rag-pipeline

RAG Pipeline Example

Full Retrieval-Augmented Generation pipeline using the AI Gateway and MongoDB Atlas Vector Search.

Architecture

                    
                      Markdown   
                       Docs      
                    
                            ingest.ts
                         
                       Chunk &   >  AI Gateway   
                       Embed     <  (embeddings) 
                         
                           
                    
                      MongoDB    
                      Atlas      
                      (vectors)  
                    
                            search.ts / chat.ts
                         
                      Query ->    >  AI Gateway   
                      Retrieve -> <  (LLM chat)  
                      Generate        
                    

Scripts

Script	Description
pnpm ingest	Load sample docs, chunk, embed, store in MongoDB
pnpm search	Vector similarity search (pnpm search "your query")
pnpm chat	Full RAG chat (pnpm chat "your question")

Prerequisites

1. MongoDB Atlas cluster with vector search enabled
2. AI Gateway running with an embedding model and a chat model
3. Node.js 18+

Setup

1. MongoDB Atlas Vector Search Index

Create a vector search index on your collection:

{
  "fields": [
    {
      "type": "vector",
      "path": "embedding",
      "numDimensions": 1024,
      "similarity": "cosine"
    }
  ]
}

Name the index vector_index. Adjust numDimensions to match your embedding model.

2. Environment

cp .env.example .env
# Edit .env with your MongoDB URI, gateway URL, and model names

3. Run

pnpm install

# Step 1: Ingest sample documents
pnpm ingest

# Step 2: Search
pnpm search "What pricing plans are available?"

# Step 3: Full RAG chat
pnpm chat "How do I handle rate limit errors?"

Sample Data

The sample-data/ directory contains fake product documentation for "AcmeCloud":

• product-overview.md - Features and pricing
• getting-started.md - Installation and quick start
• troubleshooting.md - Common issues and solutions

Replace these with your own documents for real use cases.

Configuration

Variable	Description	Default
GATEWAY_URL	AI Gateway URL	http://localhost:3000
GATEWAY_API_KEY	Gateway API key	-
MONGODB_URI	MongoDB connection string	mongodb://localhost:27017
MONGODB_DATABASE	Database name	rag_demo
MONGODB_COLLECTION	Collection name	documents
EMBEDDING_MODEL	Model for embeddings	bedrock/titan-embed-v2
CHAT_MODEL	Model for chat generation	bedrock/claude-3-sonnet