RAG

RAG feature uses the following variables

RAG_EMBEDDING_MODEL

Model used to embed text to vectors are provided by RAG_EMBEDDING_MODEL env variable.

Option 1 (OpenAI):

RAG_EMBEDDING_MODEL=text-embedding-3-small
# or #
RAG_EMBEDDING_MODEL=text-embedding-3-large
# or #
RAG_EMBEDDING_MODEL=text-embedding-ada-002

Please note

If you want to use OPEN_API, please ensure you have the OPENAI env variables configured. You can find the details in section : OpenAI Setup

Option 2 (AWS titan):

RAG_EMBEDDING_MODEL=amazon.titan-embed-text-v2:0

Please note

If you want to use AWS titan model, please ensure you have the AWS env variables configured. You can find the details in section : AWS Setup

---

RAG_VECTOR_STORE_TYPE

Database used to store the vectors are provided by RAG_VECTOR_STORE_TYPE env variable.

Option 1 (duckdb):

Uses local file system to store the vectors. No external or docker container dependencies required. Good for quick testing.

RAG_VECTOR_STORE_TYPE=duckdb

Option 2 (pinecone):

Uses Pinecone cloud database to store the vectors.

RAG_VECTOR_STORE_TYPE=pinecone

Please note

If you want to use pinecone as the store type, please ensure you have the Pinecone env variables configured. You can find the details in section : Pinecone Setup

Option 3 (mongodb):

Uses Mongodb cloud database to store the vectors.

RAG_VECTOR_STORE_TYPE=mongodb

Please note

If you want to use pinecone as the store type, please ensure you have the Mongodb env variables configured. You can find the details in section : Mongodb Setup

Please note

Local mongodb does not support atlas search which is required for embeddings based searches for RAG.