This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle
What I Built
MenuMate AI is a local-first, AI-powered menu translation tool designed for independent restaurants, cafés, and small food businesses.
Restaurants welcome customers who speak different languages, but translating menus manually can be time-consuming and expensive. Generic translations can also miss the cultural meaning of regional dishes, cooking methods, and local food terminology.
MenuMate AI aims to solve this problem by helping restaurant owners create multilingual menus using AI running on their own computers.
With MenuMate AI, restaurant owners can:
- Add, edit, and manage menu items.
- Import existing menus from images and text-based PDFs.
- Extract menu text locally using OCR.
- Translate dish names and descriptions into multiple languages.
- Review and edit translations before publishing.
- Export menus as CSV files for further use.
The goal is simple: make multilingual menus accessible to small restaurants without requiring an expensive translation service or a recurring cloud AI API bill.
Who would pay for it?
Independent restaurant owners, cafés, food stalls, and local hospitality businesses that want to serve multilingual customers without adding complexity to their daily operations.
Demo
The current implementation is a locally running MVP.
- Application: Available locally after setup.
- AI processing: Powered by a locally running Ollama model.
- Live deployment: Not yet available.
I plan to demonstrate the complete workflow, from importing a menu to reviewing a translation and exporting the result.
Code
GitHub repository: MenuMate-Ai
The project uses a modular structure with a React frontend and a FastAPI backend, making it straightforward for contributors to improve OCR, translation quality, and menu formatting.
How I Built It
I built MenuMate AI using a local-first architecture.
Frontend
- React
- Vite
- Tailwind CSS
Backend
- Python
- FastAPI
- SQLAlchemy
- SQLite
Open-source AI and local processing
- Ollama for running an open-weight language model locally.
- Qwen2.5 3B as the initial model choice for menu translation.
- Tesseract OCR for extracting text from menu images.
- PyPDF for extracting selectable text from PDFs.
The application sends translation requests to the local Ollama service instead of a hosted AI API. Menu items and translations are stored in a local SQLite database.
The translation prompt instructs the model to preserve culturally specific dish names and avoid inventing ingredients, cooking methods, prices, or allergen information.
The workflow is designed to keep the restaurant owner in control: AI generates a draft, and the owner reviews it before using it.
This is an early MVP, so translation quality, particularly for regional languages, still needs systematic testing.
Why Does Open Innovation Matter?
For a small restaurant, every recurring expense matters. A translation tool that depends entirely on a paid AI API can become less attractive as the number of menu items and translation requests grows.
Open-weight models and local inference make a different approach possible.
1. Privacy by design
Menu content can contain business-specific descriptions, pricing, and proprietary recipes. Processing translation requests locally reduces the need to send that information to an external AI provider.
2. Lower ongoing AI costs
Once the model is downloaded, inference can run locally without a per-request charge from a hosted AI API. Hardware, electricity, maintenance, and model setup still have costs.
3. Freedom to experiment
An open model makes it possible to compare different models, adjust prompts, and eventually fine-tune for regional cuisines and language pairs.
4. Less vendor lock-in
The project can evolve without depending entirely on one provider's pricing, API availability, or service policies.
5. A more accessible business tool
My long-term goal is to make useful AI tools accessible to small businesses that may not have the budget or technical expertise to integrate commercial AI services.
Open innovation is not just about using free software. It is about having the freedom to inspect, adapt, and improve the technology around a real-world problem.
What's Next?
The next milestones for MenuMate AI are:
- Improve structured extraction of dish names, descriptions, and prices.
- Add an in-app translation editor.
- Generate printable, professionally designed multilingual menus.
- Benchmark translation quality and processing time across models.
- Test the application with independent restaurant owners.
- Explore a sustainable pricing model based on real customer feedback.
MenuMate AI is an early step toward making practical, privacy-conscious AI available to small businesses.
Top comments (0)