AI Knowledge Management Platform

The Challenge

Researchers, professionals, and teams were drowning in unstructured AI conversation threads. Critical insights from ChatGPT and other AI assistants were buried in endless chat histories, making it impossible to find, organize, and leverage knowledge effectively. Users needed a way to transform conversational AI chaos into actionable, searchable knowledge.

The Solution

Paneglass.ai is a comprehensive knowledge management platform that uses advanced AI processing to automatically organize, synthesize, and structure conversations into intelligent knowledge bases. The system employs hierarchical categorization, semantic search, and unified knowledge trees to create a living repository of insights.

A Chrome browser extension enables one-click import of ChatGPT conversations with automatic detection, secure authentication, and real-time status notifications. The extension extracts conversation data from ChatGPT’s internal structures, supporting both new imports and incremental updates to existing conversations.

The AI processing pipeline uses GPT-4o-mini, GPT-4.1-nano, or GPT-5-mini to analyze conversations and create hierarchical structures (Macro Topics → Subtopics → Chunks). An optional two-stage parallel processing mode achieves 3x performance improvement on complex conversations by identifying structure first, then generating content simultaneously.

Vector embeddings power semantic search across all knowledge chunks with cosine similarity matching, finding relevant content even when exact keywords don’t match. The unified knowledge tree intelligently merges conversations from multiple sources using LLM-powered logic that determines when to merge, create subtopics, or add new topics based on similarity thresholds.

All content is protected with end-to-end encryption using user-specific master keys, applied to raw conversation text, Q&A pairs, and organized knowledge structures before database storage.

Key Features

• Browser Extension — One-click ChatGPT import with automatic detection and incremental updates
• AI Organization — Hierarchical structure creation with variable depth based on content complexity
• Parallel Processing — Two-stage mode for 3x faster processing on complex conversations
• Semantic Search — Vector embeddings with cosine similarity for natural language queries
• Knowledge Trees — LLM-powered merging of conversations into unified, growing knowledge bases
• Source Tracking — Visual indicators showing which conversations contributed to each node
• Content Synthesis — Markdown-formatted summaries with educational content extraction
• Project Workspaces — Collaborative management with isolation and access controls
• E2E Encryption — User-specific master keys for all stored content
• Flexible Views — Column-based tiles with color-coded topics and zoom controls
• Smart Deduplication — Automatic detection and merging of similar topics

Technologies

• Frontend — Next.js 15, React 19, TypeScript, TailwindCSS
• Backend — Supabase (PostgreSQL, Auth, Row Level Security)
• AI/ML — OpenAI GPT-4o-mini/GPT-4.1-nano/GPT-5-mini, text-embedding-3-small
• Extension — Chrome Manifest V3, Content Scripts, Background Service Worker
• UI — Radix UI, React Flow, D3.js for tree visualization
• Security — AES encryption, user-specific master keys
• Validation — Zod schemas for type-safe data handling