🎙️ AuraArchive

AuraArchive is an intelligent audio-to-blog platform that transforms raw audio discussions into high-quality, structured, and publishable tech articles.

Video Demo - Click Here

Made with ❤️ by Nirvan and Dhairya

---

🏗️ Architecture

AuraArchive utilizes a modern, event-driven architecture that follows a Cloud-Native, Mobile-First approach. It separates heavy AI processing on the backend from seamless content management and consumption on the Android client.

🔄 System Data Flow

graph TD
    %% Mobile Client
    subgraph "Mobile Client (Android)"
        App["Jetpack Compose UI"] -->|Upload Audio| ViewModel["Hilt ViewModel"]
        ViewModel -->|Retrofit| API["FastAPI Backend"]
        App -->|Export| PDF["Native PDF/HTML Engine"]
    end

    %% Backend Service
    API -->|Save Temp File| LocalStorage["Temp Storage"]
    API -->|Create Record: PROCESSING| Qdrant["Qdrant Vector DB"]
    API -->|Trigger Background Task| Worker["Background Processor"]

    subgraph "AI Processing Pipeline"
        Worker -->|Upload Audio| Gemini["Google Gemini 2.5 Flash"]
        Gemini -->|Transcribe & Analyze| Context["Context Window"]
        Context -->|Generate JSON| Blog["Structured Blog Post"]
        Context -->|Extract Links| Links["External Resources"]
    end

    Worker -->|Update Record: REVIEW_PENDING| Qdrant

    Admin["Admin Dashboard"] -->|Fetch Drafts| API
    Admin -->|Edit & Save| API
    Admin -->|Publish| API

    API -->|Update Status: PUBLISHED| Qdrant

    Public["Public Feed"] -->|Fetch Published| API

Loading

⚙️ App Specifications & Tech Stack

📱 Mobile App (Android)

• Language: Kotlin
• UI Framework: Jetpack Compose with Material 3
• Architecture: MVVM (Model-View-ViewModel)
• Dependency Injection: Dagger Hilt
• Networking: Retrofit + OkHttp
• Image Loading: Coil
• Export Engine: Native Android PdfDocument and a custom HTML-to-Doc Intent system for seamless Google Docs integration

🖥️ Backend Core

• Framework: FastAPI (Python) - High performance, async-first
• Database: Qdrant (Cloud) - Vector database used for scalable document storage and (future) semantic search
• AI Model: Google Gemini 2.5 Flash - Multimodal model for processing audio and generating structured JSON output

🚀 Key Features

• Audio-to-Blog: Direct upload of MP3/WAV files from mobile to the AI processing pipeline.
• Structured AI Output: Gemini 2.5 Flash ensures valid JSON with dedicated fields for title, summary, content, and external_links.
• Seamless Export: One-tap export of generated blogs directly to Google Docs for final editing or PDF for standard sharing.
• Modern Mobile Experience: Custom-themed interface featuring time-based greetings and a daily "Inspiration Engine" powered by the Quotable API.
• Review System: AI-generated drafts enter a REVIEW_PENDING state, allowing for administrative oversight before going public.

🔌 API Documentation

1. Upload & Processing

• Endpoint: POST /api/upload
• Input: file (Audio file)
• Action: Uploads file and initiates background AI reasoning task
• Response: { "id": "uuid", "message": "Upload accepted" }

2. Draft Management (Admin)

• Endpoint: GET /api/drafts

  • Returns: List of discussions with status REVIEW_PENDING

• Endpoint: PUT /api/save/{id}

  • Input: JSON { "title": "...", "summary": "...", "blog_markdown": "..." }
  • Action: Updates the content of a draft.

• Endpoint: POST /api/publish/{id}

  • Action: Changes status to PUBLISHED

3. Public Access

• Endpoint: GET /api/feed
  • Returns: List of discussions with status PUBLISHED

🚀 Setup & Installation

Android (Mobile)

1. Clone the repository.
2. Open in Android Studio (Ladybug or newer).
3. Ensure proper Base URLs are configured in your Hilt AppModule.kt.

Backend (Python)

Prerequisites: Python 3.10+, Qdrant Cloud Cluster, Google AI Studio API Key.

Environment Variables (.env):

Create a .env file in the root directory:

GEMINI_API_KEY=your_gemini_key
QDRANT_URL=your_qdrant_url
QDRANT_API_KEY=your_qdrant_key
CLOUDINARY_CLOUD_NAME=your_cloudinary_name
CLOUDINARY_API_KEY=your_cloudinary_key
CLOUDINARY_API_SECRET=your_cloudinary_secret

Running the Server:

# Install dependencies
pip install -r requirements.txt

# Start the server (Dev Mode)
python -m uvicorn backend.main:app --reload