Your first project

1. Create a project directory

aicr create my-study.project/

This scaffolds a complete project directory with template configuration files.

2. Configure endpoints

Copy config/endpoints.template.json to config/endpoints.json and fill in your endpoint URLs and auth structure:

{
  "endpoints": {
    "my_endpoint": {
      "url": "https://example.com/api",
      "auth": { "type": "api_key", "key": "my_key", "header": "Authorization", "scheme": "Bearer" },
      "fields": []
    }
  }
}

3. Configure secrets

Copy config/secrets.template.json to config/secrets.json and fill in your API key values:

{
  "keys": {
    "my_key": "sk-..."
  }
}

secrets.json is gitignored and should never be committed. For Azure deployments using DefaultAzureCredential, no secrets.json is needed — run az login once instead (see Azure prerequisites).

4. Edit deployments and inputs

  • config/deployments.json — add or adjust the model deployment arms to test
  • input/text.json — the text corpus and ground truth section headings
  • input/queries.json — query templates and task types
  • input/prompts.json — bindings between texts and queries

5. Run

aicr run my-study.project/

Output lands in my-study.project/output/<timestamp>/. See Output for details.

6. What to expect

While running, aicr prints a progress line per request showing the deployment, set, and rep. At the end of each run it prints an identity groups summary — responses grouped by content hash — and the total elapsed time.

If a request fails mid-run (network error, auth expiry), the error is logged and aicr continues with the remaining requests. The partial output directory is preserved as-is; re-running creates a new timestamped directory alongside it.

See Troubleshooting if you encounter auth errors or repeated failures.


This site uses Just the Docs, a documentation theme for Jekyll.