Lamina is an AI-powered note-taking app focused on fast capture, structured notes, and a fluid writing experience. Below are the product-grade features and where they live in the codebase.

• Live AI streaming

  • Frontend stream reader: streamAiResponse (lib/prompt.js)
  • Backend streaming endpoint: app/api/ai-stream/route.js (streamText integration)
  • Typing UX: incremental chunks are rendered to the editor as they arrive.

• Multi-source parsing (convert sources to text before summarizing)

  • PDF parsing: parsePdf → app/api/parse-pdf/route.js
  • YouTube transcripts: getYoutubeTranscript (lib/prompt.js)
  • DOC/DOCX parsing: parseDoc (lib/prompt.js)
  • Webpage extraction: fetchWebpageText (lib/prompt.js)
  • Paste/long text: getLongText (lib/prompt.js)

• Rich WYSIWYG Markdown editor

  • Editor component: components/MarkdownEditor.jsx
  • Live-rendering: streamed AI output is processed via marked and injected so markdown formatting appears as the AI types.
  • Tiptap extensions included (lists, links, images, highlights, tasks).

• Productivity UX

  • Tabbed editor system: components/TabSystem.jsx
  • Sidebar file explorer with folders and breadcrumbs: components/Sidebar.jsx
  • Auto-save and Supabase sync (client created in components using env vars)

• Clear separation:

  • Prompt and streaming logic centralized in lib/prompt.js (makeNotesPrompt, streamAiResponse).
  • API routes in app/api.
  • Editor UI in components/MarkdownEditor.jsx.

• Streaming-first design:

  • Backend returns a streaming response to the client so the frontend can render tokens in real time.
  • The editor receives tokens via streamAiResponse and appends them using Tiptap commands.

• Safe defaults and debugging:

  • Token truncation to prevent model context errors.
  • Console logging hooks in API routes (add logs in app/api/parse-pdf/route.js and app/api/ai-stream/route.js while debugging).

1. Install:

npm install

2. Add environment variables in .env.local:

• NEXT_PUBLIC_SUPABASE_URL
• NEXT_PUBLIC_SUPABASE_ANON_KEY
• OPENROUTER_API_KEY

3. Run:

npm run dev

• Generate notes from a PDF:

  1. Upload via the custom dialog in components/MarkdownEditor.jsx.
  2. Frontend calls parsePdf → POST /api/parse-pdf (app/api/parse-pdf/route.js).
  3. Build prompt via makeNotesPrompt and call streamAiResponse.
  4. AI streams notes into the editor live.

• Convert a webpage or YouTube transcript:

  • Use fetchWebpageText or getYoutubeTranscript, then stream notes as above.

• Prompt & streaming: lib/prompt.js — see makeNotesPrompt and streamAiResponse.
• Editor: components/MarkdownEditor.jsx.
• Streaming API: app/api/ai-stream/route.js.
• PDF parser: app/api/parse-pdf/route.js.
• Tabs and layout: components/TabSystem.jsx and app/Application.jsx.