AutoAlign

Dealing with dusty, 50-page policy documents is a universally frustrating experience for developers. It is incredibly difficult to manually cross-reference every new feature against massive governance docs, which often leads to avoidable security risks, compliance rework, and delayed projects.

AutoAlign is designed to solve this exact reality. It is an AI-powered tool built by the "Ninja Turtles" team for HackFest 2.0 that transforms static policy documents into an active, autonomous governance system.

Here is a brief breakdown of what the project does and how it works:

Core Concept

Instead of relying on humans to manually check Business Requirement Documents (BRDs) against company policies, AutoAlign uses an AI agentic workflow to automatically review, critique, and rewrite the documents until they are fully compliant with internal rules (like data privacy, security standards, and coding practices).

How It Works: The "Debate Loop"

The heart of AutoAlign is a multi-agent state machine built with LangGraph. It relies on two AI agents that "debate" with each other to fix your document:

1. The Defender Agent (Policy Guardian): This agent uses Retrieval-Augmented Generation (RAG) to pull rules from your company's knowledge base. It scans the provided BRD and flags any policy violations (e.g., storing passwords in plaintext or missing rate limits).

2. The Drafter Agent (Architect Solution): When the Defender flags an issue, the Drafter steps in. It rewrites the BRD to fix the violations while preserving the original business intent of the feature.

3. The Loop: The revised document goes back to the Defender for another check. This cycle repeats until the document is 100% compliant or hits a maximum iteration limit.

Once finished, the system outputs the newly aligned, compliant BRD along with a detailed audit report of what was changed and why.

Key Technical Details

• AI Engine: Google Gemini 1.5 Pro, chosen for its large context window to handle massive markdown files.

• Orchestration: LangGraph and LangChain for managing the back-and-forth agent workflow.

• Vector Storage: ChromaDB (local) or Vertex AI (production) to store and retrieve the policy documents.

• Interfaces: It includes a backend API, a frontend UI, a Command Line Interface (CLI), and a custom Python SDK called Turgon for easy integration into existing codebases.

Ultimately, AutoAlign acts as an automated compliance reviewer that fixes your mistakes before a single line of code is ever written.