Introduction
Chanterelle is a lightweight desktop application for data scientists and ML engineers to test, present, and share models and findings — without deploying anything to the cloud.
Project files are plain JSON and Python, so they live alongside your code in version control and open instantly in Chanterelle.
Project Types
| Type | What you write | What Chanterelle gives you | |
|---|---|---|---|
| 🧠 | Model | model_meta.json + handler_io.py | A generated UI form wired to your Python model |
| 📊 | Analytics | analytics.json | A rich static dashboard with charts, tables, and images |
| 💬 | Interactive | interactive.json + handler_io.py | A multi-turn conversational agent with dynamic forms |
Which should I use?
- Model — You have a trained ML model and want to test it interactively, or hand it off to a stakeholder without building a full UI.
- Analytics — You have plots, tables, and KPI numbers you want to package into a shareable report. No live Python required.
- Interactive — You want a back-and-forth conversation interface: a SQL assistant, a reporting workflow, an LLM wrapper, or similar.
How It Works
Model Projects
Three components, all in your project folder:
1. model_meta.json — defines inputs, outputs, presets, and UI settings.
➡️ Model Metadata reference
2. handler_io.py — Python functions that load, transform, and run your model.
➡️ Handler functions reference
3. model_findings.json (optional) — a dedicated insights page with charts, confusion matrices, KPIs, and more.
➡️ Model Findings reference
Analytics Projects
Write an analytics.json with sections and visualizations. Chanterelle renders it as a navigable report with a table of contents sidebar. No Python runtime required.
➡️ Analytics reference
Interactive Projects
Write an interactive.json and a handler_io.py with initialize() and on_input() functions. The Python process stays alive across turns so you can maintain state and stream responses.
➡️ Interactive reference
Chanterelle includes a built-in skill that teaches GitHub Copilot, Cursor, Windsurf, and others how to scaffold valid project files from natural language descriptions.
➡️ Learn more