Skip to content

Latest commit

 

History

History
123 lines (78 loc) · 3.7 KB

README.md

File metadata and controls

123 lines (78 loc) · 3.7 KB

Leadership Calibration Crew

Welcome to the Leadership Calibration project, powered by crewAI.

This crew models a structured debate between two senior leadership personas:

  • Senior Architect – focuses on technical design, scalability, and system quality.
  • Engineering Manager – focuses on people, delivery, and business impact.

Given a topic, the crew runs a multi-step debate and produces:

  • Architect position → architect.md
  • Engineering manager position → manager.md
  • Final synthesized conclusion → conclusion.md

You can run the crew through an interactive Gradio UI.

Installation

  • Python: requires >=3.10,<3.14
  • Dependencies: managed with UV and pyproject.toml

Install uv (once, globally):

pip install uv

From the project root (3_crew/leadership_calibration), install dependencies:

uv sync

Or, if you prefer the crewAI CLI workflow:

crewai install

Make sure your OPENAI_API_KEY (or other LLM provider keys) is available in the environment or defined in a local .env file.

Project Structure

Key components of the crew live under src/leadership_calibration:

  • Agents configuration: config/agents.yaml

    • engineering_manager_agent – senior engineering manager & technical leader
    • senior_architect_agent – senior software engineer & architect

  • Tasks configuration: config/tasks.yaml

    • architect_position_statement → writes output/architect.md
    • engineering_manager_position_statement → writes output/manager.md
    • final_alignment_and_resolution → writes output/conclusion.md

  • Crew orchestration: crew.py

    • Defines the LeadershipCalibration crew and wires agents + tasks.

  • CLI entrypoints: main.py

    • Provides run, train, replay, test, and run_with_trigger functions.

  • Gradio app: app.py

    • Defines a small UI where you can type a topic and watch the debate stream.

Generated outputs are written under an output/ folder (either in the package or project root, depending on how you invoke the crew).

Running the Gradio UI

The interactive web UI is defined in src/leadership_calibration/app.py using Gradio. To launch it:

uv run app.py

Then:

  1. Open the local URL printed by Gradio in your browser.
  2. Enter a debate topic (e.g. “Should we migrate to microservices?”).
  3. Click Start Debate to begin streaming the discussion.
  4. Use Cancel Debate to stop an in-progress run.

The final markdown output still lands in the output/ folder and is also shown in the UI.

Customizing the Leadership Calibration

  • Agents – edit src/leadership_calibration/config/agents.yaml

    • Update roles, goals, and instructional style (instructions block).

  • Tasks & debate flow – edit src/leadership_calibration/config/tasks.yaml

    • Change task descriptions, expected outputs, or ordering.

  • Crew wiring & tools – edit src/leadership_calibration/crew.py

    • Add or modify tools, change which agent runs which task, or tweak parameters.

  • CLI behavior – edit src/leadership_calibration/main.py

    • Adjust default topics, input schema, or logging.

  • UI behavior – edit src/leadership_calibration/app.py and debate_runner.py

    • Change layout, streaming behavior, or how the debate is rendered.

Support

For more about crewAI:

  • Docs: https://docs.crewai.com
  • GitHub: https://github.com/joaomdmoura/crewai
  • Discord: https://discord.com/invite/X4JWnZnxPb

This project is a specialized example on top of crewAI for calibrating leadership decisions via structured debate.