This is a project that integrates a chatbot with a MySQL database using Langchain and OpenAI's GPT-4-turbo model. The project is designed to interact with a database of projects, users, and tags, and respond with specific fields based on the user's prompt.
Recommendation system for projects and users: The recommendation system is an important feature of the project. Tailored recommendations are generated based on the interests, preferences and profile information of the users and the requirements of the projects. Use case - recommendation system:
The system uses an LLM-based model to understand the user's query. Based on this, the database is searched and suitable projects/users are returned. Example: A student asks the recommendation system for projects in the field of “data science” and receives suggestions for projects that cover this subject area. At the same time, the system can provide recommendations for professors who have similar interests or are supervised for relevant projects.
This project was developed as part of the InduKo Project, funded by Stiftung Innovation in der Hochschullehre. We also acknowledge the support from students, faculty and contributors who have been part of this collaborative effort.
For more iniformation about the InduKo research project, visit the official website: https://www.hs-heilbronn.de/en/projekt-induko-2cab68e84c21b797
For more iniformation about the IdeaLize project platform, visit the official website: https://www.hs-heilbronn.de/en/idealize-12f73ca0754864df
This project is licensed under Apache License, Version 2.0. Copyright 2024 Prof. Dr. Mahsa Fischer, Hochschule Heilbronn
This project includes several third-party libraries under open-source licenses. Key libraries include:
- FastAPI (MIT License)
- Pydantic (MIT License)
- Requests (Apache 2.0 License)
- PyMySQL (MIT License)
- Uvicorn (BSD-3-Clause License)
- Langchain (MIT License)
- OpenAI (Apache 2.0 License)
- You need the idealize_server project running for this project to work
- Python 3.10+
- MySQL or MariaDB installed locally
- MySQL client (
pymysql)
Create a file named apikey.py in the project root directory and add your OpenAI API key in it:
APIKEY = 'your-openai-api-key'
Create a file named bearer_token.py in the project root directory and add your bearer_token from your idealize_server in it:
TOKEN = 'your-bearer-token'
Run the following command to install the necessary Python dependencies:
pip install -r requirements.txt
Create a new MySQL database with your preferred name. Use the following SQL to create the necessary tables:
-- SQL Template to create tables for recsys database
CREATE DATABASE IF NOT EXISTS recsys;
USE recsys;
-- Table: chat_log
CREATE TABLE IF NOT EXISTS chat_log (
id INT(11) NOT NULL AUTO_INCREMENT,
prompt TEXT NOT NULL,
response TEXT NOT NULL,
timestamp DATETIME DEFAULT CURRENT_TIMESTAMP,
PRIMARY KEY (id)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_general_ci;
-- Table: Projects
CREATE TABLE IF NOT EXISTS Projects (
_id VARCHAR(255) NOT NULL,
title VARCHAR(255) NOT NULL,
description TEXT DEFAULT NULL,
tags LONGTEXT DEFAULT NULL,
owner_id VARCHAR(255) DEFAULT NULL,
isDraft TINYINT(1) DEFAULT NULL,
links LONGTEXT CHARACTER SET utf8mb4 COLLATE utf8mb4_bin DEFAULT NULL CHECK (JSON_VALID(links)),
attachments LONGTEXT CHARACTER SET utf8mb4 COLLATE utf8mb4_bin DEFAULT NULL CHECK (JSON_VALID(attachments)),
createdAt DATETIME DEFAULT NULL,
updatedAt DATETIME DEFAULT NULL,
PRIMARY KEY (_id)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_general_ci;
-- Table: Tags
CREATE TABLE IF NOT EXISTS Tags (
_id VARCHAR(255) NOT NULL,
name VARCHAR(255) NOT NULL,
type VARCHAR(50) DEFAULT NULL,
createdAt DATETIME DEFAULT NULL,
updatedAt DATETIME DEFAULT NULL,
PRIMARY KEY (_id)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_general_ci;
-- Table: Users
CREATE TABLE IF NOT EXISTS Users (
_id VARCHAR(255) NOT NULL,
firstName VARCHAR(100) DEFAULT NULL,
lastName VARCHAR(100) DEFAULT NULL,
email VARCHAR(255) NOT NULL,
username VARCHAR(100) NOT NULL,
status TINYINT(1) DEFAULT NULL,
userType VARCHAR(50) DEFAULT NULL,
interestedTags LONGTEXT CHARACTER SET utf8mb4 COLLATE utf8mb4_bin DEFAULT NULL CHECK (JSON_VALID(interestedTags)),
interestedCourses LONGTEXT CHARACTER SET utf8mb4 COLLATE utf8mb4_bin DEFAULT NULL CHECK (JSON_VALID(interestedCourses)),
studyPrograms LONGTEXT CHARACTER SET utf8mb4 COLLATE utf8mb4_bin DEFAULT NULL CHECK (JSON_VALID(studyPrograms)),
isBlockedByAdmin TINYINT(1) DEFAULT NULL,
createdAt DATETIME DEFAULT NULL,
updatedAt DATETIME DEFAULT NULL,
PRIMARY KEY (_id)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_general_ci;
To run the FastAPI server, use:
uvicorn main:app --reload
This will start the server at http://127.0.0.1:8000.
Ensure you are in the project directory (where the Dockerfile is located) and run the following command:
docker build -t idealize-recommendation .
Start the container with the following command:
docker run -d -p 8000:8000 --name idealize-recommendation-container idealize-recommendation
The project will run in the background (-d) and will be accessible at port 8000. You can adjust the port mapping if necessary.
Once the container is running, you can access the FastAPI application at http://localhost:8000.
Ensure that the OpenAI API key is correctly set. You can add the API key directly in the Dockerfile:
ENV OPENAI_API_KEY="your-openai_key"
If you're using an external MySQL database, make sure the container can access it. You can set up the MySQL connection environment variables inside the container or use a docker-compose.yml file to define a connected database service.
POST request to /api/chatbot with JSON body:
{ "message": "your question here" }
Example response:
{ "response": { "message": "Your response here", "projects": [ { "_id": "project_id", "title": "Project title", "createdAt": "2024-10-21 10:30:00" } ], "users": [ { "_id": "user_id", "firstName": "First", "lastName": "Last", "interestedTags": ["tag1", "tag2"] } ] } }