for improving search relevance (BAAI models tested)USER_ID- Email for ChatGPT integration authenticationPASSWORD- Password for ChatGPT integration authentication
Usage Examples
Ask any questions and you will get answers based on local files in the specified folder
Notes
Requires Python >=3.10 and installed 'uv' for MCP usage. Chat UI is available at http://localhost:3000 for local use. License: Mozilla Public License v2.0 (MPLv2).