himanijha/foldednotes

1. Overall shape

- This is a Next.js Pages Router application with both UI pages and backend API routes in the same repo.

- The architecture is split into:

- Client-facing pages/components under src/pages and src/components

- Server-side API endpoints under src/pages/api

- Shared domain/infrastructure helpers under src/lib

- A small custom server layer under src/server plus server.js

2. Primary product flow

- The app is built around anonymous voice messages: users record audio, submit it for transcription/processing, and then consume or generate content from that transcript.

- The README also calls out a notes feed where users can post short notes, including a side-panel “write a note” workflow.

3. Frontend entry points

- Main UI screens live in:

- src/pages/index.tsx and src/pages/home.tsx for landing/home

- src/pages/record.tsx for audio capture

- src/pages/generate.tsx for transcript-driven generation

- src/pages/memories.tsx for saved/archived content

- src/pages/login.tsx and src/pages/settings.tsx

- _app.tsx and _document.tsx provide shared app-level wiring and document setup.

- Styling is page/component-scoped via CSS modules in src/styles plus global styles.

4. Client components and UX boundaries

- AudioRecorder.tsx is the main recording UI component; it likely owns microphone capture, recording state, and upload/submit actions.

- GenerateFromTranscript.tsx appears to handle turning a transcript into generated content or follow-up artifacts.

- These components are the primary bridge from user interaction into the API layer.

5. API layer as the backend boundary

- The app uses Next.js API routes as its server boundary. Key endpoints include:

- POST /api/auth/login, signup, me

- POST /api/notes

- POST /api/transcribe

- POST /api/summary

- POST /api/classify

- POST /api/speech

- GET/POST /api/voices, /api/pick-voice

- GET/POST /api/recordings

- This means the browser never talks directly to MongoDB or third-party AI services; it goes through these routes.

6. Identity and session subsystem

- src/lib/auth.ts and the auth API routes indicate a dedicated authentication subsystem.

- The presence of /api/auth/me suggests the frontend can hydrate the current user/session state from the backend.

- The repo supports authenticated flows for some actions while still emphasizing anonymous contributions as a product concept.

7. Persistence layer

- MongoDB is the main database, with connection helpers in src/lib/mongodb.ts and src/server/mongodb.ts.

- Data likely persisted here includes:

- Users/auth records

- Notes/feed items

- Recordings and transcript metadata

- Possibly summaries, classifications, and selected voice preferences

- src/pages/api/recordings.ts and src/pages/api/notes.ts are the clearest persistence-facing endpoints.

8. AI/LLM processing subsystem

- src/lib/gemini.ts and src/lib/prompt.ts show an LLM-backed generation layer, likely used for summarization, rewriting, or content generation from transcripts.

- /api/summary and /api/classify are the service boundaries for turning raw transcript content into structured outputs.

- src/lib/emotions.ts and src/types/emotion.ts suggest a dedicated emotion/labeling model used in classification or content tagging.

9. Audio and speech subsystem

- src/lib/audio.ts handles audio-specific transformations or transport concerns.

- The README documents POST /api/transcribe as an ElevenLabs Speech-to-Text integration using base64-encoded audio.

- POST /api/speech, /api/voices, and /api/pick-voice indicate a text-to-speech / voice-selection pipeline, also likely external-service-backed.

- This makes audio an explicit data pipeline:

- browser recording → API upload → transcription → transcript processing → optional speech/voice output

10. External service dependencies

- ElevenLabs is a required external service for transcription, and likely also voice generation.

- Gemini is used for AI text processing/generation.

- MongoDB is the persistence backend.

- These are the main off-repo dependencies that affect architecture and deployment configuration.

11. Custom server and runtime

- server.js plus src/server/index.ts suggest there is a custom Node server entrypoint or server bootstrap in addition to Next.js defaults.

- This is relevant because it may control startup behavior, database initialization, or route mounting beyond standard Next.js deployment.

- The project still remains fundamentally a Next.js app, but with server-side bootstrap logic layered on top.

12. End-to-end data flow summary

- Typical flow:

- User records audio in the browser

- Client submits audio to /api/transcribe

- Transcript may then be sent to /api/summary, /api/classify, or /api/speech

- Results and/or notes are persisted through MongoDB-backed API routes

- The UI retrieves user state and content through API routes like /api/me, /api/notes, and /api/recordings

13. Architectural boundaries to show in a diagram

- Presentation layer: pages + reusable React components

- API layer: Next.js route handlers under /api/*

- Domain helpers: auth, audio, prompts, emotions, AI integration

- Data layer: MongoDB

- External AI services: ElevenLabs and Gemini

- Static assets/styles: public and src/styles