Skip to main content

Use Cases & Tutorials

This section contains practical, real-world workflows for CIRI. Pick a tutorial to get hands-on with common tasks:

Real-world Scenarios​

Use case 1: Research assistant with browser-based crawling and summarization​

  • Goal: Given a topic, CIRI will browse the web (using Playwright + browser CDP), extract relevant pages, and summarize them into a brief.

Flow​

Implementation pointers​

  • The web researcher agent is built in src/copilot.py via build_web_researcher_agent() and uses Playwright and the crawler browser config.
  • Ensure Playwright browsers are installed (playwright install) and CDP is available if using the user's browser profile.

Use case 2: Local codebase assistant that modifies code with human-in-the-loop​

  • Goal: Make a small code change (refactor a function), run tests, and produce a PR.

Flow​

Notes​

  • Human-in-the-loop middleware ensures tool actions (edit_file/write_file) require approval before execution.
  • PatchToolCallsMiddleware can create patches for the user to inspect.

Use case 3: Automating release packaging and publishing to PyPI​

Flow​

Implementation pointers​

  • Packaging is handled by hatchling (pyproject.toml). For release automation consider adding a GitHub Actions workflow that runs build and twine in CI with secure secrets.