"Can a generative AI automate tech contents translation?"
This repository tries to make it clear and tries to show you how much money we need to automate and how much sales we need to get from the generated contents in order to run the whole system sustainably. β»οΈβ¨
railsguides.jp has finally verified not-huge but a sustainable business model for documentation project in late 2022:
Now at least Rails documentation/translation project is confirmed and sponsored as a valued project by several economically-impacted companies like:
- SmartHR: T2D3-achived unicorn startup.
- Money Forward: A comprehensive financial management platform empowering users with tools for budgeting, expense tracking, and financial planning, all in one seamless digital ecosystem.
- Coincheck: A user-friendly cryptocurrency exchange platform offering secure trading, wallet services, and diverse digital asset options for both novice and experienced investors.
technically In 2023 and later, we will go into next advanced project, which tries to automate documentation process by using the obtained budget and LLM/Git/GitHub. The point would be how much we use LLM to automate in a cost-effective way because it costs. So this project, in other words, is going to clarify an appropriate balance and how much we can automate documentation project technically and sustainably.
β Sample files for PoC
Sample files below are fetched from 
yasslab/railsguides.jp, which is licensed under the CC BY-SA 4.0 International. They are used as an experiment for ChatGPT to detect the diff in english and possible match up with japanese translation.
- π AI Error Support (GPT-4 to support Rails Tutorial learners)
- yasulab/gpt-linter (GPT-4 to support Mitou Junior creators)
- yasslab/railsguides.jp
- yasulab/ipa2warp (Just make GPT-4 output codes to support IPA for missing-link problems)
- @yasulab's ranking at commits.top in Japan, most active GitHub users in Japan.