This project provides a dynamic workflow tool for the Pi AI assistant, enabling complex, multi-agent task execution. It targets users who need to automate intricate processes like codebase audits, large-scale refactors, or multi-perspective research, offering a significant benefit by fanning out work across isolated subagents rather than relying on a single sequential assistant.

How It Works

The core approach involves writing plain JavaScript scripts that define workflows. These scripts leverage several global functions within a Node.js vm sandbox: agent() to spawn isolated subagents, parallel() to run tasks concurrently, and pipeline() for sequential processing across items. Workflows are structured into phases for clear progress tracking, and subagents can return validated JSON output via opts.schema. This design allows for reproducible execution by intentionally restricting access to volatile APIs like Date.now() and Math.random(), ensuring deterministic outcomes.

Quick Start & Requirements

• Install: pi install npm:pi-dynamic-workflows or from a local checkout.
• Activation: Run /reload within Pi.
• Prerequisites: A functional Pi environment.
• Links: None provided beyond the README.

Highlighted Details

• Subagent Orchestration: Dynamically coordinates multiple isolated Pi sessions via JavaScript scripts.
• Structured Output: Subagents can output validated JSON objects using provided JSON Schemas.
• Deterministic Execution: Enforces reproducible workflows by sandboxing script execution and limiting API access.
• Live Progress Visualization: Provides inline, real-time updates on workflow and subagent status.

Maintenance & Community

No specific details regarding contributors, sponsorships, or community channels (e.g., Discord, Slack) are present in the provided README.

Licensing & Compatibility

• License: MIT.
• Compatibility: Designed as a Pi extension; no explicit restrictions for commercial use are noted.

Limitations & Caveats

This project is explicitly labeled as a prototype. It currently lacks features such as persisted or resumable workflow runs, and a dedicated manager interface for workflows.