Architecture

Docxa is built with a modular, extensible architecture designed for high-fidelity documentation generation.

Component Overview

• Core Engine: Manages the workspace state, document graph, and template library.
• Agents: Specialized AI personas (PM, Architect, Developer) driven by Ax-LLM.
• Analysis Layer: Pluggable framework detectors and repository scanners.
• Interview Engine: Handles role-aware conversations and persists session state.
• Generation Planner: The "Brain" that evaluates if enough evidence exists to safely generate a document.

Data Flow

graph TD
    Repo[Source Code] -->|Discover| Analysis[.docxa/analysis.json]
    User[Stakeholder] -->|Interview| Session[.docxa/sessions/]
    Analysis --> Planner[Generation Planner]
    Session --> Planner
    Templates[Templates] --> Planner
    Planner -->|Execute| LLM[Ax-LLM]
    LLM -->|Output| MD[Generated Markdown]

State Management

Docxa maintains all its state in a local .docxa/ directory within your project. This includes analysis metadata, active interview sessions, and cached AI responses. This directory should typically be ignored by your version control system.