RAGMill Setup

This is a standalone configuration tool — separate from wherever RAGMill actually runs. Fill in what you need, save, done.

Vector store backend optional

Where document chunks + embeddings are stored. Default is SQLite — a local file, no setup needed.

SQLite (default) Qdrant Pinecone

Qdrant cluster URL — required for Qdrant

Qdrant API key — required for Qdrant

Collection name optional, default "ragmill"

Pinecone API key — required for Pinecone

Pinecone environment optional

Index name optional, default "ragmill"

Answer generation (chat) optional

How RAGMill turns retrieved chunks into an answer. Default is a local model — no API key, ever. Switch to Gemini or ChatGPT if you want cloud-quality answers instead.

Local model (default) Gemini ChatGPT (OpenAI)

Model repo (Hugging Face) optional

Model file optional

Context size (tokens) optional, default 4096

First use downloads the model (~1.1GB by default) to ~/.cache/ragmill/models.

Gemini API key — required for Gemini

Model optional, default "gemini-flash-latest"

OpenAI API key — required for ChatGPT

Model optional, default "gpt-4o-mini"

Server settings optional

Host/port for ragmill serve (the REST API + chatbox), not this setup page.

Host optional, default 0.0.0.0

Port optional, default 8000