NPC irl

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built

NPC IRL — Your world is the map. Your AI gives you side quests.**

What if your boring neighborhood was actually an open-world RPG waiting to be explored?

NPC IRL is a real-world adventure game powered by locally running AI. Instead of spending more time staring at a screen, you use AI to generate short, weird, and fun missions that get you outside and interacting with the world around you.

Examples of quests:
-Detective Mode: Find something in your neighborhood that looks suspiciously out of place.

• Nature Explorer: Find three different leaf shapes.
• Alien Mode: Find an everyday object an alien would completely misunderstand.
• Chaos Mode: Find an object that looks like it knows a secret. Give it a name.

Complete quests, earn XP, level up, and turn an ordinary walk into a real-life adventure.

The goal is simple: make the screen the starting point of the adventure, not the destination.

Demo

https://npc-irl-dnra.onrender.com/

Code

https://github.com/rohit060321/npc-irl

The project uses a lightweight Python backend and a simple web frontend, making it easy to run and experiment with.

How I Built It

I built NPC IRL using a minimal stack to keep the project accessible and easy to modify.

Tech stack:

• Python: Backend server and AI integration.
• HTML, CSS, JavaScript: Interactive RPG-style interface.
• Ollama: Local inference runtime.
• Llama 3.2 3B: Open-weight language model for generating quests.
• LocalStorage: Saves XP and completed quest counts in the browser.

The backend sends a prompt to the locally running model, which generates a quest containing a title, mission, XP reward, and fun fact.

The frontend displays the mission and tracks progress as the player completes quests.

The architecture is intentionally simple: no paid AI API, no complicated agent framework, and no cloud database required.

Once the model has been downloaded, quest generation can work without an internet connection, provided the local Ollama service is running.

Why Does Open Innovation Matter?

For a project like NPC IRL, open innovation isn't just a technical choice. It's part of the experience.

1. AI without a cloud dependency

A locally running model means quests don't need to be generated by sending every request to a remote AI provider. After downloading the model, the core quest-generation experience can work without internet access.

2. Privacy by design

Your quest prompts can stay on your own machine instead of being sent to a third-party AI API. The application doesn't need a cloud database to track basic game progress.

3. Freedom to experiment

I can change the system prompt, experiment with different quest personalities, or swap in another compatible local model without rebuilding the entire application around a proprietary API.

Want a chaotic game master? A nature expert? A mysterious detective? The model and prompts can evolve with the project.

4. Lower running costs

There are no per-request AI API charges. The trade-off is that local inference uses your machine's processing power and memory.

For an application designed to encourage offline exploration, this is a meaningful advantage.

The bigger idea: Open AI makes it possible to build small, personal experiences that don't need to depend on a centralized service. Technology can help us reconnect with the physical world without demanding more of our attention.

My Agent Session

I haven't published a DevRelay session yet.

The project can be explored through the source code in the GitHub repository.

Prize Categories

Open-source AI / open-weight models / local inference.