Let Eat Go AI Service

Let Eat Goのコミュニティ機能を支えるDistilBERTテキスト分類API

About

This service exposes a small FastAPI interface for profanity classification. At startup it loads a fine-tuned DistilBERT model from the local filesystem or downloads the model artifact from Amazon S3.

The Let Eat Go Backend API sends user-generated text to this service before processing community content.

Responsibilities

• Load a fine-tuned DistilBERT tokenizer and sequence-classification model
• Download and safely extract the model artifact from Amazon S3
• Classify text through a typed POST /predict endpoint
• Expose model readiness through GET /health
• Configure CORS, model paths, logging, and thresholds with environment variables

Architecture

flowchart TD
    API[NestJS API] -->|POST /predict| AI[FastAPI Service]
    AI --> Model[DistilBERT Model]
    S3[Amazon S3] -->|Startup download| Model

Loading

API

Predict

POST /predict
Content-Type: application/json

{
  "text": "Text to classify"
}

Response:

{
  "is_profanity": false,
  "confidence": 0.04
}

Health

GET /health

{
  "status": "healthy",
  "model_loaded": true
}

Getting Started

Prerequisites

• Python 3.9+
• A local fine-tuned model directory, or AWS access to the configured S3 object

Installation

git clone https://github.com/YJU-5/ai-service.git
cd ai-service
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
uvicorn main:app --reload --port 8000

FastAPI documentation is available at http://localhost:8000/docs.

Environment Variables

Variable	Required	Description
MODEL_DIR	No	Extracted model directory; defaults to ./profanity_filter_model
MODEL_ARCHIVE	No	Temporary archive path
S3_BUCKET_NAME	Without local model	S3 bucket containing the model
S3_MODEL_KEY	No	Model archive object key
AWS_DEFAULT_REGION	For S3	AWS region
AWS_ACCESS_KEY_ID	Local S3 only	Optional local AWS credential
AWS_SECRET_ACCESS_KEY	Local S3 only	Optional local AWS credential
CORS_ORIGINS	No	Comma-separated allowed origins
PROFANITY_THRESHOLD	No	Classification threshold; defaults to 0.5
LOG_LEVEL	No	Python logging level

Docker

docker build -t let-eat-go-ai .
docker run --rm -p 8000:8000 --env-file .env let-eat-go-ai

Use an IAM role in AWS environments instead of storing long-lived credentials in the image.

Tests

pip install -r requirements-dev.txt
pytest

The basic test suite validates configuration parsing and the health response without downloading the model.

License

This repository was created as an educational team project. No open-source license has been declared.