Contents hide

1 Building a Smart Indian Food Recommender Using TinyLlama, Ollama, and Python

1.3 Understanding the Code

1.4 Simulating the Weather

1.5 Mathematical Representation

1.6 Example Output

1.7 Why Use TinyLlama with Ollama?

1.8 Benefits of This Approach

1.9 Extending the Project

Building a Smart Indian Food Recommender Using TinyLlama, Ollama, and Python

In this tutorial, we’ll build a simple yet smart food recommender system that suggests Indian dishes based on the current humidity and temperature values. The system uses a local language model TinyLlama running via Ollama, and Python’s random.randint() function to simulate real-time weather.

Why Indian Food?

India has a rich and diverse culinary tradition. Different foods are consumed based on seasons, regional temperatures, and personal preferences. For example, during hot summers, people prefer light, hydrating dishes like buttermilk, lassi, or fruit chaat. In contrast, winter calls for spicier and warmer dishes like parathas, curries, and soups.

Setup Requirements

Ollama: Install from https://ollama.com
TinyLlama: Run ollama pull tinyllama
Python with basic packages like random

Understanding the Code

Here’s the core code we’ll use:

import ollama
import random

while True:
    humid = random.randint(30, 100)
    temp = random.randint(7, 45)
    
    print("Humidity:", humid)
    print("Temperature:", temp)
    
    response = ollama.generate(
        model="tinyllama",
        prompt=f"Humidity is {humid}% and temp is {temp}C. Suggest a list of Indian foods."
    )
    
    print(response["response"])

This Python script performs the following:

Generates a random temperature and humidity level
Prints the values
Uses TinyLlama to suggest relevant Indian food

Simulating the Weather

We simulate weather using the function:

random.randint(lower, upper)

For instance, temperature is selected randomly between 7°C and 45°C, while humidity ranges from 30% to 100%. These values represent the typical weather conditions found in various parts of India.

Mathematical Representation

Let us define two variables:

$ T $: Temperature in Celsius
$ H $: Humidity in %

We can define a composite index $ S $ as:

$$ S = \frac{T + H}{2} $$

Based on this score, we can categorize the climate and guide food suggestions:

If $ S > 70 $: Recommend cold, hydrating food
If $ 50 < S \leq 70 $: Recommend light vegetarian meals
If $ S \leq 50 $: Recommend hot, spicy, and heavier meals

This formula is optional in the code, as TinyLlama handles reasoning based on the natural language prompt.

Example Output

Let’s say our simulated values are:

Temperature: 39°C
Humidity: 90%

The prompt becomes:

"Humidity is 90% and temp is 39C. Suggest a list of Indian foods."

Sample output from TinyLlama might be:

During hot and humid weather, you can enjoy:

Buttermilk (Chaas)

Fresh coconut water

Fruit chaat

Khichdi with curd

Watermelon juice

Why Use TinyLlama with Ollama?

TinyLlama is a 1.1B parameter language model trained to run locally. It’s designed for small environments like laptops and edge devices. Ollama provides an easy way to run these models with a simple CLI and local API.

Benefits of This Approach

No Internet Required: Everything runs offline
Fast and Lightweight: Ideal for quick prototyping
Customizable: Prompt engineering makes it versatile

Extending the Project

Here are a few ways you can extend this script:

Integrate real-time weather data using an API like OpenWeatherMap
Build a web interface using Flask or Django
Add regional preferences based on Indian states
Use voice output with text-to-speech libraries

Video

Conclusion

Combining Python, TinyLlama, and Ollama gives you the power to build offline AI assistants with very little setup. This food recommendation project is just one of many possible applications. You can also adapt this logic for clothing suggestions, travel plans, or even smart home automation based on weather data.

As a closing note, here’s the simplified score formula again: