Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.veri.studio/llms.txt

Use this file to discover all available pages before exploring further.

This example walks through the minimum dataset shape for a first training run. Use it when you already have prompts and target responses, or when you are converting examples from another system.

​
JSONL Format

Each line should be a standalone JSON object. 
{"prompt":"Summarize this support ticket in one sentence: ...","response":"Customer cannot access billing settings after changing workspaces."}
{"prompt":"Rewrite this answer for a technical buyer: ...","response":"The API supports signed requests, per-user keys, and request-level tracing."}

​
Upload

curl -X POST https://api.veri.studio/v1/datasets \
  -H "Authorization: Bearer $VERI_API_KEY" \
  -F "file=@training-data.jsonl" \
  -F "name=support-response-examples"

​
Start a Training Job

curl -X POST https://api.veri.studio/v1/training_jobs \
  -H "Authorization: Bearer $VERI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "base_model": "Qwen/Qwen3-4B",
    "dataset_id": "ds_support_examples",
    "reward_function_id": "rf_response_quality",
    "output_name": "support-response-model",
    "gpu": { "gpu_type": "A100-80GB", "gpu_count": 1 }
  }'

​
Quality Checklist

  • Keep prompts close to production traffic.
  • Include difficult edge cases, not just ideal examples.
  • Remove secrets, credentials, and personal data before upload.
  • Start with hundreds of strong examples before scaling to larger datasets.