Live app: auroraq.onrender.com

secret key : tombat

Multimodal Gemini App

A FastAPI service that exposes a clean REST API for multimodal (text, image, and audio) conversations powered by Google Gemini.

---

Render Deployment

This project is deployed on Render as a web service. The deployment flow is:

1. Connect the GitHub repository to Render.
2. Create a new Web Service from the repo.
3. Set the build command to install dependencies, for example pip install -r requirements.txt.
4. Set the start command to launch the FastAPI app on Render's assigned port, for example uvicorn main:app --host 0.0.0.0 --port $PORT.
5. Add the required environment variables in the Render dashboard, especially GEMINI_API_KEY and any optional settings you use locally.
6. Deploy the service and wait for Render to finish the build and start the app.

After deployment, the public app is available at https://auroraq.onrender.com/.

Features

• Text chat – multi-turn conversations via /api/v1/chat
• Image + text – send an image URL or upload a file with a prompt
• Audio + text – upload an audio file and ask Gemini about it
• File upload – store images and audio in static storage and get back a URL
• Optional API-key auth – bearer-token guard on every endpoint
• Rate limiting – configurable via environment variables
• Optional chat history – SQLite schema included (scripts/init_db.py)

---

Project Structure

multimodal-gemini-app/
├── .env                        # Environment variables (copy from this file)
├── requirements.txt            # Python dependencies
├── README.md                   # This file
├── main.py                     # FastAPI entrypoint
├── config/
│   ├── __init__.py
│   └── settings.py             # GEMINI_API_KEY, model name, timeouts, etc.
├── api/
│   ├── __init__.py
│   └── v1/
│       ├── __init__.py
│       ├── base.py             # Shared base router
│       ├── routes.py           # /chat, /upload endpoints
│       └── models.py           # Pydantic models (TextChatRequest, ImageChatRequest, …)
├── services/
│   ├── __init__.py
│   ├── gemini_service.py       # Gemini API calls (text, image + text, audio + text)
│   ├── image_service.py        # Download / validate images, prepare for Gemini
│   ├── audio_service.py        # Validate / upload audio, prepare for Gemini
│   └── orchestrator.py         # Glue: image + text + audio → Gemini prompt
├── utils/
│   ├── __init__.py
│   ├── file_helpers.py         # Save uploaded files, generate public URLs
│   ├── security.py             # Bearer-token auth dependency
│   └── logger.py               # Centralised logging
├── static/
│   └── uploads/
│       ├── images/             # Uploaded images (git-ignored except .gitkeep)
│       └── audio/              # Uploaded audio  (git-ignored except .gitkeep)
└── scripts/
    └── init_db.py              # Optional SQLite schema init for chat history

---

Quick Start

1. Clone & install dependencies

git clone https://github.com/Karn2898/Multimodal_LLM.git
cd Multimodal_LLM
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

2. Configure environment variables

Copy the .env file and fill in your Google Gemini API key:

cp .env .env.local   # or just edit .env directly

Variable	Default	Description
GEMINI_API_KEY	(required)	Your Google AI Studio API key
GEMINI_MODEL	gemini-1.5-pro	Model to use
GEMINI_TIMEOUT	60	Request timeout in seconds
APP_HOST	0.0.0.0	Uvicorn bind host
APP_PORT	8000	Uvicorn bind port
APP_RELOAD	true	Hot-reload in development
MAX_IMAGE_SIZE_MB	10	Maximum image upload size
MAX_AUDIO_SIZE_MB	25	Maximum audio upload size
API_SECRET_KEY	(empty = disabled)	Bearer token for auth

3. (Optional) Initialise the chat-history database

python scripts/init_db.py

4. Run the server

python main.py
# or
uvicorn main:app --host 0.0.0.0 --port 8000 --reload

Open the interactive docs at http://localhost:8000/docs.

---

API Endpoints

Method	Path	Description
GET	/	Health check
POST	/api/v1/chat	Text-only chat
POST	/api/v1/chat/image	Image URL + text chat
POST	/api/v1/chat/image/upload	Uploaded image + text chat
POST	/api/v1/chat/audio	Uploaded audio + text chat
POST	/api/v1/upload/image	Upload an image file
POST	/api/v1/upload/audio	Upload an audio file

Example – text chat

curl -X POST http://localhost:8000/api/v1/chat \
  -H "Content-Type: application/json" \
  -d '{"message": "What is multimodal AI?"}'

Example – image chat (URL)

curl -X POST http://localhost:8000/api/v1/chat/image \
  -H "Content-Type: application/json" \
  -d '{"message": "Describe this image.", "image_url": "https://example.com/photo.jpg"}'

Example – audio chat (file upload)

curl -X POST http://localhost:8000/api/v1/chat/audio \
  -F "message=Transcribe this audio." \
  -F "file=@/path/to/audio.mp3"

---

License

MIT