Turn prompts into training-ready datasets

Upload a list of prompts, point at any OpenAI-compatible model, and get back a clean JSONL file ready for fine-tuning. No glue code. No babysitting.

No credit cardEncrypted keys at restJSONL out of the box
datagen · jobs/run
prompts.txt
1  Explain quicksort in 2 sentences.
2  Write a haiku about debugging.
3  Summarize REST in one paragraph.
4  Compare TCP and UDP briefly.
…2,996 more
dataset.jsonlrunning · 2,847 / 3,000
{"messages": ["user", "assistant"]}
{"messages": ["user", "assistant"]}
{"messages": ["user", "assistant"]}
94%~12s remaining

Four steps from prompt to dataset

Bring your own model and key. We handle orchestration, retries, and packaging.

01

Save a key

Store your provider credentials. Works with any OpenAI-compatible endpoint.

02

Upload prompts

Drop in a .txt file — one prompt per line. Up to 5MB per file.

03

Run a job

Pick a key, pick a prompt file, choose a model. We handle the rest.

04

Download JSONL

Each line is a chat conversation pair, ready for fine-tuning.

Fast by default

Concurrent execution with smart retries. Watch progress live as prompts complete.

Your keys, encrypted

Provider credentials are encrypted at rest and only decrypted in-memory at job start.

Provider agnostic

OpenAI, OpenRouter, Fireworks, vLLM, llama.cpp — anything that speaks the OpenAI API.

Ready to build your next dataset?

Spin up a job in under a minute. Bring your own key — we never charge for tokens.