quick_pp

Python package to assist in providing quick-look/ preliminary petrophysical estimation.

Quick Start (Jupyter Notebook Examples)

1. Create virtual environment (tested working with Python3.11)

    python -m venv venv
   

2. Activate virtual environment

    > venv\Scripts\activate (Windows)
   
    > source venv/bin/activate (Linux)
   

3. Install requirements

    pip install -r requirements.txt
   

4. Launch the notebook and run the cells

   • 01_data_handler: create the MOCK qppp project file.

   • 02_EDA: quick look on the data

   • 03_*: quick petropohysical interpretation of the MOCK wells.

   • For API notebook, need to run the following before running the cells

       python main.py app
     

Install

To install, use the following command:

pip install quick_pp

To use qpp_assistant, you would need to;

1. Run git clone https://github.com/imranfadhil/quick_pp.git
2. Run pip install -r requirements.txt
3. Specify the required credentials in .env (based on .env copy file)
4. Run docker-compose up -d
5. Go to Langflow at http://localhost:7860 and build your flow.
6. Run python main.py app and go to the qpp Assistant at http://localhost:8888/qpp_assistant to test your flow.

CLI

To train an ML model, these are the requirements;

1. The input file in parquet format need to be available; /data/input/<data_hash>___.parquet

2. The parquet file need to have the input and target features as specified in MODELLING_CONFIG in config.py.

quick_pp train <model_config> <data_hash>

quick_pp train mock mock

To run the MLflow server

quick_pp mlflow-server

You can access the mlflow server at http://localhost:5015

To run prediction, the trained models need to be registered in MLflow first.

quick_pp predict <model_config> <data_hash>

quick_pp predict mock mock

To deploy the trained ML models

quick_pp model-deployment

You can access the deployed model Swagger UI at http://localhost:5555/docs

To start the App

quick_pp app

You can then access the Swagger UI at http://localhost:8888/docs and qpp_assistant at http://localhost:8888/qpp_assistant. You can enter any user name and password to login the qpp_assistant.

To use the mcp tools, you would need to first add the following SSE URLS through the interface; http://localhost:8888/mcp - quick_pp tools.

http://localhost:5555/mcp - quick_pp ML model prediction tools (need to run quick_pp model-deployment first).

Documentation

Documentation is available at: https://quick-pp.readthedocs.io/en/latest/index.html