https://github.com/thangbuiq/booking-bot
Booking hotels recommendation chatbot on Vietnamese hotels using GraphRag
https://github.com/thangbuiq/booking-bot
graphrag louvain neo4j
Last synced: 7 months ago
JSON representation
Booking hotels recommendation chatbot on Vietnamese hotels using GraphRag
- Host: GitHub
- URL: https://github.com/thangbuiq/booking-bot
- Owner: thangbuiq
- License: apache-2.0
- Created: 2024-12-22T01:41:11.000Z (10 months ago)
- Default Branch: main
- Last Pushed: 2025-01-04T03:49:59.000Z (9 months ago)
- Last Synced: 2025-01-25T04:48:14.076Z (8 months ago)
- Topics: graphrag, louvain, neo4j
- Language: Python
- Homepage:
- Size: 9.29 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# DaLat Booking Recommendation Chatbot
   
- This project is about booking hotels recommendation chatbot on Vietnamese hotels using GraphRag.
## Usage
### Prerequisites
#### 1. Python packages
- Install package [uv](https://github.com/astral-sh/uv):
```bash
# On macOS and Linux.
curl -LsSf https://astral.sh/uv/install.sh | sh
# On Windows.
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
```
- Activate virtual environment:
```bash
# Create virtual environment.
uv venv --python 3.11
source .venv/bin/activate
```
- Install dependencies:
```bash
uv sync
```
#### 2. Services integration
- Install [Docker](https://docs.docker.com/get-docker/).
- Install `build-essential`:
```bash
sudo apt-get update && sudo apt-get install build-essential -y
```
- Create `.env` file:
```bash
cp .env.example .env
# Then, update the values.
```
> [!NOTE]
> `NEO4J_PASSWORD` must not be `neo4j`, then the Neo4j database will not start.
- Start services:
```bash
docker compose up -d
# or
make up
```
- Check the Neo4j database at `http://localhost:7474/browser/` or `http://localhost:7474/browser/preview/`. Login with your password and username specified in the `.env` file.
### Initialize Neo4j Graph Database
> [!IMPORTANT]
> You can quickly start the services and load the data into the Neo4j database with the following command:
>
> ```bash
> make full-load
> ```
#### Step by step
- The above services are running, then initialize the dimension and fact data with:
```bash
make transform
```
- Then we will have the graph data at these paths:
```bash
scraper/booking/output/warehouse/dim_hotel.parquet
scraper/booking/output/warehouse/dim_user.parquet
scraper/booking/output/warehouse/fact_review.parquet
```
- Load the data into Neo4j:
```bash
make load
```
- If you receive something like this, it means the data is loaded successfully:
```bash
Recommendations for user: d49aedca2af303f3439c9ddcfaa6c534
[{'hotel_id': 'a678288de41a8ae09740c785ce68cb19', 'name': 'Hotel Làng Kò Hơ', 'avg_rating': 10.0, 'review_count': 10}, {'hotel_id': 'b146903aee22801c928654c2e5ea5158', 'name': '296 Cetete - Home in Dalat', 'avg_rating': 10.0, 'review_count': 10}, {'hotel_id': '72229f4f874507f22c2e625ebc969dc8', 'name': 'Tuan Thuy Hotel', 'avg_rating': 10.0, 'review_count': 10}, {'hotel_id': '2a5e2e19df11a5afd54712afdff21502', 'name': 'Khách Sạn Đà Lạt View- Đà Lạt View Hotel', 'avg_rating': 10.0, 'review_count': 10}, {'hotel_id': '48eedf27f35cba1e5948c35be44ffb59', 'name': 'EURO HOUSE', 'avg_rating': 10.0, 'review_count': 10}]
Similar hotels to: d25d0d6a3f0f42b19482c3d6f16b7fbf
[{'hotel_id': 'b4af9e494c116dd2cdbdc3ea8e43a7d9', 'name': 'Đà Lạt Sunshine Garden', 'avg_rating': 10.0, 'review_count': 1, 'common_users': 1}, {'hotel_id': 'e522226e425ec0f82c0f77d389df2e21', 'name': 'NK.Dalat Hotel', 'avg_rating': 10.0, 'review_count': 1, 'common_users': 1}, {'hotel_id': '23bc46575562083fe796854a0f976987', 'name': 'Seven A Hotel', 'avg_rating': 10.0, 'review_count': 1, 'common_users': 1}, {'hotel_id': '7dc015e78f62b21a5616aba5d192aabc', 'name': 'Hostel-John Cafe and Beer', 'avg_rating': 9.0, 'review_count': 1, 'common_users': 1}, {'hotel_id': 'c0770a843f039342865a9dc11277cd2c', 'name': 'Dalat Terrasse Des Roses Villa', 'avg_rating': 9.0, 'review_count': 1, 'common_users': 1}]
Sample query:
[{'hotel_id': 'f8dd9fa66a227ffede2937d71ad05921', 'name': 'Dalat Wind Deluxe Hotel'}, {'hotel_id': 'fb401270b8babc2109a6dea17d1ebba2', 'name': 'Miền Nhiệt Đới 2 Hotel'}, {'hotel_id': '03adc0096493aa76d4e51842f8a15aa8', 'name': 'Raon Dalat'}, {'hotel_id': '9bb7114f0c3873e5398496dd7e140f5e', 'name': 'Royal Palace'}, {'hotel_id': '4a23f6c840d5b9f8d9f38fe289418a31', 'name': 'Jolie House'}]
```
- Go to `http://localhost:7474/browser/` or `http://localhost:7474/browser/preview/`, you can see some UI like this:
> Check the number of nodes and relationships in the UI to see if the data is loaded successfully.
- Then, you can query the data with Cypher query language:
> Example query just return Table of 5 hotels
```cypher
MATCH (h:Hotel)
RETURN h.hotel_id as hotel_id, h.name as name
LIMIT 5
```
### Recommendation
- Get the recommendations for a user query:
```bash
make recommend message="I am looking for a hotel with air conditioning and TV."
```