Clawrag

What This Skill Does

ClawRAG acts as the cognitive engine for your OpenClaw agent, enabling it to retrieve information from a private, self-hosted knowledge base. By utilizing advanced hybrid search techniques—which combine traditional keyword (BM25) search with high-dimensional vector similarity—it ensures that OpenClaw finds relevant information even if the terminology doesn't match perfectly. It supports a variety of document formats, including PDFs, Microsoft Office documents, and Markdown files, making it an ideal solution for users wanting to query personal archives, technical manuals, or project-specific documentation without exposing sensitive data to third-party cloud RAG providers.

Installation

To integrate ClawRAG, first ensure Docker and Docker Compose are installed on your machine. Clone the official repository from 2dogsandanerd, copy the environment configuration file, and run 'docker compose up -d' to initialize the local server. Once the service is healthy at localhost:8080, register the connector with your OpenClaw environment by running the command: 'openclaw mcp add --transport stdio clawrag npx -y @clawrag/mcp-server'. This establishes a secure communication bridge between the agent and your local Vector database.

Use Cases

• Technical Documentation Search: Quickly query local engineering manuals or code docs when troubleshooting complex systems.
• Personal Research Assistant: Upload PDFs and research papers to have OpenClaw synthesize information across multiple sources.
• Project Management: Store project meeting transcripts, requirements, and roadmaps in a local collection for consistent status updates.

Example Prompts

1. "Check the documents in my 'Project Alpha' collection and tell me the primary deadline for the phase two deployment."
2. "Summarize the security protocols listed in the local onboarding PDF, focusing specifically on data encryption requirements."
3. "Search my local knowledge base for references to the legacy API and explain how they impact the current integration architecture."

Tips & Limitations

• Resource Management: ClawRAG is memory-intensive; ensure you have at least 4GB of RAM available, and ideally 8GB if you are running local embedding models within Docker.
• Data Privacy: Because all data is stored locally within ChromaDB, ensure you manage your own backups; if the Docker container volume is deleted, your vector index is lost.
• Optimizing Search: For the best results, use descriptive filenames for your documents, as the hybrid search index performs better when both metadata and content are clear and organized.