PyaGPT - Virtual Assistant of Instituto Piaget

🚀 The PyaGPT project is an innovative application that leverages various technologies to create a virtual assistant 🤖 specialized in providing information about Instituto Piaget. The user interface is developed using Streamlit 💻, while the backend is managed by FastAPI ⚡, ensuring efficient communication between the system's components. The Neo4j graph database 🗄️ stores all essential information about the institute, such as contacts and courses, as well as user-specific data, like personal information and class schedules 📝, allowing for contextual and personalized responses. Integration with the Ollama API enables the use of advanced language models 💡 to answer user queries more accurately and relevantly. PyaGPT is thus an interactive, easy-to-use, and efficient tool, ideal for facilitating access to institutional information about Piaget.

Architecture

The following diagram shows the architecture of the PyaGPT system:

Technologies

• User Interface 💻: Developed with Streamlit, providing an interactive and user-friendly interface.
• Backend ⚡: Managed by FastAPI, ensuring fast and efficient communication between system components.
• Database 🗄️: Uses the Neo4j graph database to store essential information about the institute, such as contacts and courses, as well as user-specific data, like personal information and class schedules. This allows the system to provide contextual and personalized responses.
• Language Models 🤖💡: Integrated with the Ollama API to use advanced language models, offering precise and relevant answers to user queries.

Demonstration

PyaGPT Chatbot (Llama 3.1 Model)

Interact with the PyaGPT chatbot to get contextual answers about Instituto Piaget:

PyaGPT Multimodal Analysis (Llava Model)

Image Analysis

Analyze images and receive descriptive output from the model:

PDF Analysis

Extract text or summarize content from PDFs:

How to Use PyaGPT

Clone the PyaGPT Repository

To get the source code of the PyaGPT project, execute the following command in your shell:

git clone https://github.com/gui-gaspar/PyaGPT

Install Ollama Models

The Ollama models must be installed beforehand to ensure PyaGPT works correctly. We recommend installing the following models:

• Llama 3.1: For general chat interactions.
• Llava: For image and PDF analysis.

Installation Commands:

To install the models, use the following commands:

ollama run llama3.1
ollama run llava

After installation, locate the directory where the models were installed on your system:

• Windows: C:\Users\USER\.ollama\models (replace "USER" with your username).
• Linux: /usr/share/ollama/.ollama/models

Copy the blobs and manifests folders to the project directory under ollama\models. This directory should be created at the root of the PyaGPT project if it does not exist. This step is essential for the models to be loaded correctly during execution.

Run the Project

To run PyaGPT, follow one of the procedures below:

1. In the main project directory, execute:

   docker-compose up --build

2. Alternatively, use the script located at the root of the project:

   run_docker.bat

Import Data into the Neo4j Database

The Neo4j data should be loaded using the SQL scripts available in the sql folder:

1. importar_dados.sql
2. importar_relacoes.sql

Simply copy their content into the Neo4j prompt to import all data related to Instituto Piaget and user-specific information.

Try It Out

The project is now ready to run! 🚀 Feel free to explore PyaGPT, test its features, and share any suggestions or report bugs. 📝 Your feedback is greatly appreciated and helps improve the project for everyone. 💡

Note

Response times from the models may vary depending on your computer’s performance and available resources. Some queries, especially those involving multimodal inputs like images or PDFs, may take a few extra seconds to process. Please be patient while PyaGPT generates the response.