A Flask-based REST API service that integrates Fish Audio's ASR (Automatic Speech Recognition) and TTS (Text-to-Speech) capabilities. Built for CalHacks 2025.

• ASR Endpoint: Convert audio files to text transcriptions
• TTS Endpoint: Convert text to synthesized speech audio
• JanitorAI Integration: Example script for accessing JanitorAI completions API
• Mock Mode: Development mode with mock responses (no API calls)
• Health Check: Monitor service status and configuration

calhacks25/
├── src/
│   ├── fish/                  # Fish Audio integration module
│   │   ├── __init__.py
│   │   ├── client.py          # Session manager & API configuration
│   │   ├── asr.py             # Speech-to-text functionality
│   │   └── tts.py             # Text-to-speech functionality
│   └── app.py                 # Flask REST API server
├── app/                       # Web frontend (optional)
│   ├── static/
│   │   ├── css/
│   │   ├── js/
│   │   └── images/
│   └── templates/
│       └── index.html
├── main.py                    # JanitorAI example script
├── requirements.txt
├── .env                       # Environment configuration
└── README.md

source venv/bin/activate  # On macOS/Linux
# or
venv\Scripts\activate  # On Windows

pip install -r requirements.txt

Create or edit .env file with your Fish Audio API credentials:

# Fish Audio API Configuration
FISH_API_KEY=your_api_key_here
FISH_API_BASE=https://api.fish.audio
FISH_MOCK=false  # Set to 'true' for mock mode (no API calls)

# Flask Server Configuration
HOST=0.0.0.0
PORT=5000
FLASK_DEBUG=true

cd src
python app.py

The API will be available at http://0.0.0.0:5000

python main.py

GET /health

Returns service status and configuration info.

POST /asr
Content-Type: application/octet-stream
# OR
Content-Type: multipart/form-data (with 'audio' field)

# Body: audio file bytes (WAV, MP3, etc.)

Returns JSON with transcribed text.

Example:

curl -X POST http://localhost:5000/asr \
  -H "Content-Type: application/octet-stream" \
  --data-binary @audio.wav

Response:

{
  "transcript": "Hello world",
  "mock": false
}

POST /tts
Content-Type: application/json

{
  "text": "Hello world"
}

Returns JSON with base64-encoded audio data.

Example:

curl -X POST http://localhost:5000/tts \
  -H "Content-Type: application/json" \
  -d '{"text": "Hello world"}'

Response:

{
  "audio": "base64_encoded_audio_data...",
  "format": "wav",
  "mock": false
}

Set FISH_MOCK=true in .env to use mock responses without making real API calls:

• ASR returns: "This is a mock transcription of your audio."
• TTS returns: A simple 440Hz tone (1 second WAV file)

The project makes actual HTTP calls to Fish Audio API endpoints:

• ASR: POST {FISH_API_BASE}/v1/asr
• TTS: POST {FISH_API_BASE}/v1/tts

Authentication uses Bearer token from FISH_API_KEY.

• Python 3.8+
• Flask 3.0.0
• python-dotenv 1.0.0
• requests 2.31.0

The main.py script demonstrates how to access the JanitorAI completions API for CalHacks 2025:

• Endpoint: https://janitorai.com/hackathon/completions
• API Key: calhacks2047
• Model: OpenAI-compatible chat completions format

• The .env file contains API keys and should not be committed to version control
• Audio files should be in standard formats (WAV, MP3, etc.)
• TTS responses are returned as base64-encoded audio for easy JSON transmission
• The service runs on 0.0.0.0:5000 by default (configurable via environment variables)