Make your own private GPT with python

Although ChatGPT is incredible, its understanding is limited to the data used for training.

Wouldn’t it be amazing if we could interact with our own private documents using the power of Large Language Models (LLMs) without posting them online?

The good news is that we can get started right away! Let me demonstrate.

An open source project called privateGPT enables you to parse your own documents and communicate with them using LLM. The LLM will use your materials to produce answers to any queries you have.

(Note: To use the code, we must have Python3.10 or later installed. Earlier versions of python will not compile.)

1. Clone the repository

Navigate to folder of your choice and clone the repository https://github.com/imartinez/privateGPT.git . That will create a folder called “privateGPT”, so change the working directory (cd privateGPT).

git clone https://github.com/imartinez/privateGPT.git
cd privateGPT

2. Create and activate environment

Create virtual environment using the command (virtualenv venv) and activate the environment.

# creates virtual environment
virtualenv venv

# activates the environment

# for mac and linux
source venv/bin/activate

# for windows
call venv/scripts/activate

3. Install packages

Install all the required package listed in requirements.txt

pip install -r requirements.txt

Use pip freeze command to check if the package has been installed in the virtual environment.

  pip freeze

4. Download LLM model

Download the file ggml-gpt4all-j-v1.3-groovy.bin (3.5GB in size).

Then, create a subfolder called “models” and move the downloaded file to “models” folder.

5. Copy environment (example.env)

There is a file called “example.env” in the “privateGPT” folder. Copy that file into new file named “.env”

cp example.env .env

6. Add documents

Now add your private documents inside “source_documents” folder.

7. Ingest documents

Now run the command python ingest.py. It will parse your source documents and can take some few times to get completed.

python ingest.py

8. Interact with documents

Run the command python pivateGPT.py and start querying your documents. We can see a prompt with “Enter a query” once it is loaded and we are ready to ask question.

python pivateGPT.py