This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle
What I Built
Have you ever received a community event poster, a public notice, or a PDF and had to search through the entire document just to find the date, venue, registration details, or things you need to bring?
I wanted to explore a simple idea: What if AI could turn these everyday documents into something genuinely useful without sending our data to a cloud service?
That's the idea behind LocalLoop AI — Your Neighbourhood, Organised.
LocalLoop AI is a local-first AI assistant designed to help residents, student clubs, community volunteers, and small nonprofit organisations organise information from community notices and event flyers.
The goal is to let users import a document, extract its important details, generate a short summary, and turn the information into an actionable checklist.
Instead of another chatbot that answers general questions, I wanted to build something that takes an actual input and helps the user complete a practical task.
Who would use it?
Community volunteers managing event announcements.
Student clubs organising workshops and activities.
Local organisations handling public notices.
Everyday users who want to keep important information organised.
For these users, saving time and keeping information private can be valuable enough to make a focused tool worth paying for.
Code
Github url - https://github.com/Praveen-Raj-Thulasi/sidehustle
How I Built It
The main technical challenge is not just getting an LLM to summarise a document. It is making the extracted information useful, structured, and trustworthy.
The planned workflow is:
Document input: Import a community notice as an image or PDF.
Local OCR: Extract text using Tesseract OCR.
Local AI inference: Use Ollama with an open-weight language model to extract structured information.
Evidence validation: Associate extracted details with supporting text and flag missing or ambiguous information.
Human review: Let users correct the results before saving them.
Local storage: Store reviewed notices and their associated information using SQLite.
Actionable output: Generate concise summaries and preparation checklists.
Technology stack
Frontend: React and Vite
Backend: Node.js and Express
OCR: Tesseract OCR
Local inference: Ollama with Qwen3 4B
Database: SQLite
The initial application structure separates the client and server, allowing the frontend, document-processing pipeline, and local AI integration to be developed independently.
The focus is on getting one workflow right before adding unnecessary complexity.
Why Does Open Innovation Matter?
For this project, open innovation is about more than using a different model.
Community notices can contain personal contact details, registration information, and other information users may not want to upload to an external service.
A local-first approach gives users more control over where their documents are processed and stored.
Using open-weight models also makes it possible to experiment with different models, inspect their behaviour, and compare their performance without building the entire product around a single hosted AI API.
Another important part of this approach is being honest about uncertainty.
If a notice does not clearly mention a date or venue, the application should flag that information rather than invent an answer. The aim is to combine local inference with source evidence, validation, and human review.
Of course, local AI comes with trade-offs. Hardware requirements, inference speed, OCR quality, and model accuracy all matter. Those are part of what makes this project interesting to build.
What I Want to Explore Next:
My immediate priority is to complete the document-to-action workflow and test it with different kinds of notices, including unclear scans and incomplete information.
I'd also like to get feedback from a community volunteer or student club organiser to understand whether the tool actually saves time in a real situation.
The bigger idea is simple: AI doesn't always need to be a giant assistant that does everything. Sometimes, a small tool that solves one annoying problem well is more useful.
That's what I'm exploring with LocalLoop AI.
Top comments (0)