https://github.com/shaadclt/ed-tech-company-qa-assistant
This project is a QA (Question-Answering) assistant designed for an Ed-Tech company. It leverages vector databases, state-of-the-art language models, and custom document embeddings to provide accurate responses to user queries based on a pre-defined knowledge base.
https://github.com/shaadclt/ed-tech-company-qa-assistant
embeddings langchain question-answering vector-database
Last synced: 6 months ago
JSON representation
This project is a QA (Question-Answering) assistant designed for an Ed-Tech company. It leverages vector databases, state-of-the-art language models, and custom document embeddings to provide accurate responses to user queries based on a pre-defined knowledge base.
- Host: GitHub
- URL: https://github.com/shaadclt/ed-tech-company-qa-assistant
- Owner: shaadclt
- License: mit
- Created: 2024-08-25T14:55:35.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2024-08-25T15:17:02.000Z (about 1 year ago)
- Last Synced: 2025-04-10T01:10:36.609Z (6 months ago)
- Topics: embeddings, langchain, question-answering, vector-database
- Language: Python
- Homepage:
- Size: 29.3 KB
- Stars: 3
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE.txt
Awesome Lists containing this project
README
# Ed-Tech Company QA Assistant
This project is a QA (Question-Answering) assistant designed for an Ed-Tech company. It leverages vector databases, state-of-the-art language models, and custom document embeddings to provide accurate responses to user queries based on a pre-defined knowledge base.
## Table of Contents
- [Features](#Features)
- [Installation](#Installation)
- [Usage](#Usage)
- [Project Structure](#Project-Structure)
- [Future Enhancements](#Future-Enhancements)
- [Contributing](#Contributing)
- [License](#License)
## Features
- **Vector Database**: Utilizes FAISS for efficient similarity search across the knowledge base.
- **Custom Embeddings**: Powered by HuggingFaceEmbeddings using the sentence-transformers/all-MiniLM-L6-v2 model.
- **Large Language Model (LLM)**: Integrates Groq's LLaMA3 model (llama3-8b-8192) for generating context-aware responses.
- **CSV Loader**: Loads FAQs or other structured data from CSV files to populate the knowledge base.
- **Streamlit Interface**: A simple and interactive UI for asking questions and receiving answers.
## Installation
1. Clone the Repository:
```python
git clone https://github.com/shaadclt/Ed-Tech-Company-QA-Assistant.git
cd Ed-Tech-Company-QA-Assistant
```
2. Set Up a Virtual Environment:
```python
python -m venv venv
source venv/bin/activate # On Windows, use `venv\Scripts\activate`
```
3. Install Dependencies:
```python
pip install -r requirements.txt
```
4. Environment Variables: Create a .env file in the project root and add your Groq API key:
```python
GROQ_API_KEY=your_groq_api_key_here
```
## Usage
1. Run the Streamlit App
```python
streamlit run main.py
```
2. Create the Knowledge Base:
To create the knowledge base from a CSV file, click the button Create Knowledgebase
3. Ask a Question:
Enter a question in the Streamlit interface, and the assistant will provide a contextually accurate response based on the loaded knowledge base.
## Project Structure
- **langchain_helper.py**: Contains the logic for loading data, creating the vector database, and generating responses.
- **main.py**: The Streamlit application file.
- **faqs.csv**: A sample CSV file used to populate the knowledge base.
- **requirements.txt**: Lists all required Python packages.
- **.env**: Stores environment variables like API keys (not included in the repository for security reasons).
## Future Enhancements
- Expand Knowledge Base: Add more documents or data sources.
- Fine-Tune Models: Adjust the embeddings or LLM for more accurate responses.
- Advanced UI: Improve the Streamlit interface with additional features like response filtering or feedback mechanisms.
## Contributing
Feel free to fork this repository, create a feature branch, and submit a pull request for any enhancements or bug fixes.
## License
This project is licensed under the MIT License. See the [License](/LICENSE.txt) file for details.