SimpleRAG - Setup

Configure Enhanced SimpleRAG

Enter your API keys and settings to configure the system with Normal and Graph RAG support.

🔒

Your keys are safe. Keys are encrypted at rest, used only to make API calls on your behalf, and are never transmitted to us, never logged, and never shared with anyone.

API Keys

Gemini API Key *

Required for embeddings and Graph RAG entity extraction. Get a key from Google AI Studio.

Qdrant API Key

Required for vector database. Get a key from Qdrant Cloud.

Qdrant URL

URL of your Qdrant instance (e.g., https://your-instance.qdrant.io)

Voyage AI API Key

Optional. Enables cross-encoder reranking (+8–15% retrieval accuracy). Get a key from voyageai.com. Leave blank to disable reranking.

Neo4j Database Configuration

Enable Neo4j Integration

Enable to use Neo4j graph database for storing knowledge graphs

Neo4j URI

Connection URI for your Neo4j instance. Get from Neo4j Aura or your self-hosted instance.

Neo4j Username

Username for Neo4j authentication (default is usually 'neo4j')

Neo4j Password

Password for Neo4j authentication

ℹ️ Neo4j Setup Instructions:

Create a free Neo4j Aura instance at neo4j.com/cloud/aura
Copy the connection URI, username, and password from your instance details
Enable Neo4j integration and enter the credentials above
You can then use the "Neo4j Only" upload mode to store knowledge graphs

RAG Mode Settings

Default RAG Mode

Normal RAG - Traditional semantic search with document chunks

Fast, efficient, good for straightforward Q&A

Graph RAG - Knowledge graph with entities and relationships

Advanced reasoning, better for complex queries and relationships

Neo4j Mode - Store directly in Neo4j graph database

Pure graph storage without vector embeddings, requires Neo4j configuration

LLM Settings

Preferred LLM

Gemini (gemini-2.5-flash)

Raw Results (No LLM)

Advanced Settings

Chunk Size

Size of text chunks in characters (500-5000 recommended).

Chunk Overlap

Overlap between adjacent chunks (50-500 recommended).

Results Count (Top K)

Number of results to retrieve for each query (1-20).

Graph RAG Settings

Max Entities per Chunk

Maximum entities to extract from each text chunk (5-50).

Graph Reasoning Depth

How many relationship hops to consider (1-5).

Entity Similarity Threshold

Threshold for merging similar entities (0.5-1.0).