NLP Restaurant Service

This repository contains a restaurant reviews platform split into two services:

• reviews-api-service: Node.js + TypeScript REST API for users, restaurants, and reviews.
• sentiment-grpc-service: Python gRPC service that performs sentiment analysis with a Transformers model.

The API service sends review text to the gRPC service, stores reviews in PostgreSQL via Prisma, and maintains review sentiment aggregates/rankings in Redis.

Repository Structure

• reviews-api-service/: Express API, Prisma schema/migrations, Swagger docs, queue workers, tests.
• sentiment-grpc-service/: gRPC server and sentiment inference logic.
• shared-proto/: shared gRPC contract (sentiment.proto).

High-Level Flow

1. Client calls POST /reviews or PATCH /reviews/{id}.
2. API enqueues a BullMQ job and responds immediately with 202 and jobId.
3. BullMQ worker processes the job:

• creates/updates review in PostgreSQL
• calls gRPC NLP service for sentiment prediction
• persists sentiment and confidence
• updates Redis restaurant sentiment stats and ranking sorted set

Tech Stack

• API: Express, TypeScript, Zod, Prisma, BullMQ
• Data: PostgreSQL
• Cache/Queue backend: Redis
• NLP service: Python, gRPC, Hugging Face Transformers (distilbert-base-uncased-finetuned-sst-2-english)
• API docs: Swagger
• Tests: Jest

Prerequisites

Install and run these dependencies locally:

• Node.js 18+
• Python 3.10+
• PostgreSQL
• Redis

Environment Variables

reviews-api-service

Create a .env file in reviews-api-service/ with:

PORT=3000
DATABASE_URL=postgresql://postgres:postgres@localhost:5432/nlp_restaurant
JWT_SECRET=change_me

# Redis (used by BullMQ and sentiment stats)
REDIS_URL=redis://127.0.0.1:6379

# gRPC NLP service target
NLP_GRPC_HOST=127.0.0.1
NLP_GRPC_PORT=50051
NLP_GRPC_TARGET=127.0.0.1:50051
# Optional override to proto location
# NLP_GRPC_PROTO_PATH=/absolute/path/to/shared-proto/sentiment.proto

sentiment-grpc-service

Create a .env file in sentiment-grpc-service/ with:

NLP_GRPC_HOST=0.0.0.0
NLP_GRPC_PORT=50051

Setup and Run

1. Start infrastructure

Start PostgreSQL and Redis first.

2. Start the gRPC sentiment service

cd sentiment-grpc-service
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python main.py

The service listens on 0.0.0.0:50051 by default.

3. Start the API service

cd reviews-api-service
npm install
npm run prisma:generate
npm run prisma:migrate
npm run dev

API default URL: http://localhost:3000

Swagger UI: http://localhost:3000/api-docs

Useful Commands

From reviews-api-service/:

npm run dev
npm test
npm run test:coverage
npm run prisma:generate
npm run prisma:migrate
npm run prisma:studio

From sentiment-grpc-service/:

python main.py
python test_client.py "The food was amazing"

Redis Data Model

• Per-restaurant sentiment hash:

• key: restaurant:sentiment:{restaurantId}

• fields: positiveReviews, negativeReviews

• Ranking sorted set:

• key: restaurants:ranking:positive-reviews

• score: number of positive reviews

• member: restaurantId

Notes

• Review create/update API calls are asynchronous by design and return immediately after queueing.
• Worker startup is triggered in API bootstrap.
• If the NLP service is down, jobs will retry according to BullMQ retry settings.