research-deep

You can automate deep research by defining an outline.yaml file and running the tool to spawn independent background agents for batch processing. It auto-locates the configuration and launches item-level investigations to perform parallel web-search tasks.

To validate structured JSON outputs, you need to define a fields.yaml specification. The research process automatically checks the gathered web-search data against this file to ensure your batch processing results match the required schema.

The best way to resume deep research from partial results is to run the tool in the same directory. It automatically detects previously completed items in your outline and skips them, continuing the batch execution from where it left off.

Yes, you need an outline.yaml file in your current working directory to start batch processing. The tool extracts the research items and execution configuration directly from this file to spawn the autonomous agents.

Yes, you can configure per-item settings for autonomous research agents. The tool reads the execution configuration from your outline.yaml, allowing customized parallel deep-dives on multiple topics with specific parameters for each item.

Batch processing fails to validate structured output when your JSON results do not match the fields.yaml specification. Ensure your autonomous agents gather data strictly conforming to the defined schema fields to pass validation.