MongoDB-RAG

A semantic search pipeline that combines MongoDB Atlas Vector Search with VoyageAI embeddings and reranking for high-precision RAG retrieval. User input is embedded using voyage-3.5, searched against a MongoDB Atlas vector index, and reranked with rerank-2.5 to surface the most relevant documents. Results are formatted via a template node for downstream LLM use. Suitable for knowledge bases, document Q&A, and AI agent memory retrieval.

pash

155 used·Last update 2 week ago

Open in Dify

SHARE WITH

100%

React Flow

Setup steps

Prerequisites

A MongoDB Atlas cluster (Free tier possible and 0.0.0.0/0 IP access listed) with a collection containing documents pre-embedded using VoyageAI (voyage-3.5 or compatible model) stored in an field
An Atlas Vector Search index named created on the field of that collection

Setup

Install the MongoDB Atlas plugin from the Marketplace:
Install the VoyageAI plugin from the Marketplace:
Authorize both plugins with your API keys under Tools → Authorize
In the Embed Text node, set to
In the Vector Search node, set your , — and are pre-configured as / to match the prerequisite index
In the Rerank node, confirm model is set to and configure to your desired top-K
Update the Template node to format results for your LLM or output node

TYPE

Workflow