How kaleido34/n2n works
Project Type & Purpose
• Full‐stack web application for ingesting various content types (PDF, video, audio, image), processing them with AI (OCR, transcription, summarization, flashcards, mind maps, quizzes, etc.), and organizing results in user-specific “spaces.”
• Frontend: Next.js/React/Tailwind
• Backend: Next.js API routes + Python extraction server
• Persistence: PostgreSQL (via Prisma), vector search (Pinecone)
• Auth & BaaS: Supabase
• External AI: Google AI SDK
External Entities:
• User (web/mobile client)
• Admin (reindexing, management tasks)
• Google AI / OCR / Transcription service
• Pinecone Vector DB
• Supabase Auth service
Processes:
1.0 User Management
2.0 Content Ingestion
3.0 Content Processing
4.0 Workspace Management
5.0 Admin & Reindex
Data Stores:
A. User Store (PostgreSQL)
B. Raw Content Store (file blobs or object storage)
C. Processed Content Store (PostgreSQL + Prisma)
D. Embeddings Store (Pinecone)
Data Flows:
– User → 1.0 [signup/signin/profile data] → A
– User → 2.0 [upload requests + files] → B
– 2.0 → 3.0 [raw content pointer]
– 3.0 ↔ Google AI / OCR / Transcribe → 3.0
– 3.0 → C [summaries, transcripts, flashcards, etc.]
– 3.0 → D [vector embeddings]
– User → 4.0 [create/read/update spaces] → C
– Admin → 5.0 [reindex commands] → C/D
DFD Level 1 (Breakout of Context Processes)
1.0 User Management
1.1 Signup
1.2 Signin
1.3 Update Profile
1.4 Deactivate Account
2.0 Content Ingestion
2.1 Ingest PDF
2.2 Ingest YouTube Video
2.3 Ingest Audio File
2.4 Ingest Image (OCR)
3.0 Content Processing
3.1 Transcribe Audio/Video
3.2 OCR Documents/Images
3.3 Summarize Text/Audio
3.4 Generate Flashcards
3.5 Generate Mind Map
3.6 Generate Quiz
3.7 Generate Concept Matches
3.8 Generate Terms
4.0 Workspace Management
4.1 Create/Retrieve/Update/Delete Spaces
4.2 Assign Content to Spaces
5.0 Admin & Reindex
5.1 Reindex All Embeddings
5.2 Reindex Single Content Embeddings