Workforces AI

Project Overview

While single-turn LLM chat assistants can generate isolated code snippets, they struggle with large-scale refactors, cross-file architectural integrity, and automated quality assurance.

Workforces was developed as an end-to-end framework that transforms AI coding agents into a cohesive, autonomous software engineering organization:

• dev (Clean Coder): Enforces strict TDD, symbol graph discovery, and post-edit AST verification.
• advisor & strategy: Unpacks root problems, unbundles monolithic architectures, and designs validation experiments.
• scribe: Maintains immutable session lineage, persistent task tracking, and structured decision context.

Architectural Highlights

1. AST Code-Graph Indexer & Symbol Discovery

• Before modifying or generating code, subagents invoke an AST graph parser to locate existing utilities and classes.
• Completely prevents duplicate helper functions and enforces DRY principles across multi-repo workspaces.

2. Automated Post-Code Review & Self-Healing Hooks

• Immediately analyzes modified diffs against contract signatures, swallowed error patterns, and test boundaries.
• Triggers autonomous self-healing loops to remediate issues before finalizing tasks.

3. Open Knowledge Format (OKF) & Lineage Persistence

• Stores architectural context, ADRs, and session states in clean, token-efficient Markdown tables.
• Enables seamless context handover across asynchronous agent sessions.