This is a submission for the Hacktoberfest Open-Source AI Challenge — Week 2: Side Hustle.
The problem
Running a small online business means doing several jobs at once.
You are the product manager, marketer, sales analyst, inventory manager, and sometimes the entire customer support team. You need to write product descriptions, create social media content, understand which channels are performing, and decide what to do next.
The difficult part isn't always knowing what to do. It's finding the time to do everything.
I wanted to explore a simple question: What if a solo seller had an AI-powered workspace that could help with both creating marketing content and making better business decisions?
So I built HustleKit AI, an AI growth assistant for solo online sellers and small businesses.
Rather than building another generic chatbot, I wanted to connect AI to practical business workflows: product marketing, sales analysis, and structured growth recommendations.
Here it is in action
The video walks through the live dashboard, Product Studio, Sales Insights, Growth Advisor, and the underlying architecture.
What I Built
HustleKit AI brings several small-business workflows into one workspace.
Product Studio
A seller can provide product details and generate marketing copy using Gemma 3 12B through OpenRouter.
Instead of starting with a blank page, the seller gets AI-generated content that can help them prepare product descriptions, promotional copy, hashtags, and channel-specific marketing material.
The goal is to reduce the time between having a product and having something useful to publish about it.
Sales Insights
The dashboard brings sales metrics and channel performance into a more accessible view.
Instead of manually interpreting every number, a seller can review key metrics, examine sales trends, compare channels, and use the available forecast to think about upcoming performance.
The dashboard's displayed figures represent the application's demo workspace data; they should not be interpreted as independently verified results from a real business.
Growth Advisor
This is where HustleKit AI goes beyond generating text.
The Growth Advisor uses a multi-step Mastra workflow to load sales records from MongoDB, calculate relevant metrics, assemble supporting evidence, generate recommendations using the AI provider, validate the structured output with Zod, and persist the result.
The idea is to make recommendations more useful by grounding them in available business data instead of asking a chatbot for generic business advice.
One workspace, connected workflows
HustleKit AI combines these capabilities with a dashboard, inventory and product views, and settings.
The aim isn't to replace the seller's judgment. It's to give them a starting point for content creation and business decisions without having to switch between several disconnected tools.
Demo
Try the live application:
https://hustlekit-ai.onrender.com
The deployed application lets you explore the interface and its available workflows.
Code
GitHub repository:
https://github.com/Pranjal-debug/hustlekit-ai
The project was built with a React and TypeScript frontend, a Node.js and Express backend, MongoDB Atlas for persistence, Gemma 3 12B through OpenRouter, and Mastra for the growth-advisor workflow.
The repository contains the implementation and setup information for developers who want to explore the project or build on it.
How I Built It
1. A full-stack foundation
I used React 18, Vite, TypeScript, Tailwind CSS, and Lucide for the frontend, with Node.js, Express, and TypeScript on the backend.
MongoDB Atlas provides persistent storage for the application's business data.
Keeping the frontend and backend separate made it easier to develop the interface while preserving the existing API integrations and data behavior.
2. Integrating Gemma 3 12B
For the AI generation features, I integrated Google's Gemma 3 12B through OpenRouter.
During development, I had to account for the fact that model availability can change. The original model identifier I used, google/gemma-2-9b-it, was retired, so I moved to google/gemma-3-12b-it.
That was a useful reminder that integrating an AI model isn't just about making the first request succeed. The application also needs a working configuration, sensible error handling, and a way to verify that it is actually using the intended provider rather than silently falling back to a demo response.
3. Turning business data into recommendations with Mastra
For the Growth Advisor, I added a Mastra workflow named sales-growth-advisor.
The workflow follows a sequence:
- Load sales records from MongoDB.
- Calculate metrics and collect supporting evidence.
- Generate growth recommendations using the AI provider.
- Validate the structured response with Zod.
- Save the resulting recommendations to MongoDB.
This separation makes the workflow easier to inspect and test than a single large prompt that attempts to do everything at once.
It also creates clearer boundaries between business data, AI-generated output, validation, and persistence.
4. Making the interface feel like a real product
Once the core functionality was working, I focused on the UI.
The redesign uses a warm ivory background, black typography, and acid-lime accents, with an editorial commerce aesthetic. I wanted the interface to feel like a focused business workspace rather than a generic AI dashboard.
The redesign was implemented with an important constraint: preserve the working backend integrations and existing data behavior while improving how the application presents its features.
5. Testing the complete application
I didn't want to rely only on a successful frontend build.
The release-readiness process included production builds, automated backend tests, API health checks, and verification of the deployed application's main workflows.
The final audit reported passing production builds and tests, along with live checks for product generation, sales insights, the Growth Advisor, settings, and responsive navigation.
These checks helped me evaluate the application as a deployed product, not just as code that runs on my machine.
Why Does Open Innovation Matter?
I chose Gemma because open-weight models make it possible to build AI-powered applications around models whose weights are available for developers to use under their applicable terms.
For a small-business tool, that matters because AI should be something developers can experiment with, integrate into practical workflows, and adapt as their applications evolve.
HustleKit AI combines an open-weight model with an application stack that developers can inspect and extend. The project also demonstrates that useful AI integration involves more than model inference: it requires data handling, workflow orchestration, validation, persistence, testing, and a usable interface.
The model is one component of the system. The surrounding engineering is what turns it into a product.
What's next?
There is room to take HustleKit AI further with richer sales integrations, more personalized recommendations, and additional ways to evaluate whether suggested actions actually help a seller.
For now, I focused on building a working end-to-end foundation rather than claiming business outcomes that the project has not yet established.
Prize Categories
Best Use of Gemma: Gemma 3 12B powers product marketing content generation and AI-driven growth recommendations through OpenRouter.
Best Use of Open Source AI: HustleKit AI combines an open-weight model with a full-stack application, structured AI workflows, business-data processing, and output validation.
Best Use of Render: The application is deployed on Render, bringing the React frontend and Express API together in a publicly accessible, production-style application. The deployed workflow includes product generation, sales insights, and the Mastra-powered Growth Advisor, backed by MongoDB Atlas.
Built solo.
Built solo
HustleKit AI was built for the Hacktoberfest Open-Source AI Challenge, Week 2: Side Hustle.
If you're building a side project, running a small store, or experimenting with AI-powered business tools, I'd love to hear what you'd add to a workspace like this.
Try the demo, explore the code, and let me know what you think!
Live demo: https://hustlekit-ai.onrender.com
Source code: https://github.com/Pranjal-debug/hustlekit-ai
Top comments (1)
Nice product