DEV Community: Stephen Chukwunonso

LazHelper: A Calmer Way to Understand and Organise Your Digital Workspace

Stephen Chukwunonso — Mon, 26 Oct 2026 06:59:13 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady

What I Built

LazHelper is an open-source, AI-assisted file organisation and workspace understanding application designed to reduce the time and effort people spend managing digital files.
Over time, our computers become crowded with documents, spreadsheets, presentations, research materials, source code, downloaded files, and multiple versions of the same document. Finding the right file or deciding where it belongs can become a frustrating, repetitive task.
I started building LazHelper to make that process easier.
Rather than simply moving files automatically, LazHelper helps users inspect their workspace, understand supported file formats, discover exact duplicates, search document contents, identify possible relationships between files, and review suggested organisation plans.
Its guiding workflow is:
Scan → Understand → Recommend → Review → Organise → Undo
The goal is to reduce repetitive work without taking control away from the user.
LazHelper includes workspace insights, content-aware search, a review queue for uncertain files, duplicate detection, reusable read-only workflows, and an optional AI integration for recommendations. It also includes safeguards designed to prevent accidental file changes.

Demo

Live application:https://laz-helper.onrender.com/

Code

GitHub repository:https://github.com/Stephen804-creator/LazHelper.git

How I Built It

I approached LazHelper as a practical software engineering project rather than a simple AI chatbot.
The first priority was building a reliable file-processing foundation. The application needs to inspect files safely, respect workspace boundaries, handle supported formats, and operate within configurable limits.
From there, I developed the organisation workflow around user approval. LazHelper can prepare suggestions, but a suggestion does not automatically grant permission to change a file.
I also worked on features that help users understand their workspaces, including duplicate and version insights, content-aware search, relationships between files, workspace health reports, and a lightweight knowledge map.
Another important part of the design is recovery. The application maintains an operation journal and supports undo for eligible moves when the original path remains available.
For AI-assisted recommendations, LazHelper is designed to integrate with Ollama and open-weight models. The deployment configuration being prepared targets Gemma 4 31B through Ollama Cloud, while local Ollama remains an option for suitable environments.
The application also uses PostgreSQL to persist application records rather than relying on SQLite as its runtime database.
Throughout development, I have focused on bounded file access, explicit approval, safe handling of destination conflicts, and avoiding automatic deletion.

Why Does Open Innovation Matter?

Open innovation matters because people should have more control over the tools they use to manage their digital lives.
An application that depends entirely on a closed AI service can limit how developers adapt its behaviour, change models, manage costs, and control where information is processed.
By building LazHelper around open-source software and supporting open-weight AI models, I can continue improving the application and explore different deployment options.
For example, users with a suitable local setup can use local Ollama inference. Other deployments can use a configured remote Ollama service. These options have different privacy and infrastructure implications: when remote AI is enabled, the relevant extracted content sent for a recommendation reaches that configured service.
This distinction matters. Local file processing and remote AI inference should not be treated as the same thing.
Open development also makes the implementation easier for other developers to inspect, test, improve, and extend. The longer-term goal is to create a useful tool that helps people work with their own files without surrendering control over file-changing actions.
For me, open innovation is not just about making source code available. It is about making software more adaptable, understandable, and respectful of the people who use it.

Prize Categories

Best Use of Render
LazHelper uses Render to deploy its web application, bringing together the FastAPI backend and browser interface in a single web service, with PostgreSQL for persistent application data.
Best Use of Gemma
LazHelper integrates Gemma 4 through Ollama Cloud to provide AI-assisted file recommendations, helping users understand their digital workspaces and prepare file-organisation suggestions for review.

Exprofit: Turning Everyday Transactions into Smarter Profit Decisions

Stephen Chukwunonso — Mon, 19 Oct 2026 06:58:05 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built
Every business records money coming in and going out. But recording transactions is only the beginning. The harder question is understanding what those numbers reveal: Is the business profitable? Where is money being lost? What is changing, and which decisions could improve the outcome?
I built Exprofit, an open-source financial intelligence platform designed to help small businesses, independent earners, retailers, and side-hustle operators answer those questions with greater clarity.
Exprofit brings financial tracking, receipt management, profitability analysis, and AI-assisted explanations into one workspace. Instead of treating financial records as entries to be stored and forgotten, it turns them into evidence that business owners can use to understand their financial position and investigate changes in performance.
The platform combines several practical capabilities:
Expense and income tracking: Record, organise, and review financial transactions in one place.
Receipt management: Upload receipts and review extracted information before saving it as a financial record.
Financial dashboard: Understand recorded income, expenses, cash movement, and financial trends across selected periods.
Profitability analysis: Examine revenue, recognised direct costs, gross profit, and gross margin when the necessary data is available.
MarginGuard: Investigate changes between financial periods, identify important findings, and examine the evidence behind them.
What-if simulator: Explore how changes in prices, sales volume, or unit costs could affect profitability without modifying actual financial records.
AI-assisted financial briefs: Turn verified financial findings into clearer explanations and practical next steps.
Workspace management: Keep personal financial activity separate from business workspaces, with collaboration and access controls for business teams.
What makes Exprofit different is its emphasis on evidence. Financial calculations are handled by application logic, while AI helps explain the results. The goal is not to replace financial judgement with a chatbot, but to give business owners a clearer starting point for making informed decisions.
For a small retailer, that could mean investigating rising expenses. For a freelancer, it could mean understanding whether revenue is translating into meaningful profit. For someone running a side hustle, it could mean testing a pricing decision before making a real-world change.
Exprofit turns financial record-keeping into a more useful process: record, understand, investigate, and decide.

Explore Exprofit here:https://exprofit.onrender.com
The application brings the financial dashboard, transaction workflows, receipt management, profitability insights, and financial intelligence tools together in one experience.

Exprofit is open source and available on GitHub.
Repository: https://github.com/Stephen804-creator/Exprofit.git
The project is released under the MIT License. Its codebase includes the frontend, backend API, database models and migrations, financial analysis logic, authentication, workspace functionality, and financial intelligence services.
I wanted the project to be more than a visual prototype. The underlying calculations, financial records, and application workflows matter just as much as the interface because users need to understand where their numbers come from.

How I Built It
Exprofit combines a modern web interface with a structured backend and a relational database.
The frontend uses React, TypeScript, and Vite to provide the dashboard and financial workflows. The backend uses Python and FastAPI to handle application requests, authentication, financial operations, and the financial intelligence layer. PostgreSQL provides persistent storage for financial records and related application data.
One of the most important design decisions was separating financial calculations from AI-generated explanations.
The application calculates supported financial measures using explicit logic and decimal-based arithmetic. MarginGuard compares financial periods and produces structured findings. The financial intelligence layer uses those results to prepare explanations, keeping the underlying calculations distinct from the language used to describe them.
Exprofit also integrates an opt-in path for Gemma through Hugging Face Inference Providers. This allows the application to use an open-weight model to explain structured financial findings when AI assistance is enabled. The model is configurable, so the AI provider can be changed without replacing the financial calculation engine.
The AI layer is deliberately constrained. It receives a bounded representation of calculated financial results rather than unrestricted access to the transaction system. Financial arithmetic remains under the application's control, and a deterministic explanation path provides useful results when AI assistance is unavailable.
The what-if simulator follows the same principle. Users can explore hypothetical changes to prices, sales volume, and direct costs without writing those assumptions into the real transaction ledger.
Authentication, workspace-scoped permissions, database migrations, and automated checks support the application's wider architecture.
The result is a working financial application in which the interface, business logic, database, and optional AI assistance serve one clear purpose: helping people understand their business finances more effectively.

Why Does Open Innovation Matter?
Small businesses should not need expensive financial infrastructure to begin understanding their numbers.
Open innovation makes it possible to build useful tools on accessible technologies, inspect how those tools work, adapt them to local needs, and improve them without depending entirely on a single proprietary platform.
For Exprofit, this matters in three ways.
First, financial intelligence should be understandable. By keeping calculations in explicit application logic and separating them from AI explanations, Exprofit makes its financial reasoning easier to inspect. Users can examine the underlying findings instead of being asked to trust an unexplained answer from a model.
Second, open-weight AI creates flexibility. Integrating Gemma through Hugging Face gives Exprofit a model option that can be configured independently of its core financial engine. This creates room to evaluate different models, control how they are used, and adapt the AI layer as requirements change.
Third, affordability matters to the people this product serves. A side-hustle operator or small retailer needs tools that help protect and improve already limited margins. An architecture that separates essential financial calculations from optional AI assistance makes it possible to preserve the core analytical workflow without making every financial operation dependent on a model request.
Open innovation is not simply about choosing an open model. It is about retaining the freedom to understand, adapt, and improve the technology behind a product.
With Exprofit, that freedom supports a practical mission: making financial insight more accessible to the people who need it to run and grow their businesses.

Prize Categories
Gemma: Exprofit integrates an opt-in Gemma-based financial explanation path through Hugging Face Inference Providers.
Render: Exprofit is deployed on Render, providing a live, publicly accessible application that brings its frontend and backend together under one URL. Render hosts the application while PostgreSQL provides persistent financial data storage. This deployment makes Exprofit accessible to users through a single entry point and demonstrates a practical approach to deploying a full-stack, open-source AI application.

How do you keep prompts, tools, and tests working when you switch models?

Stephen Chukwunonso — Fri, 16 Oct 2026 21:54:05 +0000

I've been thinking about what happens to an AI agent when the model underneath it changes. The code doesn't change, but the behavior can. A prompt that worked well on one model may give different answers on another, and a tool call that was reliable may start failing quietly.
How are you handling this today? Some questions I'm curious about:
–When you upgraded or switched models, what broke first?
–Did you have tests for agent behavior, or did you find out from users?
–Are your prompts, tool definitions, and test cases stored together, or in different places?
–Did you build any manual process to catch these changes, and what does it look like?
–What's the last model change that surprised you?
–How do you know when an agent has gotten worse?
–What do you do before you upgrade a model?
–What tool do you wish existed for this?
If you've solved this, I'd like to hear how. If you haven't, I'd like to hear that too. Share what you've run into in the comments.