Skip to main content

Mongodb Setup

tip

Mongodb is used to save vector embeddings for RAG feature. This is required when mongodb is used as RAG vector store i.e. RAG_VECTOR_STORE_TYPE=mongodb

Env variables​

apps/api/.env
RAG_VECTOR_STORE_MONGODB_URI=...
RAG_VECTOR_STORE_MONGODB_SEARCH_INDEX=...

Visit mongodb atlas platform to create database and search index

1. Create a database​

create-database

2. Create search index​

create-search-index

Use the json editor to setup the search index as follows. Please take note of the search index name to use in .env

{
  "fields": [
    {
      "numDimensions": 512,
      "path": "embeddings",
      "similarity": "cosine",
      "type": "vector"
    },
    {
      "path": "documentId",
      "type": "filter"
    }
  ]
}