Skip to main content

​
1. Sign in and create an assistant

  1. Open the Promptify app at https://app.promptify.one.
  2. Sign up / log in from the auth screen.
  3. Navigate to Onboarding in the sidebar.
  4. Walk through the onboarding chat and describe:
    • who your assistant is for,
    • what it should and shouldn’t do,
    • tone, style, and formatting preferences.
When you’re done, Promptify creates an assistants row in Supabase with your description, system prompt, and base model.

​
2. Start a fine‑tuning job

  1. Go to the Dashboard.
  2. Click Start Training on the Fine‑Tuning Job card.
  3. Behind the scenes, Promptify will:
    • generate a JSONL training dataset via OpenAI (real-fine-tune.ts),
    • upload it to OpenAI Files,
    • start a fine‑tuning job,
    • poll status via /api/fine-tune/status,
    • update your assistant to point to the fine‑tuned model when ready.
Stay on the Dashboard to watch log output and progress.

​
3. Talk to your assistant

Once training has finished (job status SUCCEEDED):
  1. Open the Assistant page.
  2. Send a few prompts and verify:
    • responses follow your spec and tone,
    • edge cases are handled as expected.
This page calls your authenticated internal API (/api/assistant/message) using the same assistant and model that the public API will use.

​
4. Use the Chat API

Now you can call the Chat API directly from your own code:
  • Endpoint: POST /api/v1/chat
  • Auth: Authorization: Bearer <access_token>
  • Body:
{
  "message": "How do I reset my password?",
  "conversationId": "optional-conversation-id"
}
See the dedicated Chat API reference for full request/response details and example snippets.