Veg_Grow+

Basis Query (Uses file retrieval to give custom advice + gardening tips based on user input)
Basic Query (Uses function calling and WeatherAPI)
Basic Query (Uses file retrieval to get relevant info based on user input)
Specific Query (Uses WeatherAPI and function calling to give custom care guide based on daily forecasted weather in given location)
Specific Query (Uses function calling and WeatherAPI to grab relevant info with little user input)

Inspiration

The main purpose of this project is to create a user-friendly AI for students to learn more about agriculture, cosmetology, and plant life. A lot of plants (especially in Florida) rot out because they are over-watered or out in the sun too long. Veg_Grow+ aims to remedy that. Students could use Veg_Grow+ to find the relationship between the current weather and plant growth. In areas like Florida where rainfall is frequent and the weather changes sporadically, students could use this AI to learn new gardening habits. Likewise, biology teachers could use this tool to better guide their students during experiments, such as when to water their plants (or do other care activities) depending on where they live and the weather forecast for that day.

What it does

Veg_Grow+ is a customized OpenAI Assitant that uses function calling and file retrieval to answer questions about different types of plants and how to care for them based on the weather in a given location. It can also provide gardening (harvesting, planting, etc.) tips for any given plant and tips on tools, outdoor wear, etc.

How we built it

I created an OpenAI assistant with specific instructions and used functional calling and file retrieval. The AI gets information from Purdue agricultural reports on various vegetable and fruit plants. I also gave the Veg_Grow+ Assistant access to a weather_api and created a get_weather function to get the daily forecast weather for any given location. The threads and runs for the AI are inside a handle_input function that runs through the Gradio Interface.

Challenges we ran into

This was my first time working with OpenAI and APIs. While I am proficient in Python, I am new to creating AI assistants and customizing their functionality. It was challenging to format my code to extract the right data from the OpenWeatherAPI and find files/databases that my AI could retrieve information from.

Accomplishments that we're proud of

I am most proud of how my AI responds to the user. I found regular AI bots to be less specific about plant care, and a lot of them gave inaccurate data about the weather.

What we learned

I learned that time management is one of the most important parts of a quick project. I spent most of the first day rewriting code, trying different ideas, and seeing what could plausibly come together in such a short time. The second day was spent making work as one.

What's next for Veg_Grow+

I plan to add more files to the vector store of my AI so it has more data to work with for user queries. I would also like to incorporate more functions to give specific advice based on other factors such as soil type (as this varies on location as well).