Redesigning OpenStack I18n for the AI Era: Weblate & Zero-GPU AI
Talkat KubeCon + CloudNativeCon + OpenInfra Summit + PyTorch Conference China 2026 · 2026-09-07..09-09·
summary
Migrated decades of PO/POT data across 50+ languages into Weblate migration while preserving plural rules, header normalization and compileability using automated integrity checks and scripts; CI sync runs via Zuul to push translated PO files back to repositories. Built a zero-GPU CPU LLMs pipeline that runs open-source models on donated OpenDev VMs (up to 16 cores, ~16 GB RAM) to produce commit-time draft translations without fine-tuning, guided by language-specific prompts and a glossary, yields drafts that need human review and currently take ~30+ minutes to cover 40+ languages.
AI-written summary · about
chapters (5)
- 0:00OpenStack I18n redesign & AI
- 3:58Current translation pipeline & challenges
- 11:10Migration to Weblate
- 21:20CPU AI translation workflows
- 29:14Community collaboration & future steps
related talks
46mTackling Language Barriers in Open Source Docs: A Case Study of openEuler's...
48mContinuous Localization in Enterprise Web Projects
12mBuilt-in AI: how Mastodon uses translation APIs
04mGlobalization Pipeline: Translations at cloud speeds
18mAI translation + AI agents = i18n made easy
11mEngineering Real-Time Translation Platforms
37mEmpowering Developers with local AI
16mW2GPU: Toward WebAssembly-to-WebGPU Program Translation via Small Language Models