INIT.md

AI Note Taker App - Project Overview

Application Overview

An AI-powered note taker app that automatically joins meetings, transcribes conversations, and generates AI summaries.

Core Flow

1. User authentication via Google OAuth 2
2. User grants Google Calendar access on home screen
3. Webhook syncs Google Calendar events
4. Bot automatically joins meetings with the user
5. Live transcription during meetings
6. Post-meeting transcription storage and AI summary generation

Tech Stack

Backend - Java Service

• Language: Java
• Framework: Spring Boot
• Build Tool: Maven
• Database: PostgreSQL
• DB Migrations: Flyway

Integration & Communication

• Message Queue: Kafka (for triggering Python AI service)
• Schema Management: Protocol Buffers (protobuf)
• Service Communication: gRPC (primary), HTTP, Webhooks
• External APIs: Google OAuth 2, Google Calendar webhooks

AI Service (External)

• Language: Python (separate service)
• Triggered via: Kafka events
• Purpose: Transcription and AI summary generation

Architecture Components

• Java Spring Boot main service
• PostgreSQL database with Flyway migrations
• Kafka message broker
• Python AI service (external)
• Google OAuth 2 integration
• Google Calendar webhook integration
• gRPC service definitions
• Protocol Buffer schemas

MVP Features

1. Google OAuth 2 Authentication

• Endpoint 1: Initiate OAuth flow with client ID, secret, and redirect URI
• Endpoint 2: Handle redirect callback and receive access token from Google
• Standard OAuth 2 authorization code flow implementation

2. Google Calendar Integration

• Use access token to request Google Calendar scope
• Implement webhook-based calendar sync
• Real-time event synchronization from Google Calendar

3. Meeting Bot Integration

• Domain Event: Emit meeting start event when calendar meeting begins
• External Service Consumption: Python service consumes meeting start event
• Bot Behavior: External service deploys bot to join meeting
• Meeting End Event: External service emits meeting end event with transcription
• Event Consumption: Java service consumes meeting end event and processes transcription

4. AI Summary Generation (Temporal Workflow)

• Temporal Integration: Use Temporal workflow engine for long-running processes
• RPC Call: Make gRPC call to external AI service for summary generation
• Workflow Completion: Store generated summary and complete Temporal workflow
• Async Processing: Handle AI summary generation as background workflow

Event Flow

1. Meeting Start → Domain Event → External Service Bot Joins
2. Meeting End → External Service Event → Transcription Received
3. Transcription → Temporal Workflow → AI Summary RPC → Summary Stored

Microservices Architecture (3 Services)

1. Auth Service

• Google OAuth 2 flow implementation
• User authentication & JWT management
• Token validation for other services
• Database: User auth data, OAuth tokens

2. User Service

• User profiles, settings, preferences
• Account management
• Database: User profiles, settings

3. Calendar Service (Extended Responsibilities)

• Google Calendar integration & webhooks
• Meeting lifecycle management (creation, status tracking)
• Transcription handling (storage, processing)
• Bot coordination (domain events for external service)
• Temporal workflow integration (AI summary generation)
• gRPC calls to external AI service
• Database: Calendar events, meetings, transcriptions, summaries, temporal workflows

Event-Driven Communication

• No separate Event Service
• Kafka consumers/producers embedded in each service
• Each service configuration includes:
  • Kafka broker URLs
  • Schema Registry URL
  • Topic configurations
  • Consumer/producer properties

Kafka Topics

• meeting-start-events (Calendar → External Bot Service)
• meeting-end-events (External Bot Service → Calendar)
• transcription-ready-events (Internal Calendar Service workflows)

Next Steps

Ready to implement features step by step as directed.