2.1 KiB

Raw Blame History

Orchestrated Discussions

Multi-agent AI discussion orchestration with voting and phases.

Conduct structured discussions between multiple AI personas, each with distinct perspectives, expertise, and voting behavior.

# Create a discussion
discussions new "Add user authentication" --template feature

# Run a turn with specific participants
discussions turn auth-discussion.md @architect @security @pragmatist

# Check status
discussions status auth-discussion.md
# Output: Phase: initial_feedback, Votes: READY: 1, CHANGES: 2

# Interactive mode
discussions ui auth-discussion.md

Installation

pip install orchestrated-discussions

# For TUI support
pip install orchestrated-discussions[tui]

Requirements

Python 3.10+
SmartTools (installed automatically)
At least one AI CLI tool (Claude, Codex, OpenCode, etc.)

Quick Start

# Create your first discussion
discussions new "My Feature" --template feature

# See bundled participants
discussions participants list

# Run a turn
discussions turn my-feature.md @architect @pragmatist

# Add your own comment
discussions comment my-feature.md "I think we should..." --vote READY

How It Works

Discussions are markdown files with structured comments
Participants are AI personas with distinct perspectives (architect, security, pragmatist, etc.)
Phases guide the discussion through stages (feedback → review → vote)
Votes (READY/CHANGES/REJECT) determine consensus
Markers (Q:, TODO:, DECISION:) capture structured information

Documentation

Design Document - Full architecture and implementation details
API Reference - Python API documentation
Participant Guide - Creating custom personas

Project Context

This is part of a three-project ecosystem:

SmartTools - AI provider abstraction and tool execution
Orchestrated Discussions (this project) - Multi-agent conversation orchestration
CascadingDev - Git-driven automation (uses both above)

License

MIT

2.1 KiB Raw Blame History