engram

What This Skill Does

The Engram skill acts as a cognitive bridge between your AI agent and your personal knowledge base. By leveraging Pinecone's vector database and Google's Gemini embeddings, Engram transforms a collection of static Markdown files into a dynamic, searchable, and semantically aware knowledge store. Unlike traditional grep-based searches that rely on exact keyword matches, Engram interprets the underlying intent and meaning of your queries. Whether you are using a digital garden, a research vault, or a collection of technical documentation, this skill allows your OpenClaw agent to retrieve contextually relevant information instantly, ensuring your local intelligence is always at your fingertips.

Installation

To integrate Engram into your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/anwitch/engram

Once installed, ensure your environment variables for Pinecone API keys and Gemini service credentials are configured to allow the agent to read your documents and communicate with the embedding service.

Use Cases

1. Personal Knowledge Management (PKM): Query your Obsidian vault to recall specific project details or philosophical notes you recorded months ago without remembering exact phrasing.
2. Documentation Retrieval: Quickly locate specific troubleshooting steps or configuration parameters buried within hundreds of local markdown files in a development repository.
3. Synthesis & Research: Ask the agent to find connections between different documents in your vault, effectively acting as an automated research assistant that links fragmented ideas.

Example Prompts

1. "Engram, search my notes for any information regarding the project architecture decisions we made last quarter."
2. "Use engram_search to find what I wrote about the interaction between meditation and productivity, and summarize the key takeaways."
3. "Please run engram_index to update the knowledge base with my newly added research notes."

Tips & Limitations

To get the best results, ensure your Markdown files are well-structured with clear headings. The engram_index tool is crucial; establish a routine to trigger it after major documentation updates to avoid stale results. Be aware that this skill requires external network access to process embeddings via Gemini and communicate with Pinecone, meaning search functionality will be unavailable when offline. Additionally, large vaults may require careful management of token limits when retrieving multiple high-scoring snippets.