Lumen RAG Framework

Lumen is a high-performance, modular, and database-agnostic RAG (Retrieval-Augmented Generation) framework written in Rust.

It abstracts the complexity of vector storage and retrieval, allowing you to switch seamlessy between MongoDB, CosmosDB, and Qdrant, while providing built-in support for state-of-the-art embeddings (BERT) via candle.

🚀 Features

🔌 Modular Backends: Switch between MongoDB (Atlas/Local) and Qdrant with Feature Flags.
⚡ High Performance: Built on Tokio, Actix-web, and Rayon for async and parallel processing.
🧠 Local Embeddings: Integrated BERT support using Hugging Face's candle (no external API needed for embeddings).
📄 Smart Chunking: Intelligent text segmentation preserving semantic context.
🤖 LLM Agnostic: Compatible with any OpenAI-compatible API (Ollama, vLLM, OpenAI, Mistral, etc.).

📦 Installation

Add lumen-rag to your Cargo.toml. Select the database backend you need:

[dependencies]

# For MongoDB or CosmosDB support

lumen-rag = { version = "0.1.0", features = ["mongodb"] }



# OR for Qdrant support

lumen-rag = { version = "0.1.0", features = ["qdrant"] }

🛠️ Configuration

Lumen uses environment variables for configuration. Create a .env file in your project root:

# --- LLM Settings ---
# URL to your LLM provider (Ollama, OpenAI, Groq, etc.)
LLM_URI=[https://api.openai.com/v1/chat/completions](https://api.openai.com/v1/chat/completions)
# Model name
MODEL=gpt-3.5-turbo
# (Optional) API Key if using a cloud provider
LLM_API_KEY=sk-your-api-key-here

# --- Database Settings ---
# MongoDB / CosmosDB
COSMOS_URI=mongodb://admin:password@localhost:27017
DATABASE=lumen_db
COLLECTION=knowledge_base

# Qdrant
QDRANT_URI=http://localhost:6334

⚡ Quick Start

1. Setup Infrastructure

You can easily start the required databases using Docker:

# Start MongoDB and Qdrant

docker run -d -p 27017:27017 -e MONGO_INITDB_ROOT_USERNAME=admin -e MONGO_INITDB_ROOT_PASSWORD=password mongo:latest

docker run -d -p 6333:6333 -p 6334:6334 qdrant/qdrant:latest

2. Run the Examples

We provide full server examples in the examples/ folder.

Using MongoDB:

cargo run --example server_mongo --features mongodb

Using Qdrant:

cargo run --example server_qdrant --features qdrant

3. Test the API

Ingest a Document (Indexing)

curl -X POST [http://127.0.0.1:8080/ingest](http://127.0.0.1:8080/ingest) \

   -H "Content-Type: application/json" \

   -d '{
     "text": "Rust is a systems programming language focused on safety and performance.",
     "metadata": {"source": "wikipedia"}
   }'

Ask a Question (RAG)

curl -X POST [http://127.0.0.1:8080/ask](http://127.0.0.1:8080/ask) \

   -H "Content-Type: application/json" \

   -d '{
     "question": "What is Rust focused on?"
   }'

🏗️ Architecture

Lumen is built around the VectorStore trait, enabling easy integration of new vector databases.

#[async_trait]
pub trait VectorStore: Send + Sync {
    async fn add_passages(&self, passages: Vec<Passage>) -> Result<Vec<String>>;
    async fn search(&self, query_embedding: &[f32], limit: usize) -> Result<Vec<Passage>>;
}

Supported Stores

Database	Feature Flag	Search Type
MongoDB	`mongodb`	Hybrid (Fetch + In-memory Cosine Similarity)
CosmosDB	`mongodb`	Hybrid (Mongo API Compatible)
Qdrant	`qdrant`	Native HNSW Vector Search

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Fork the project
Create your feature branch (git checkout -b feature/AmazingFeature)
Commit your changes (git commit -m 'Add some AmazingFeature')
Push to the branch (git push origin feature/AmazingFeature)
Open a Pull Request

📄 License

Distributed under the MIT License. See LICENSE for more information.

lumen-rag 0.1.0