A Generative AI based project deployed through Streamlit Framework.

Before proceeding to steps make sure you have: Anaconda Navigator, VS Code with C++ Redistributable as LLamaCPP utilised cpp compiler. If you have anaconda environment and vscode: If an error arises "Error while installing python package: llama-cpp-python" then, try to install c++ as an extension in vscode and try to run the code OR Reinstall VS Code with c++ build tools and other necessary tools, here's a reference:

Just in case you don't have VS installer, you can find it with Download Build Tools from visualstudio.microsoft.com/visual-cpp-build-tools.

Then, use the following command: !pip install llama-cpp-python

Additionally, create a separate folder named model and download the llm model (Quantized) from huggingface - sqlcoder Link - https://huggingface.co/TheBloke/sqlcoder2-GGUF/blob/main/sqlcoder2.Q4_K_M.gguf

Then change the path in "model_path" variable which is present inside the function.

Create a new environment after anaconda installation - conda create -n <your_env_name> and also activate the environment - conda activate <your_env_name> then, Use the command:"pip install -r requirements" to install all necessary dependencies.

The commands are to be run in anaconda prompt after creating and activating the environment

This is an Generative AI based Application which is deployed through Streamlit Framework, based on the prompt we define answers are seen accordingly. This is a "Prompt" based chatbot or Prompt Engineering based application utilised to convert a question in the form of query to a SQl (Structured Query Language) which is used to interact with the metabases and other databases. This is a local based application where we do not require any APIs or additional information to access or share our information thus putting a privacy of our personal information.

According to the above diagram, gpt-4 a closed Model requires API key and other details to access it whereas our selected model performance is about as good as the predecessor of GPT-4 and since it is open-sourced we can use it as a local model for any text generation related tasks.