Filter Crowdin translations with AI
Turn an editorial question into a CroQL expression and check how language, approval and unresolved feedback affect the results.
Filter Crowdin translations with AI when a reviewer wants the French strings that have a draft translation and still need approval. A project-wide count of translations cannot answer that question. The croql skill helps an agent choose the right record type and make the language scope explicit before combining conditions.
A good filter has a counterexample. If a source message has an approved Spanish version and an unapproved French one, decide in advance whether it belongs in your result. That small example can expose an expression that accidentally accepts an approval in any language.
What it does
- Choose fields appropriate to source strings, translations, glossary terms or memory segments.
- Scope nested translation conditions to the relevant language.
- Group alternatives so an unrelated approval does not change the result.
- Explain a primary query and a narrower variation.
- Prepare the expression for the appropriate API parameter if a live check is authorised.
Install and use
- Give the agent the data context and a redacted example of a match. Running the resulting filter against a private project requires authorised Crowdin access.
- Ask, “Find translated French strings that still lack approval. Explain why an approved Spanish version should not remove a French item from the result.”
- Try the expression on familiar records and compare the matches with your expected list.
FAQ
Source
Made by Crowdin. Original skill: croql, published on GitHub under the MIT licence.
View on GitHub