science-storyteller / docs /LOCAL_TESTING.md
tuhulab's picture
chore: Organize repository structure - move docs to docs/ and tests to tests/
28b3cfa
# Local Testing Guide
## Quick Test (Recommended)
Test just the audio generation:
```bash
# Activate virtual environment
source .venv/bin/activate
# Run quick test
python quick_test.py
```
This will generate a short audio sample to verify Kokoro-82M TTS works.
## Full Gradio App
To run the complete app locally:
```bash
# Activate virtual environment
source .venv/bin/activate
# Set environment variables (or create .env file)
export HUGGINGFACE_TOKEN=your_token_here
export ANTHROPIC_API_KEY=your_anthropic_key_here
export SEMANTIC_SCHOLAR_API=your_scholar_key_here # Optional
# Run the app
python app.py
```
Then open your browser to `http://localhost:7860`
## What to Test
1. **Audio Generation** - Run `quick_test.py` to verify Kokoro-82M works
2. **Paper Search** - Enter a topic like "AlphaFold" or "CRISPR"
3. **Full Workflow** - Generate a complete podcast from a research paper
4. **UI/UX** - Check that all tabs display correctly
5. **Error Handling** - Try invalid inputs to see error messages
## Expected Output
- **Audio files** saved to `assets/audio/`
- **WAV format** (Kokoro-82M native format)
- **File sizes** ~900KB-1MB per minute of audio
- **Quality** Much better than gTTS, similar to commercial TTS
## Troubleshooting
### "No module named 'huggingface_hub'"
```bash
pip install huggingface-hub
```
### "HUGGINGFACE_TOKEN not set"
Make sure you've exported the environment variable or created a `.env` file.
### "Port 7860 already in use"
```bash
GRADIO_SERVER_PORT=7861 python app.py
```
## Requirements
- Python 3.10+
- Internet connection (for API calls)
- Hugging Face token (free at https://huggingface.co/settings/tokens)
- Anthropic API key (for Claude AI)