v0.1.0-alpha.1 — GraphRAG Workbench (First alpha)

GraphRAG Workbench — Initial Alpha

Entire project initialization (first public release).

Demo video:
https://github.com/user-attachments/assets/1f588a45-07ca-4953-92ed-fc888fe28cff

Overview

A modern web app for building and visualizing knowledge graphs using Microsoft’s GraphRAG. Convert documents into an explorable 3D knowledge graph, run indexing pipelines, and manage archived knowledge graph snapshots.

How It Works

• Add PDFs to input/ (UI uploads supported). PDFs are converted to .txt and originals archived to input/_pdfs.
• Run the GraphRAG pipeline (OpenAI models via settings.yaml).
• Parquet outputs are mirrored to JSON in output/, which the visualizer loads.
• Archives let you snapshot and restore KGs; Restore performs a true swap (current ↔ selected archive).
• Terminal Logs show live, colorized logs while indexing.

Features

• End‑to‑end GraphRAG indexing (OpenAI) and JSON export
• 3D graph viewer with filters, search, and community insights
• Archive management: create/list/rename/delete; swap‑restore
• Terminal Logs: sticky bottom, level colors (INFO/WARN/ERROR), inline JSON highlighting

Security/Hardening

• Upload filename sanitization (prevents path traversal)
• File serving rooted at repo (process.cwd())
• pdf-parse import fix to avoid test‑asset ENOENT
• ESLint clean

Known Issues

• Some TypeScript types in 3D/graph files still need refinement (does not block typical usage)
• Optional Python/pyarrow may be needed for certain Parquet variants

Getting Started

1. Set OPENAI_API_KEY in .env
2. Put PDFs in input/ (or upload via UI)
3. Click “Run Index” and watch Terminal Logs
4. Explore the graph and manage archives

Changelog

• Initial project creation and first alpha release
• PDF→TXT ingestion, parquet→JSON conversion
• Archive swap‑restore and improved Terminal Logs