由「你好，我想落單」到完成：認識 OrderSaathi

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

OrderSaathi: A Little AI Assistant for Small Businesses

What I Built

Running a small business means doing a lot of little things yourself: answering customer messages, figuring out exactly what someone wants to order, checking whether important details are missing, and writing replies while trying to manage everything else.

"OrderSaathi is my attempt to make that daily work a little easier."

It is a browser-based AI assistant designed for small shop owners, home bakers, Instagram sellers, resellers, and other small businesses that receive customer orders through chat messages.

Imagine a customer sends:

Hi, I need 2 chocolate cakes for tomorrow. One should be half kg and the other 1 kg. Please deliver in the evening.

Instead of reading the message several times and manually organising the details, the seller can paste it into OrderSaathi. The app uses AI to help turn the message into a more structured order summary, identify potentially missing information, and draft a reply for the customer.

The goal is simple: less time organising messages, fewer overlooked details, and smoother customer communication.

Who could pay for it?

OrderSaathi is designed for people who may not have the budget for a dedicated order-management system, especially:

• Home bakers and home-based food businesses.
• Instagram and WhatsApp sellers.
• Small retailers managing orders through messages.
• Freelancers and solo entrepreneurs who handle customer requests themselves.

A possible future business model would be a free basic version with an affordable monthly plan for features such as saved order history, reusable reply templates, and support for multiple business workflows. This is a potential monetisation plan, not an existing paid service.

Demo

Try OrderSaathi here:

The project is deployed on GitHub Pages. The browser-based AI model may need an internet connection to download the model the first time it runs.

Code

The complete source code is available on GitHub:

OrderSaathi — Small Business, Smarter Orders 🌿

Turn messy customer messages into business-ready order briefs using open-weight AI in your browser.

OrderSaathi is a privacy-conscious AI assistant designed for home bakers, tiffin-service providers, handmade-product sellers, freelancers, and other small business owners who receive customer orders through informal messages.

Instead of manually interpreting every message, sellers can paste a customer's request and generate a structured order brief, identify missing details, and draft a professional reply.

Built for the Side Hustle: Open-Source AI Challenge.

✨ The Problem

Small business owners often receive orders in informal English, Indian English, Hinglish, abbreviations, and incomplete messages.

For example:

Didi 2 chocolate cakes chahiye Sunday ko, one is 1 kg and dusra half kg, naam Aanya likhna hai, budget 1200 ke andar ho toh batao.

A seller needs to understand the quantities, cake sizes, requested name, delivery date, and budget while figuring out what still…

The project is built using three main application files:

• index.html — the application structure.
• style.css — the user interface and styling.
• script.js — the application logic and AI-powered message processing.

I kept the project lightweight so that it is easy to explore, understand, modify, and contribute to.

How I Built It

OrderSaathi uses 'Transformers.js and the open-weight Xenova/flan-t5-small model' to bring AI-powered text processing into the browser.

The application is built with HTML, CSS, and JavaScript. Instead of sending every customer message to a paid, hosted AI inference API, it uses browser-based inference through Transformers.js.

The basic workflow is:

1. The user selects a business category and enters a customer message.
2. The application passes the text to the AI-powered processing workflow.
3. The model helps generate a structured summary and a customer reply, including information about details that may need clarification.
4. The user reviews the result and can copy it or download an order brief.

The project is intentionally small. I wanted to explore how an open-weight model could support a specific, everyday business task without requiring a complicated backend or a paid inference service.

It is important to be clear about the current limitations: the model is small, generated details can be imperfect, and the output should be reviewed before it is used for a real order. The current version is a prototype, not a replacement for a complete order-management system.

Why Does Open Innovation Matter?

For a small business, the cost of an AI feature matters just as much as the feature itself.

A conventional approach would be to connect the application to a closed, hosted AI API. That can be useful, but it also introduces an external service dependency, usage-based costs, API-key management, and a new place where customer messages may be processed.

For OrderSaathi, I wanted to explore a different approach.

1. No per-message hosted inference bill

Because inference runs in the browser, the app does not need to pay a hosted AI inference provider for every message processed. That makes the idea more practical to experiment with when serving users who may have very limited budgets.

1. A better starting point for customer privacy

The app's browser-based inference approach means customer message text can be processed locally rather than being sent to a hosted inference API by the application. This is useful for messages that may contain names, addresses, delivery instructions, or other order details.

This does not automatically guarantee complete privacy: the website still loads its application assets and model files, and production use would need careful privacy and security testing.

1. More freedom to experiment

Using an open-weight model makes it possible to study the model, evaluate alternatives, and experiment with different models or workflows as the project grows, subject to their respective licences and technical requirements.

1. Lower barriers to building useful AI

A student or independent developer can experiment with AI-powered products without immediately needing a paid inference account, a dedicated GPU server, or a complex backend.

That is what open innovation means to me: making AI easier to experiment with, adapt, and build upon.

A closed API might provide stronger performance or more advanced capabilities in some situations. I am not claiming that a small browser model beats every hosted model. For this project, the advantage is the opportunity to explore a useful, affordable, and potentially privacy-conscious workflow with fewer infrastructure dependencies.

My First User

OrderSaathi is currently a prototype, and I am still looking for its first real-world users.

I do not have a verified paying customer or user-feedback story to share yet. My next step is to put it in front of a few home bakers and small online sellers, observe how they handle real order messages, and use their feedback to improve the workflow.

If you run a small business or know someone who does, I would love to hear whether a tool like this would genuinely save time—and what would make it useful enough to pay for.

Prize Categories

Overall Hacktoberfest Open-Source AI Challenge entry.

I am not claiming a partner prize category at this stage because I have not established that the current project meets a specific partner category's requirements.

What's Next?

I would like to explore:

• Better handling of Hinglish and informal customer messages.
• More reliable detection of missing order details.
• Custom reply templates for different business types.
• Optional order history and export features.
• Testing with real small-business owners to understand what they actually need.

I built OrderSaathi around a simple question: "Can a small, accessible AI tool take one repetitive task off a small business owner's plate?"
It is an early experiment, but I hope to turn it into something genuinely useful.

If you try it, I would appreciate your feedback. What would you change, and would you use it in your own small business?

devchallenge #hf26challenge