@sprucelabs/chroma-data-store

Chroma Data Store

Give your skill the ability to store and retrieve data from a Chroma database.

Running the Chroma Database

• Clone this rep
• Run yarn start.chroma.docker

Setting an embedding model

By default, the ChromaDabatase class will use llama3.2 hosted through Ollama to generate embeddings

Installing Ollama

• Visit https://ollama.com
• Click "Download"
• Select your OS

Installing Llama3.2

Llama 3.2 is the newest version of Llama (as of this writing) that supports embeddings.

• Inside of terminal, run ollama run llama3.2
• You should be able to visit http://localhost:11434/api/embeddings and get a 405 response (this is because the route only accepts POST requests)

Improving embeddings with nomic-embed-text

We have seen significantly better search performance when using nomic-embed-text to generate embeddings.

Run ollama pull nomic-embed-text

Using in your skill

Add the following to your env:

DB_CONNECTION_STRING="chromadb://localhost:8000"
DB_ADAPTER="@sprucelabs/chroma-data-store"