Summary

AICartographer addresses the complexity of large Unreal Engine 5 projects by transforming Blueprints into navigable, narrative "maps." It targets developers needing rapid codebase comprehension, offering LLM-driven insights and visualizations to drastically reduce onboarding time and improve understanding.

How It Works

This UE5 React plugin and Python FastAPI backend analyzes Blueprint code via AST parsing. It generates detailed Markdown narratives and force-directed graphs, storing them in a local .aicartographer/vault/ directory. This local-first approach ensures user control over LLM keys and avoids Git merge conflicts, prioritizing developer workflow.

Quick Start & Requirements

The recommended portable package (~1.6 MB zip) offers a ~10-minute setup on Windows 10/11, requiring UE 5.6+, VS 2022 (C++ workload), Python 3.11+ (added to PATH), and an LLM API key. Alternatively, a source build (~30 min) requires cloning the repo, npm install, and pip install. Links: GitHub Releases, HANDOFF.md.

Highlighted Details

• Five-Level View: Navigates from project overview (Lv0) down to function execution flow (Lv3) and cross-blueprint call traces (Lv4).
• AI Narrative: LLM generates structured Markdown for each Blueprint (Intent, Flow, Interactions, Coupling, Risk).
• AIChat Tools: Multi-turn RAG agent with tools for search, code analysis, and structured blueprint edit suggestions (propose_edit).
• Multi-Vendor LLM: Supports Volcengine, Anthropic Claude, and OpenAI-compatible endpoints.
• Local Vault: .md files are stored locally, not in Git, preserving developer-specific LLM configurations and notes.

Maintenance & Community

Primarily maintained by Liamour. Community interaction via GitHub Issues.

Licensing & Compatibility

Licensed under the permissive MIT License, allowing for commercial use and integration into closed-source projects.

Limitations & Caveats

The propose_edit feature currently outputs suggestions (D0); full in-editor node composition is planned (D1). L3 function flow visualization lacks LLM analysis. Blueprint scanning is prioritized over C++ or Interface analysis. Live Coding limitations require full C++ module rebuilds for new UFUNCTIONs. Language switching requires rescanning to update existing .md files.