Soul QA Bot: Smarter AI That Works Offline

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

This is an AI-Assistant and a smart one it can work with and without internet connection. It can solve problems, write simple code, summaries your task, give world wide information even in an offline mode. But it has one flaw because it can't give real time (live) information without internet.
I build this for my sister because she needs AI for studies but we can't buy the expensive subscription of AI models and there are too many students who needs AI for different work but they can't afford it so this is for them an AI which works with and without internet connection.

Demo

My youtube demo link - https://youtu.be/Ngq9Llm6L7E?si=4-5bmeA5ESNwQ54y

Code

My GitHub URL for this repo is https://github.com/dilesh596/soul-qa-bot-ai-assistant.

How I Built It

Soul QA Bot is architected specifically to balance low hardware requirements with robust AI capabilities. Here is a breakdown of the technical stack under the hood:

Open-Weight Model: I used Google's Gemma 3:1b as the core local brain. At only 1 billion parameters, it has an incredibly small memory footprint (~820MB), making it small enough to run smoothly on budget student laptops or older hardware without requiring an expensive, dedicated GPU.
Local Inference:** The offline mode is powered by Ollama. Ollama allows the system to run local LLMs efficiently on standard CPUs, serving the model locally via a lightweight API backend (localhost:11434).
Hybrid Routing Logic: The project utilizes custom application logic that detects the user's internet connectivity state. When connected, it routes requests to an online API for deep research; when disconnected, it seamlessly switches to the local Ollama backend to fetch answers instantly from the local gemma3:1b environment without interrupting the student's workflow.

Why Does Open Innovation Matter?

Open innovation is the absolute foundation of Soul QA Bot. Without it, this project simply could not exist. Here is why open-source technology makes all the difference compared to closed-source, premium APIs:

Breaking the Paywall Barrier: Closed APIs (like OpenAI or Anthropic) require ongoing token fees and active subscriptions. By utilizing open-weight models like Google's Gemma 3, open innovation allows me to provide quality AI capabilities completely free of charge to students who cannot afford monthly premiums.
True Offline Independence: Closed APIs strictly demand a constant, high-speed internet connection. Open-source innovation allows us to download the model directly onto a local machine. Once downloaded, the AI is completely decentralized—allowing a student to study in a rural area or an offline zone without spending a single rupee on data packs.
Community-Driven Optimization: Because the open-source community created tools like Ollama, complex models are compressed and optimized to run on standard, consumer-grade student laptops. Closed API companies focus on massive models that require server farms; open innovation focuses on making AI small, efficient, and accessible to everyone.

My Agent Session

Here is a look at how the Soul QA Bot behaves during a live user session, showcasing its ability to process prompts locally and handle hybrid state switches.
What Happens in a Session:

1. Connectivity Check: The agent initialises by verifying if the host machine has an active internet connection.
2. Context Retrieval: It prepares standard student-focused system prompts to guide the response style.
3. Execution: If online, it leverages external resources; if offline, it instantly routes the prompt to the local gemma3:1b instance managed via Ollama, completing the query entirely on-device.

Prize Categories

I am entering Soul QA Bot into the following official Hacktoberfest challenge categories:
Best Use of Gemma (Featured Category): This project is built natively around Google's open-weight Gemma 3:1b model. By optimizing the app to run a 1-billion parameter Gemma model locally via Ollama, it delivers zero-cost, high-performance AI directly on consumer-grade student hardware.
Overall Best Open-Source AI Project: Soul QA Bot introduces an impact-driven, hybrid online/offline orchestration layer explicitly designed to solve accessibility and financial barriers for students globally.

The Team

Dilesh Zingare — DEV Profile: https://dev.to/dilesh596
I am submitting this project as a solo developer, building this out to support student accessibility and open-source innovation!