Hi folks! Not that I have anything against German, but when I use the “Rename to Chat suggestion,” it changes the filename to German. Can I set this to only give results in English?
You need to provide more information for us to assess.
- What version of DEVONthink?
- What OS?
- What AI provider and model?
- What is the language of the document you’re renaming?
Definitely!
What version of DEVONthink?
Devonthink 4.02
What OS?
Sequoia 15.6.1
What AI provider and model?
ChatGPT 4.1 Nano
What is the language of the document you’re renaming?
English
Neither the default prompt nor DEVONthink specify a language in this case. You could revise the prompt and explicitly mention the desired language.
I’m experiencing the same issue. In Settings > AI > Summarization I’ve set to English. Even added to the custom prompt “Any returned content shall be in English.” I don’t find any other place I can hint at this. Getting all kinds of returned languages when the source OCR content is clearly English.
These settings control only summarizing, e.g. via Edit > Summarize via Chat… or Tools > Summarize Documents via Chat….
The Rename to Chat suggestion batch processing configuration (see Tools > Batch Processing…) demonstrates how to use your own prompt instead of the placeholder which is used by the smart rule.
I will try that… thank you!
Is there a way to make a Batch Process run automatically on a condition like you can with a Smart Rule?
I have an incoming scans smart rule that automatically runs updating multiple items, but they are all almost guaranteed to be in a non-English language when results are returned, though the source documents are all English. Appears I need to rebuild all of this in Batch Processing with custom prompts simply to state “return results in English” but desire the automatic processing.
You could convert these documents to plain text to check the quality of the PDF’s text layer. Maybe there are lots of OCR issues or uncommon words?
Smart rules support the Chat - Query action of course too.
Batch processing is built for user-interaction, e.g., processing selected items, not unintended background handling.
