# Dopp

> Dopp is a drop-in proxy for any Jev-compatible decision API: TypeSafe's Jev, Kev, GLiNER, simple-jev, or your own endpoint that answers typed questions about a state. You change one URL; every request still gets your current API's answer and is recorded as training data. When you have enough, one button trains a small open model on your own requests, measures it against the API it learned from on requests it never saw, and lets you choose who answers: the API, your model with the API as backup, or your model alone, on our servers, or offline, even in your users' browsers.

## For agents

Set up with one call, no sign-up form: POST https://dopp.sh/agent/start returns a working key and a claim link for the person. Instructions: https://dopp.sh/skill.md · API: https://dopp.sh/docs.md

## Compare

- [Dopp vs Hugging Face AutoTrain](https://dopp.sh/blog/vs-huggingface-autotrain/): AutoTrain fine-tunes a Hub model on a CSV you upload. Dopp makes the CSV from your live traffic, labels it, trains a typed-decision model, measures it against that API, and routes.
- [Dopp vs Jev](https://dopp.sh/blog/vs-jev/): Jev answers typed questions about any state with a hosted model. Dopp keeps Jev in the loop and turns your Jev traffic into a small model you own that runs hosted or offline. Jev stays as the fallback.
- [Dopp vs OpenAI fine-tuning](https://dopp.sh/blog/vs-openai-fine-tuning/): OpenAI lets you fine-tune their models on your JSONL. Dopp collects the JSONL for you from live traffic, labels it, trains an open model you own, and measures it.
- [Dopp vs simple-jev and RFDT](https://dopp.sh/blog/vs-simple-jev-rfdt/): simple-jev is an open-source server that answers typed questions with open models; RFDT is its training script. Dopp is the product around that idea: data capture, labels, one-button training, measuring, routing.

## Articles

- [Stop paying for decisions your own model already gets right](https://dopp.sh/blog/stop-paying-for-decisions/): Change one URL, keep the decision API you use today, and by the afternoon own a small model trained on your own requests, measured against it, switchable in one click.
- [Every request is training data. Why a proxy is the right shape.](https://dopp.sh/blog/every-request-is-training-data/): A fine-tuned model needs examples with answers. You already produce hundreds a day. A proxy keeps every one, in your format, with your current API's answer as the label.

## Reference

- [API docs](https://dopp.sh/docs.md): the proxy endpoint, keys, and the request format.