Configuration

Models

Bring your own models. Any OpenAI-compatible endpoint works for both embedding and chat — OpenAI, Together, Groq, OpenRouter, Ollama, vLLM, LM Studio, or your own server.

Embedding model

Used to embed source chunks and questions for retrieval. Any OpenAI-compatible embedding endpoint works — OpenAI, Ollama, Together, or your own server.

Chat model

Used to answer questions grounded in your sources. Any OpenAI-compatible chat endpoint works — OpenAI, Groq, OpenRouter, Ollama, vLLM, or your own server.

Provider preset

Base URL Where the /v1/chat/completions endpoint lives. Use http://localhost:11434/v1 for Ollama.

API key Stored locally in the SQLite database. Not required for local providers.

Model

Click Refresh to query {base}/models on the server. You can also type a custom value.

Temperature

Max tokens

System prompt

Retrieval How many source chunks to feed the model per question.

Storage

All data lives in the data/ volume on the host. SQLite, uploaded files, and settings are all in notebooks.db. Back this folder up to keep your notebooks.

Path	Size	Notes
Loading…