Knowledge
Lecca.io provides a way to easily upload files or text, chunk, embed, and save the files to S3 for retrieval. RAG.
To get a deep dive into all available environment variables refer to
server.config.ts.
Roadmap Item: Use Chroma to embed and store data locally without S3 or Pinecone
Roadmap Item: Letting users selecting the AI provider and embedding model that each notebook would use.
Roadmap Item: Letting users select the chunking strategy they want to use.
Roadmap Item: Allowings users to select between Using pg vector and Pinecone.
Roadmap Item: Support multiple dimension sizes besides just 1536
We currently use S3, OpenAI's embedding model, and Pinecone to operate knowledge notebooks. We have plans to support Chroma for local embedding and vector storage.
S3 Secrets
- Create an AWS S3 bucket
- Generate an
S3_ACCESS_KEY_ID - Generate an
S3_SECRET_ACCESS_KEY - Generate a
S3_REGIONsecret;
Create Pinecone Database
- Go to Pinecone
- Create an index with a dimension of 1536
- Generate a
PINECONE_API_KEY - Generate a
PINECONE_INDEX_NAMEsecret.
Add all these secrets to the environment variables
Once added as environment variable, the ServerConfig will use it and the knowledge feature will be enabled.