Integration overview
How to put a federated model into your own product. Two paths, same platform:
| Docker sidecar | Python SDK | |
|---|---|---|
| Your app is | any language | Python |
| You run | one extra container | one extra import |
| You call | POST http://localhost:8080/predict |
client.predict([...]) |
| Best for | Node, Go, .NET, a monolith you don't want to touch | a Python service that already owns its process |
Both wrap the same core. Whatever you pick, your data never leaves the machine it is on. The platform sees model weights and anonymous counters, and nothing else. That is not a policy; it is the shape of the system — spelled out here.
The pieces
- Prepare your model — the two functions you supply, and how they are analyzed and validated before anything ships.
- The Docker sidecar or the Python SDK — wiring the client into your product.
- Configuration — the environment variables both paths share.
- Verify what you received — checking every kit and image against our signing key.
- Consent & erasure — the training switch and your GDPR tooling.
- Troubleshooting — when something doesn't behave.