Skip to content

nju-websoft/CNPDS

Repository files navigation

CN-PDS

This is source code and prompts for paper "CN-PDS: China’s Public Data Search Enabled by LLMs".

Prompts

Prompts for metadata consolidation, metadata enrichment, dataset reranking, and relevance explanation are at ./prompts/.

Vue Frontend

./cnpds-frontend-vue, based on Vue3 and Tabler: An HTML Dashboard UI Kit built on Bootstrap.

Deployment

Docker (Nginx)

Deploy the frontend using nginx inside of a docker container.

Expose: Port 80. Env-Vars: VITE_BACKEND_HOST; VITE_BEIAN_NUMBER.

Image build command:

# PWD: ./ChinaOpenDataPortal-Frontend-Vue
docker build -f docker/Dockerfile -t username/imagename .

Push image to Docker Hub:

docker push username/imagename:latest

Use this command to create a basic env.custom.sh:

echo "PYTHON_PATH=$(realpath $(which python))" >> ./scripts/env.custom.sh

API Server & Backend

Java Backend

./cnpds-backend-java, based on Sprint Boot and Thymeleaf, with ability of acting as API server for other frontend service. Default index path is indices/current.

Usage

Use scripts/start-server.java.sh to start a process as API server and backend. Append more arguments as you need like ./scripts/start-server.java.sh --server.port=9998.

Environmet variables you may want to specify in env.custom.sh:

  1. MAVEN_PATH
  2. JAVA_PATH (Java 11 Recommended)
  3. ADMIN_USER
  4. ADMIN_PSWD
  5. PYTHON_RETRY_TIMES

Python Backend

./cnpds-backend-flask, based on Flask, providing ability of communicating with Large Language Models and querying MySQL database.

Usage

Use scripts/start-server.py.sh to start a process as python backend. Append more arguments as you need like ./scripts/start-server.py.sh --host 0.0.0.0.

Environmet variables you may want to specify in env.custom.sh:

  1. FLASK_PATH
  2. FLASK_PORT

Data Processor

  1. ./cnpds-metadata: Python scripts crawling metadata from each portals, multi-threads supported. Crawled metadata will be written to database for next step usage.
  2. ./cnpds-index-builder

Metadata

Use scripts/fetch-data.sh to start a process for metadata fetching. PS: There is one table that serves as an archive table for each metadata crawl and another table for index building.

Environmet variables you may want to specify in env.custom.sh:

  1. About Crawler Control:
    1. CRAWL_WORKERS
    2. CRAWL_FILES (whether or not to download datafiles)
  2. About database (Only MySQL Supported):
    1. DB_ADDR
    2. DB_PORT
    3. DB_USER
    4. DB_PSWD
    5. DATABASE_NAME
    6. REF_TABLE_NAME (specify a table as template)
    7. PRD_TABLE_NAME (specift a table which used in production)
  3. Others:
    1. PYTHON_PATH (Python 3.6 Recommended)

Lucene Index

After writing into database, the index builder will be started. If necessary, it will link the latest index to the path indices/current. If current index has been updated,the server will receive a POST request and refresh index path.

Environmet variables you may want to specify in env.custom.sh:

  1. BACKEND_URL

Appendix

Metadata Table Template

+-------------------+--------------+------+-----+---------+----------------+
| Field             | Type         | Null | Key | Default | Extra          |
+-------------------+--------------+------+-----+---------+----------------+
| dataset_id        | int          | NO   | PRI | NULL    | auto_increment |
| title             | varchar(255) | YES  |     | NULL    |                |
| description       | text         | YES  |     | NULL    |                |
| tags              | text         | YES  |     | NULL    |                |
| department        | varchar(255) | YES  |     | NULL    |                |
| category          | varchar(255) | YES  |     | NULL    |                |
| publish_time      | varchar(255) | YES  |     | NULL    |                |
| update_time       | varchar(255) | YES  |     | NULL    |                |
| is_open           | varchar(255) | YES  |     | NULL    |                |
| data_volume       | varchar(255) | YES  |     | NULL    |                |
| industry          | varchar(255) | YES  |     | NULL    |                |
| update_frequency  | varchar(255) | YES  |     | NULL    |                |
| telephone         | varchar(255) | YES  |     | NULL    |                |
| email             | varchar(255) | YES  |     | NULL    |                |
| data_formats      | varchar(255) | YES  |     | NULL    |                |
| url               | text         | YES  |     | NULL    |                |
| province          | varchar(255) | YES  |     | NULL    |                |
| city              | varchar(255) | YES  |     | NULL    |                |
| standard_industry | varchar(255) | YES  |     | NULL    |                |
+-------------------+--------------+------+-----+---------+----------------+

Metadata Table for Production Template

+-------------------+--------------+------+-----+---------+----------------+
| Field             | Type         | Null | Key | Default | Extra          |
+-------------------+--------------+------+-----+---------+----------------+
| dataset_id        | int          | NO   | PRI | NULL    | auto_increment |
| title             | varchar(255) | YES  |     | NULL    |                |
| description       | text         | YES  |     | NULL    |                |
| tags              | text         | YES  |     | NULL    |                |
| publish_time      | varchar(255) | YES  |     | NULL    |                |
| update_time       | varchar(255) | YES  |     | NULL    |                |
| industry          | varchar(255) | YES  |     | NULL    |                |
| department        | varchar(255) | YES  |     | NULL    |                |
| is_open           | varchar(255) | YES  |     | NULL    |                |
| data_volume       | varchar(255) | YES  |     | NULL    |                |
| update_frequency  | varchar(255) | YES  |     | NULL    |                |
| data_formats      | varchar(255) | YES  |     | NULL    |                |
| url               | text         | NO   |     | NULL    |                |
| province          | varchar(255) | YES  |     | NULL    |                |
| city              | varchar(255) | YES  |     | NULL    |                |
| standard_industry | varchar(255) | YES  |     | NULL    |                |
| url_hash          | varchar(255) | NO   | UNI | NULL    |                |
| origin_metadata   | mediumtext   | YES  |     | NULL    |                |
| fetch_time        | varchar(255) | YES  |     | NULL    |                |
+-------------------+--------------+------+-----+---------+----------------+

Enhanced Descripion Table Template

+----------------------+------+------+-----+---------+-------+
| Field                | Type | Null | Key | Default | Extra |
+----------------------+------+------+-----+---------+-------+
| dataset_id           | int  | YES  |     | NULL    |       |
| description_enhanced | text | YES  |     | NULL    |       |
+----------------------+------+------+-----+---------+-------+

Data File Path Table Template

+------------+------+------+-----+---------+-------+
| Field      | Type | Null | Key | Default | Extra |
+------------+------+------+-----+---------+-------+
| dataset_id | int  | NO   |     | NULL    |       |
| path       | text | YES  |     | NULL    |       |
+------------+------+------+-----+---------+-------+

Docker Compose YAML

version: '3.4'
services:
  frontend:
    image: 'username/imagename:latest'
    container_name: codp-frontend-vue
    restart: unless-stopped
    ports:
      - 'HOST_PORT:80'
    environment:
      - 'VITE_BACKEND_HOST=VITE_BACKEND_HOST_PLACEHOLDER'
      - 'VITE_BEIAN_NUMBER=VITE_BEIAN_NUMBER_PLACEHOLDER'

Params:

  1. Image Name.
  2. Host Port.
  3. Env-Var: VITE_BACKEND_HOST.
  4. Env-Var: VITE_BEIAN_NUMBER.

Script Usages

start-server

Java

Recommended:

nohup bash ./scripts/start-server.java.sh >> ./logs/server.java.txt 2>&1 &
Python

Recommended:

nohup bash ./scripts/start-server.py.sh >> ./logs/server.py.txt 2>&1 &

fetch-data

Recommended:

# the first of every month at 0:00
echo "0 0 1 */1 * __ROOT=\"`realpath .`\"; bash \${__ROOT}/scripts/fetch-data.sh > \"\${__ROOT}/logs/fd-\`date --i\`.txt\" 2>&1" >> logs/auto-task.txt
crontab logs/auto-task.txt