fraggle 0.1.0

A simple RAG API - build a question and answer interface to your own content in minutes

Fraggle

A simple RAG (Retrieval-Augmented Generation) API - build a question and answer interface to your own content in minutes.

Fraggle is the spiritual successor to microllama, modernized with:

• Provider-agnostic LLM support via any-llm
• Modern LangChain (0.3+)
• Support for OpenAI and Anthropic models
• Simple deployment and usage

Installation

pip install fraggle

Quick Start

1. Prepare your content in a source.json file:

[
  {
    "content": "Fraggle is a RAG API for building Q&A interfaces.",
    "title": "What is Fraggle",
    "url": "https://example.com/docs"
  },
  {
    "content": "You can use OpenAI or Anthropic models with Fraggle.",
    "title": "Supported Models"
  }
]

2. Set your API key:

export OPENAI_API_KEY="your-key-here"
# or
export ANTHROPIC_API_KEY="your-key-here"

3. Create the search index:

fraggle index

4. Start the server:

fraggle serve

5. Query your content:

curl -X POST http://localhost:8000/api/ask \
  -H "Content-Type: application/json" \
  -d '{"question": "What is Fraggle?"}'

Configuration

Configure Fraggle with environment variables:

Variable	Default	Description
SOURCE_JSON_PATH	source.json	Path to your content JSON file
INDEX_PATH	faiss_index	Path to store/load the FAISS index
LLM_PROVIDER	openai	LLM provider: openai or anthropic
LLM_MODEL	gpt-4o-mini	Model ID (e.g., gpt-4o, claude-3-5-sonnet-20241022)
EMBEDDINGS_PROVIDER	openai	Embeddings provider
EMBEDDINGS_MODEL	text-embedding-3-small	Embeddings model
CHUNK_SIZE	1000	Text chunk size for indexing
CHUNK_OVERLAP	100	Overlap between chunks
K_CONTEXT_DOCS	4	Number of documents to retrieve
UVICORN_HOST	0.0.0.0	Host to bind the server to
UVICORN_PORT	8000	Port to bind the server to

Using Anthropic Claude

export ANTHROPIC_API_KEY="your-key-here"
export LLM_PROVIDER="anthropic"
export LLM_MODEL="claude-3-5-sonnet-20241022"
fraggle serve

CLI Commands

fraggle serve

Start the API server. Automatically serves the frontend at / if the frontend/ directory exists.

fraggle index

Create a FAISS index from your source documents.

Options:

• --source: Path to source JSON file (default: source.json)
• --output: Path to save the index (default: faiss_index)

fraggle make-front-end

Generate a simple HTML frontend for your Q&A interface.

Options:

• --output: Directory for frontend files (default: frontend)

fraggle make-dockerfile

Generate a Dockerfile for containerized deployment.

API Endpoints

POST /api/ask

Non-streaming question answering.

Request:

{
  "question": "What is Fraggle?"
}

Response:

{
  "answer": "Fraggle is a RAG API for building Q&A interfaces to your content."
}

POST /api/stream

Streaming question answering (Server-Sent Events).

Request:

{
  "question": "What is Fraggle?"
}

Response: SSE stream of text chunks.

Deployment

Docker

fraggle make-dockerfile
docker build -t fraggle .
docker run -p 8000:8000 -e OPENAI_API_KEY=your-key fraggle

Pre-building the Index

For faster startup, create the index at build time by uncommenting the RUN fraggle index line in the Dockerfile.

Development

# Clone the repo
git clone https://github.com/tomdyson/fraggle.git
cd fraggle

# Install with uv
uv pip install -e .

# Run in development mode
fraggle serve --reload

Comparison with microllama

Fraggle improves on microllama by:

•  Provider-agnostic: Use any LLM via any-llm
•  Modern dependencies: LangChain 0.3+
•  Better naming: Focus on RAG, not LLMs
•  Same simple API and deployment story

License

MIT