AI Assistant

This AI Assistant can help with code debugging, auditing and quality, coding, writing documents, etc. It uses multiple LLM models to improve predictions and reduce hallucinations. This example uses Gemini 1.5 Flash and OpenAI GPT-4o, but the models can be switched. By using the wide context window of these models, we can reduce the use of RAG. Special attention has been paid to the fact that the user can modify the system and user prompts and use prompt templates.

The application is simple based on Python and Streamlit and can easily be extended for other use cases like code generation etc.

The usage follows the following steps:

1. Select the LLM Models and define their parameters

2. Check the system prompt or define your own custom prompt

3. You may define an User prompt using a prompt template

4. Now you chat with your documents. Gemini's answer

5. GPT-4o answer

Setup

Variables that need to be defined .env file (you may copy these from .env.sample)

GEMINI_PROJECT  =
GEMINI_LOCATION  =  
GEMINI_MODEL  =  "gemini-1.5-flash"
OPENAI_API_KEY  =
GPT_MODEL  =  "gpt-4o-2024-05-13"

Local environment

Create a new virtual environment

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Clone the repository

git clone https://github.com/MLConvexAI/AI-Assistant.git

You can run the code locally as

streamlit run app.py

The page can be found from

Local URL: http://localhost:8501

Cloud

The solution can also be easily deployed to the cloud using a container and a Dockerfile.