[i18n] Summary of 2026.1 Gazpacho PTG Discussions
Hello all, I would like to share with you a summary of the recent discussions held during the 2026.1 Gazpacho virtual PTG regarding the i18n SIG. We had several productive sessions focused on our two main goals for the upcoming cycles: the Weblate Migration, and a forward-looking discussion on AI Translation. Special thanks to the OpenDev Infra Team for the infrastructure support, the OpenSource Contribution Academy mentees in Korea who worked on the Weblate migration, and the KNU students who brought great energy and ideas to our AI discussion. Please refer to the full details from our Etherpad: https://etherpad.opendev.org/p/oct2025-ptg-i18n 1. Weblate Migration Status & Challenges A significant portion of our PTG was dedicated to the ongoing migration from Zanata to Weblate. - Migration Status: DaGyeong Kim presented the current status. 15 dashboard projects were initially migrated, but we identified issues that require a retry. A detailed status page is tracking all projects [1]. - Tooling Development: We are actively developing and refining automation scripts to manage the migration, including language creation/deletion and handling plural forms (patches [2], [3]). This work is being tracked in a blueprint [4]. - Key Challenge - Language Definitions: We are addressing mismatches between Zanata's language codes (e.g., ko_KR) and Weblate's CLDR standards (e.g., ko), especially concerning plural form definitions. - Verification of Migration Completion: It was discussed that a clear method needs to be defined to verify whether all translated PO files in Zanata have been successfully migrated to Weblate. In subsequent discussions, it was agreed to establish specific criteria for accurate verification and to proceed with the work based on those criteria. 2. Zuul Integration & Infrastructure The second major topic was integrating our new Weblate instance with the OpenDev (Zuul) CI/CD pipeline. - Core Patches: We reviewed the main patches for openstack-zuul-jobs [5] and project-config [6] that will enable Weblate to propose translation updates. - New Zuul Job Testing Strategy: It was discussed to create a completely new job to minimize the impact and to apply it to specific projects first, gradually expanding its scope over time. 3. AI Translation PoC We held an exciting session with students from Kyungpook National University (KNU) to discuss a Proof-of-Concept (PoC) for using AI in our translation workflow. - Goals: The discussion focused on two potential tools: - An automated, CPU-based LLM tool (e.g., using Ollama) to catch new strings from merged patches and provide initial translations. - A local AI-assisted tool for translators to help with the initial translation of large .po files. - Next Steps: We are exploring different instance flavors for performance and comparing the pros/cons of self-hosting a model versus using external APIs (similar to the Firefox translation model). 4. I18n Repository Packaging We are working to package the openstack i18n repository [7] and publish an updated package to PyPI. This initiative aims to enhance productivity by independently managing the tools related to i18n. Best Regards Seongsoo Cho [1] https://wiki.openstack.org/wiki/I18nTeam/Migration-to-weblate#List_of_projec... [2] https://review.opendev.org/c/openstack/i18n/+/961371 [3] https://review.opendev.org/c/openstack/i18n/+/963758 [4] https://blueprints.launchpad.net/openstack-i18n/+spec/weblate-migration [5] https://review.opendev.org/c/openstack/openstack-zuul-jobs/+/921878 [6] https://review.opendev.org/c/openstack/project-config/+/961499 [7] https://review.opendev.org/c/openstack/i18n/+/955037
On 2025-11-07 00:56:03 +0900 (+0900), Seongsoo Cho wrote: [...]
3. AI Translation PoC We held an exciting session with students from Kyungpook National University (KNU) to discuss a Proof-of-Concept (PoC) for using AI in our translation workflow.
- Goals: The discussion focused on two potential tools: - An automated, CPU-based LLM tool (e.g., using Ollama) to catch new strings from merged patches and provide initial translations. - A local AI-assisted tool for translators to help with the initial translation of large .po files. - Next Steps: We are exploring different instance flavors for performance and comparing the pros/cons of self-hosting a model versus using external APIs (similar to the Firefox translation model). [...]
On this topic, I recommend exercising caution. I know the situation is somewhat different, but an article posted this week about the impact of AI translation automation on the translators for the Mozilla community gives a good list of negative factors that we should strive to avoid when implementing any similar solutions: https://support.mozilla.org/en-US/forums/contributors/717446 -- Jeremy Stanley
Hi all, First, many thanks to Seongsoo for the great summary and active participation during the 2026.1 Gazpacho virtual PTG sessions. It was one of the most active and fruitful PTGs we’ve had in some time, and I deeply appreciate all the discussions and collaboration that took place. I’d also like to thank Jeremy for sharing the caution regarding AI translation usage - it’s an important reminder for us as we explore new workflows. Regarding the current direction of the I18n SIG with AI, our top priority remains ensuring that all translations integrated into the source repositories undergo proper human review before merging. We also need to improve documentation around this workflow so contributors and reviewers clearly understand each step. At the same time, the underlying issue lies in the PO file format itself, which currently provides no official flag to indicate whether a string has been machine-translated. While the format supports flag lines starting with #, today, an official extension introducing workflow flags with #= is already recommended for reading since June 2025, with a formal revision scheduled for after 2027-01-01 as described here: https://www.gnu.org/software/gettext/manual/html_node/PO-File-Format-Evoluti... . Finally, I want to express sincere thanks to everyone across the community who continues to make OpenStack globally accessible through translation and localization work. I also encourage more feedback and perspectives from our I18n language teams as we continue to refine our processes. Thank you all, /Ian On Fri, Nov 7, 2025 at 4:30 AM Jeremy Stanley <fungi@yuggoth.org> wrote:
On 2025-11-07 00:56:03 +0900 (+0900), Seongsoo Cho wrote: [...]
3. AI Translation PoC We held an exciting session with students from Kyungpook National University (KNU) to discuss a Proof-of-Concept (PoC) for using AI in our translation workflow.
- Goals: The discussion focused on two potential tools: - An automated, CPU-based LLM tool (e.g., using Ollama) to catch new strings from merged patches and provide initial translations. - A local AI-assisted tool for translators to help with the initial translation of large .po files. - Next Steps: We are exploring different instance flavors for performance and comparing the pros/cons of self-hosting a model versus using external APIs (similar to the Firefox translation model). [...]
On this topic, I recommend exercising caution. I know the situation is somewhat different, but an article posted this week about the impact of AI translation automation on the translators for the Mozilla community gives a good list of negative factors that we should strive to avoid when implementing any similar solutions:
https://support.mozilla.org/en-US/forums/contributors/717446
-- Jeremy Stanley _______________________________________________ OpenStack-I18n mailing list -- openstack-i18n@lists.openstack.org To unsubscribe send an email to openstack-i18n-leave@lists.openstack.org
participants (3)
-
Ian Y. Choi
-
Jeremy Stanley
-
Seongsoo Cho