A minimal web-based LLM Translator for single-user local deployment. Translate text using various LLM providers (OpenAI, Anthropic, Gemini, or local LLMs) with customizable tones and styles.
My goal was to build a fully finished app that's actually useful to me and just works - without writing a single line of code myself. I offloaded the whole thing to the LLM. Basically, I was testing the agentic capabilities of AI to see what this means for developers. And I want to emphasize this again: I didn't write one bit of code for this app. I spent 1 hour and 28 minutes on base implementation of this project - prompting agents, code review, manual testing, code-generation, etc.
This project is a proof-of-concept for fully AI-driven development. Since I didn't write the lines myself, treat this as experimental software.
It works for me, but it comes with no warranties. Always audit the code before using it for anything serious.
During the development of this project, I kept records of all the prompts I used with the agents, along with tracked timings for each major task. You’ll find all of these details in docs/HISTORY.md. If you’re interested in repeating this experiment or trying something similar, feel free to use this file as a reference.
- Multi-Provider Support: OpenAI, Anthropic, Gemini, and local LLMs
- Custom Tones: Define and manage custom translation styles (e.g., 'Official', 'Slang', 'Financial')
- Translation History: View and manage past translations
- Secure API Key Storage: API keys encrypted at rest using AES-256-GCM
- Settings Management: CRUD operations for tones and API key configuration
- Backend: Node.js v24, TypeScript, Express.js, SQLite with Prisma
- Frontend: React, Vite, TypeScript, Tailwind CSS
- Project Architecture
- Development Guide
- Dependency Management
- Coding Standards
- Troubleshooting
- Security
- Testing
See LICENSE for details.