Anime Subtitle Chatbot

Few-shot chatbot trained on anime subtitles.

A flexible chatbot that generates contextual responses to anime dialogue inputs using few-shot prompting based on the Anime Subtitles Dataset. The project uses the OpenRouter API for access to multiple AI models.

This project was migrated from an older trax implementation and subsequently refactored from a local Hugging Face distilgpt2 model to use external AI APIs with the OpenRouter provider.

Provider

The chatbot uses OpenRouter as its sole AI provider, giving access to various models including DeepSeek through a unified API.

Setup

Clone the repository:

git clone https://github.com/waifuai/anime-subtitle-chatbot-trax
cd anime-subtitle-chatbot-trax

Create the uv virtual environment and ensure tooling (Windows venv shim path shown):

python -m uv venv .venv
.venv/Scripts/python.exe -m ensurepip
.venv/Scripts/python.exe -m pip install uv

Install dependencies:

.venv/Scripts/python.exe -m uv pip install -r requirements.txt
.venv/Scripts/python.exe -m uv pip install -e ./src/anime_chatbot[test]

Set up API Key:
- Preferred: set OPENROUTER_API_KEY in your environment.
- Fallback: create a single-line file at ~/.api-openrouter (Windows: C:\Users\YourUsername\.api-openrouter) containing your key.
- Get a key from: https://openrouter.ai/
Prepare Example Data (optional, improves style consistency):
- Place input.txt and output.txt under src/local_data/data/. Example files are included.

Usage

Interactive Mode

.venv/Scripts/python.exe -m python src/scripts/predict.py

Type quit or press Ctrl+D to exit.

Batch Mode

.venv/Scripts/python.exe -m python src/scripts/predict.py --input_file path/to/prompts.txt --output_file path/to/responses.txt

Custom Model

.venv/Scripts/python.exe -m python src/scripts/predict.py --model openrouter/free

Command Line Options

--input_file: text file with one prompt per line for batch processing
--output_file: optional output path (default: ./output_responses.txt)
--data_dir: optional few-shot data directory (default: src/local_data/data/)
--model: specific model to use (overrides default)

Project Structure

.
├── src/
│   ├── anime_chatbot/           # Core package
│   │   ├── __init__.py         # Package exports and convenience imports
│   │   ├── setup.py            # Package setup and metadata
│   │   ├── provider_selector.py # Provider selection and configuration logic
│   │   └── providers/          # AI provider implementations
│   │       └── openrouter.py   # OpenRouter provider
│   ├── local_data/             # Example data for few-shot prompting
│   │   └── data/
│   │       ├── input.txt       # Example dialogue inputs
│   │       ├── output.txt      # Example dialogue outputs
│   │       └── train.txt       # Training data (legacy)
│   ├── scripts/               # Executable scripts
│   │   └── predict.py         # Main prediction script with CLI
│   └── tests/                # Test suite
│       └── test_chatbot.py   # Comprehensive test coverage
├── .gitignore                 # Git ignore patterns
├── LICENSE                    # MIT-0 License
├── pytest.ini                # Pytest configuration
├── requirements.txt          # Python dependencies
└── README.md                 # This file

Models

Default Model

OpenRouter: openrouter/free

Custom Model Selection

Create ~/.model-openrouter file with custom OpenRouter model name
Use --model CLI parameter to override
Set environment variable OPENROUTER_API_KEY

Few-Shot Prompting

Uses examples from src/local_data/data/input.txt and output.txt
No fine-tuning or local training is performed
Examples improve style consistency and response quality

Development

Running Tests

# Install with test dependencies
.venv/Scripts/python.exe -m uv pip install -e ./src/anime_chatbot[test]

# Run all tests
.venv/Scripts/python.exe -m pytest

# Run with coverage
.venv/Scripts/python.exe -m pytest --cov=src/anime_chatbot --cov-report=html

Project Architecture

Provider Pattern: Clean separation between different AI services
Configuration Management: Environment variables and file-based config
Error Handling: Comprehensive error handling with user-friendly messages
Testing: Mock-based testing for reliable CI/CD
CLI Interface: Simple command-line interface with multiple options

src/anime_chatbot (1)

Files in src/anime_chatbot

Anime Chatbot Core Package