Graphify Atlas — Graph-Powered Knowledge Workspace

A cinematic, graph-backed knowledge workspace that turns a private Markdown corpus into a queryable knowledge graph with conversational retrieval, evidence inspection, and Obsidian browsing.

Overview

Graphify Atlas is a graph-powered knowledge workspace that turns a private document corpus into an explorable knowledge graph — supporting conversational retrieval, graph-backed evidence inspection, image/table previews, and optional Obsidian browsing of community clusters and node notes.

The published repo is intentionally a data-free shareable shell: it ships the product architecture, frontend, and retrieval scaffolding, while the private corpus and generated graph outputs stay local and out of version control — a clean pattern for open-sourcing a UI without exposing confidential data.

Implementation

Graph retrieval: built on Graphify, which parses a local Markdown corpus into graph outputs (graph.json, an HTML graph view, a wiki, and an Obsidian vault) used as the retrieval source
Workspace UI: a landing page plus a dedicated /home workspace split into an Atlas Rail (context & dataset stats), a Conversation Canvas (main interaction), a Signal Rail (source signals & related nodes), and a floating Evidence Viewer for focused graph exploration
Frontend: Next.js + React + TypeScript + Tailwind CSS with shadcn/ui primitives and a cinematic, motion-driven landing experience
Chat: conversational Q&A over the generated graph via the OpenAI API, available both in the web workspace and a terminal chat helper
Local-first workflow: PowerShell scripts (build_kb / chat_kb / run_frontend) that bootstrap a venv, install Graphify, build the graph from data_markdown/, and launch the app
Obsidian integration: generated node notes, community-cluster overviews, and a graph canvas that can be opened directly as an Obsidian vault for visual knowledge-map navigation