Memory Chromadb

What This Skill Does

The Memory Chromadb skill provides OpenClaw with a persistent, long-term storage solution using ChromaDB as the underlying vector database. By integrating with Ollama for text embeddings, it allows the agent to transform natural language into high-dimensional vector representations. This enables the model to perform semantic searches across previous conversations and stored data. When 'auto-recall' is enabled, the skill automatically queries the vector database for relevant past context before the model generates a response, ensuring the agent maintains continuity and historical awareness throughout multi-turn interactions.

Installation

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

clawhub install openclaw/skills/skills/broedkrummen/memory-chromadb

Ensure that you have a running instance of ChromaDB at the configured URL and an Ollama server accessible with your chosen embedding model (defaulting to nomic-embed-text) before activating the skill.

Use Cases

This skill is ideal for users requiring a personalized assistant that remembers past preferences, project requirements, or specific technical constraints across long sessions. It is particularly effective for managing documentation knowledge bases, tracking user-specific preferences, or maintaining logs of project milestones that would otherwise exceed the standard token window of an LLM.

Example Prompts

1. "Look back at our previous discussion regarding the architectural pattern for the dashboard and suggest a database schema that aligns with it."
2. "What was the primary feedback I gave you regarding the tone of voice in our last three marketing campaign drafts?"
3. "Save this project requirement document to my long-term memory so you can reference the milestones later."

Tips & Limitations

To optimize performance, adjust the minScore and autoRecallResults settings within the configuration file to balance the depth of context versus the latency of the response. Be aware that this skill performs network requests to both the ChromaDB and Ollama services, so ensure low-latency connectivity to these endpoints. As this is a vector database, it is best suited for semantic retrieval; it may struggle with exact keyword matching if the embedding model is not fine-tuned for the specific domain language. Always monitor memory usage in the database as the collection grows to ensure retrieval speeds remain efficient.