Vectorizer Dashboard
Semantic memory dashboard for Vectorizer — visualizes workspaces, embeddings, knowledge graph, and enables RAG Q&A.
Overview
A Next.js 14 dashboard for Vectorizer — the self-hosted semantic memory server. Built to visualize and interact with stored memories, embeddings, and the knowledge graph.
Features
- Dashboard Overview — health status, workspace count, embedding model, LLM brain status
- Workspace Browser — browse all workspaces, view messages with role badges
- Semantic Search — search interface with relevance scores and filters
- RAG Q&A — chat interface powered by Vectorizer's LLM brain
- Embedding Visualization — PCA projection of 768d vectors to 2D scatter plot
- Knowledge Graph — force-directed canvas visualization of GRAPH.json (1419 nodes, 7479 edges)
- Analytics — Recharts bar/pie charts for workspace and activity metrics
Tech Stack
- Next.js 14, TypeScript, Tailwind CSS (dark theme)
- Recharts for charts
- Custom PCA for client-side dimensionality reduction
- Canvas-based force-directed graph with alpha cooling
API Sources
- Vectorizer API (port 8091) — messages, search, brain/RAG
- ChromaDB v2 API (port 8100) — raw vectors for embedding visualization
- GRAPH.json — knowledge graph data
news(69)
fix: search API uses 'where' object for workspace_id, not top-level field
Date: 2026-08-30T12:24:11Z
New commit by alfirus: "fix: search API uses 'where' object for workspace_id, not top-level field" (298f33d)