raglite

What This Skill Does

RAGLite is a specialized local-first Retrieval-Augmented Generation (RAG) cache designed to optimize document indexing and retrieval for AI agents. Rather than acting as a replacement for chat context, it provides a persistent, highly structured storage layer for internal knowledge bases. The skill operates in three distinct phases: condensation, indexing, and querying. First, it distills raw documents into clean, low-fluff Markdown files that emphasize high-value information. Next, it indexes these artifacts into a Chroma database. Finally, it provides a hybrid retrieval mechanism that combines vector similarity searches with fast, keyword-based lookups powered by ripgrep. By focusing on compression before embedding, RAGLite reduces operational costs and improves the reliability of information retrieval.

Installation

Installation is handled through a skill-local virtual environment to ensure dependency isolation. Before beginning, ensure you have Python 3.11+ installed and the ripgrep utility available (e.g., 'brew install ripgrep'). To install, navigate to your OpenClaw directory and execute './scripts/install.sh'. This command triggers the installation from the official repository at 'git+https://github.com/VirajSanghvi1/raglite.git@main'. Ensure your Chroma server is active, typically at 'http://127.0.0.1:8100', and if you are using the default OpenClaw condensation engine, ensure your gateway URL is reachable and the 'OPENCLAW_GATEWAY_TOKEN' environment variable is correctly configured.

Use Cases

RAGLite is ideal for scenarios requiring private, secure, and offline access to structured knowledge. Common use cases include managing internal company runbooks, personal medical archives, academic research notes, or technical documentation that you prefer to keep off public cloud platforms. Because it version-controls the distilled Markdown, it is perfect for users who need to audit or manually verify the information their AI agent is retrieving.

Example Prompts

1. "Run the RAGLite pipeline on my local folder at /home/user/notes/docs to index all new project files."
2. "Search the 'my-docs' collection using RAGLite for any mentions of the 'deployment-rollback' procedure."
3. "Use RAGLite to analyze the contents of the 'finance-2024' collection and explain the main budgeting constraints identified in the notes."

Tips & Limitations

To get the most out of RAGLite, always prioritize the '--skip-existing' and '--skip-indexed' flags during the pipeline run to save time and compute power. If you find your keyword results are empty, verify that 'rg' is installed correctly. Note that this skill is a local cache; if your data volume expands significantly, you may eventually need to scale your Chroma infrastructure. Always ensure your distilled documentation is human-readable, as the transparency of these 'artifact' files is one of the primary advantages of this approach.