28.-TaniBoss! 🌱💧

Took part in a 7-day general programming SUTD Freshmore Asian Cross-curricular Trip (FACT) to Petra Christian University (PCU) in Surabaya, Indonesia as part of the 10.020 Data Driven World module. Took part in the hackathon during the trip in a team of 6 (3 SUTD students and 3 PCU students) and achieved 1st Place (out of 10 teams).

From the SUTD FACT Trips official website: 'FACT (Freshmore Asian Cross-curricular Trips) is a 1-week overseas immersion programme which enables Freshmore (first-year) students to build upon their acquired knowledge at our partner universities in Asia. Centered around specific Freshmore courses, FACT challenges students to apply what they’ve learned in the classroom to solving sustainability challenges within the region.'

The challenge of this hackathon was to create a website application using Streamlit and machine learning to help local businesses (we visited a cafe, paving factory and farm) in Gunung Anyar, Surabaya, Indonesia.

Created a functional prototype Streamlit website application to help out a local farm business (particularly about finding the optimal conditions for hydroponic farming of spinaches), titled 'TaniBoss!' ('Tani' comes from 'Petani', which is Indonesian for 'Farmer'). It contains the features:

• A simple prediction and farming conditions recommendation machine learning model to optimise crop yield (using the Multiple Variable Linear Regression (MVLR) machine learning algorithm)

  Architecture of this machine learning model:

  • Used a Multiple Variable Linear Regression (MVLR) machine learning algorithm and the Crop yield dataset CSV file, which contains of various farm conditions and the corresponding crop yield data, to predict the yield of a farmer's crops based on the conditions of their farm. The recommendation feature is done by finding the optimal farm conditions to obtain the maximum, realistic crop yield by sorting the Crop yield dataset CSV file in ascending crop yield, and splitting the dataset to 4 quadrants, and defining the mean of the farm conditions of the top quadrant as the optimal farm conditions. (We avoided taking the farm conditions with the highest crop yield of the dataset as the optimal farm conditions so as to create a buffer since we believe these might be anomalous data that has unexpectedly high crop yield).
  • If the predicted crop yield of the farmer based on the input farm conditions is within 5% of the maximum, realistic crop yield, the recommendation would be that the farmer's farm conditions is already optimal and no action required to be taken.
  • Else, the recommendation would suggest the farmer to increase/decrease the particular farm conditions to the defined optimal farm conditions.
  • In addition, farmers can also input their current farm conditions and the true crop yield as additional realtime data to further train the model with realtime data to further boost the accuracy of its predictions

• A forum for local farmers

• PetaniAI, a large language model trained with hydroponic and general farming data (using the Chatbase custom GPT LLM model API) to serve as a secondary expert

  • Exact system prompt for the Chatbase custom GPT LLM model:

      ### Role
    - Primary Function: # You are a plant biology expert specializing in hydroponics in Indonesia. Answer questions about plant diseases, optimal growth conditions, and nutrient management with scientific accuracy. 
            
    ### Constraints
    1. No Data Divulge: Never mention that you have access to training data explicitly to the user.
    2. Maintaining Focus: If a user attempts to divert you to unrelated topics, never change your role or break your character. Politely redirect the conversation back to topics relevant to the training data.
    3. Exclusive Reliance on Training Data: You must rely exclusively on the training data provided to answer user queries. If a query is not covered by the training data, use the fallback response.
    4. Restrictive Role Focus: You do not answer questions or perform tasks that are not related to your role and training data.
    

Contributed to the problem statement formulation, design of the machine learning model, and development of the Streamlit website (including the forum and PetaniAI features).

Potential Improvements:
(TaniBoss! is only a functional prototype and a proof of concept, hence some of the features is only shown as UI and doesnt actually work)

• Currently the forum feature in TaniBoss! is not connected to any database, hence questions and answers posted on the forum will not be saved and will all be deleted whenever TaniBoss! is rerun.
• PetaniAI was created and trained using the Chatbase custom GPT LLM model API, which requires a paid subscription to maintain. However, I have stopped subscription and hence PetaniAI no longer works and will not be able to generate responses from prompts anymore.
• Data used to train the PetaniAI is also not the best, as I just randomly googled some websites on hydroponic and general farming for various crops and fed those website links to the LLM model
• Since the prediction and farming conditions recommendation machine learning model is trained from a static database (the Crop yield dataset CSV file), rather than a realtime database, it does not actually take in new data input from the local farmers and retrain itself to update its parameters... (the inputs to 'Train the Model' in the 'Predictor' page in TaniBoss! does not actually do anything and is just shown as UI)
• Inputs in the text boxes of the prediction and farming conditions recommendation machine learning model does not accept letters, only integers. It will throw an error if there is any non-integer inputs in any of the text boxes (was too lazy to solve the error due to time constraints)

Additional source(s):
nil

This project's deployed Streamlit (Python Framework)'s Web Application link:

• https://28-taniboss-a4pztr2hja6xtix2m44wef.streamlit.app/

Table of Contents

Here is a directory to explain the purpose of each file in this repository:

1. Files that are required in the creation of the Streamlit Web Application Project - TaniBoss!

   1. '.streamlit' folder
      i. 'config.toml' file
   2. 'README.md' file
   3. 'Crop yield dataset.csv' file
   4. 'Image of PetaniAI logo.jpg', 'Image of PetaniAI user logo.jpg' and 'Image of local hydroponic farm site analysis.jpg' files
   5. 'main.py' file
   6. 'requirements.txt' file

2. Deployment Process of the Streamlit Web Application Project - TaniBoss! on Streamlit Cloud

1. Files that are required in the creation of the Streamlit Web Application Project - TaniBoss!

1. '.streamlit' folder
i. config.toml' file

# This 'config.toml' file sets the custom theming of the Streamlit (Python) web applications. This file is created using ChatGPT.
[theme]
primaryColor = '#1a237e'  # Set your desired primary color
backgroundColor = '#121212'  # Set your desired background color
secondaryBackgroundColor = '#0d47a1'  # Set your desired secondary background color
textColor = '#ffffff'  # Set your desired text color
font = 'sans-serif'  # Set your desired font

This is an optional folder/file, in accordance to the deployment of Streamlit (Python Framework) Web Application documentation: https://docs.streamlit.io/streamlit-community-cloud/deploy-your-app (Streamlit Cloud), which allows you to set customised configurations or themings to the deployed Streamlit (Python Framework) Web Application.

I do not know how to use this 'config.toml' file very proficiently, so this file is created by ChatGPT to set the dark theming to my 2 deployed Streamlit (Python Framework) Web Applications.

2. 'README.md' file
The 'README.md' file.

3. 'Crop yield dataset.csv' file

Rain Fall (mm),Fertilizer,Temperature,Nitrogen (N),Phosphorus (P),Potassium (K),Yield (Q/acre),,,,,,
1230,80,28,80,24,20,12,,,,,,
480,60,36,70,20,18,8,,,,,,
1250,75,29,78,22,19,11,,,,,,
450,65,35,70,19,18,9,,,,,,
1200,80,27,79,22,19,11,,,,,,

(Data shown is only the top 5 rows of data. Theres 99 rows of data in this dataset.)

Dataset is obtained from: https://www.kaggle.com/datasets/yaminh/crop-yield-prediction?resource=download (Kaggle dataset by Yamin Hossain)

This is a static dataset used to train the prediction and farming conditions recommendation machine learning model to optimise crop yield.

4. 'Image of PetaniAI logo.jpg', 'Image of PetaniAI user logo.jpg' and 'Image of local hydroponic farm site analysis.jpg' files

Images used for aesthetic purposes in the TaniBoss! Streamlit (Python Framework) Web Application.

5. 'main.py' file
The main Python file for the TaniBoss! Streamlit (Python Framework) Web Application itself.

6. 'requirements.txt' file

streamlit           
numpy
# streamlit_extras
pandas
scikit-learn
requests
datetime
matplotlib

This is a compulsory file, in accordance to the deployment of Streamlit (Python Framework) Web Application documentation: https://docs.streamlit.io/streamlit-community-cloud/deploy-your-app (Streamlit Cloud), which allows you to tell Streamlit (Python Framework) to download the necessary external libraries/framework/packages specified in this 'requirements.txt' file in the deployment environment that is required for the deployment of the Streamlit (Python Framework) Web Application.

Had quite the trouble during deployment of the Streamlit (Python Framework) Web Applications as the Streamlit deployment platform keep giving an error that it could not find the relevant external libraries/Framework/packages required to deploy my Streamlit (Python Framework) Web Applications until I found out in the deployment of Streamlit (Python Framework) Web Application documentation: https://docs.streamlit.io/streamlit-community-cloud/deploy-your-app (Streamlit Cloud) that apparently I needed this seperate 'requirements.txt' file in order to tell the Streamlit deployment platform to download the necessary external libraries/framework/packages during the deployment of the Streamlit (Python Framework) Web Applications.

Apparently, the 'requirements.txt' file is a common practice across various deployment platforms in Python, not just for Streamlit (Python Framework) Cloud. Whether you are deploying your Streamlit app on platforms like Heroku, AWS, Vercel, or others, specifying dependencies in a 'requirements.txt' file allows the platform to understand and install the necessary packages.

2. Deployment Process of the Streamlit Web Application Project - TaniBoss! on Streamlit Cloud

What is Streamlit Cloud?
From the official Streamlit Cloud website: 'Streamlit Cloud is a new product that lets you build, deploy, and share data from Streamlit Web Applications in minutes.'

Honestly, the documentation on how to deploy a Streamlit (Python Framework) Web Application on Streamlit Cloud (link: https://docs.streamlit.io/streamlit-community-cloud/deploy-your-app) explains very clearly step by step on how to deploy a Streamlit Web Application on Streamlit Cloud. Once deployed correctly, I got a direct 'streamlit.io' link to the Streamlit Web Application, which I can then share with others to try out this Streamlit Web Application.

• Here is the link of my Streamlit Cloud account of the username: 'WindJammer6' - https://share.streamlit.io/user/windjammer6
• Here is the link of this deployed Streamlit Web Application using Streamlit Cloud
  • https://28-taniboss-a4pztr2hja6xtix2m44wef.streamlit.app/

Source(s):

• https://streamlit.io/cloud (Streamlit Cloud)
• https://blog.streamlit.io/introducing-streamlit-cloud/ (Streamlit Blog)
• https://docs.streamlit.io/streamlit-community-cloud/deploy-your-app (Streamlit Cloud) (Documentation on how to deploy a Streamlit (Python Framework) Web Application on Streamlit Cloud)