dopp vs Hugging Face AutoTrain
AutoTrain fine-tunes a Hub model on a CSV you upload. dopp makes the CSV from your live traffic, labels it, trains a model, measures it, and routes.
Articles
dopp sits between your app and Jev, or any API that answers the same way. You change one URL, your app gets the same answers, and every request is kept. You train a small model of your own on those requests, with labels from you, an LLM, a dataset or another API, and run it hosted, on your own servers, or offline in the browser, with your API as the fallback.
Compare
dopp vs Jev
Jev answers typed questions with a hosted model. dopp keeps Jev as the fallback and turns your Jev traffic into a small model you own, hosted or offline.
dopp vs OpenAI fine-tuning
OpenAI fine-tunes their models on your JSONL. dopp collects the JSONL from your live traffic, labels it, trains an open model you own, and measures it.
dopp vs simple-jev and RFDT
simple-jev answers typed questions with open models; RFDT is its training script. dopp is the product around that idea: capture, labels, training, routing.
Read
Stop paying for decisions your own model already gets right
Change one URL, keep the decision API you use today, and by the afternoon own a small model trained on your own requests.
Every request is training data. Why a proxy is the right shape.
A fine-tuned model needs examples. You already produce hundreds a day. A proxy keeps every one, in your format, and you pick what labels them.
Fine-tuning Kev 9B past Kev 27B and Jev
Kev 9B went from 57.5% to 75% on Kev's own emotion eval after training on 2,500 labelled rows. How, start to finish, for about $1 of GPU.
How a 34 MB model runs this guard
I prototyped a castle guard on Jev, trained a 34 MB model on a few hundred labelled lines, and now the guard runs in your browser tab with no server.
Can you run Jev offline?
Train a 34 MB model on the requests your app already sends Jev and run it with no internet. I did it with Home Assistant for $0.12.