Power Platform Community Forum Thread Details

Hi everyone,

I'm currently working on a Copilot Studio agent that should use the DeepL Pro API to translate uploaded files. However, I'm new and running into several issues and would appreciate any guidance or best practices from the community.

What I'm trying to achieve:
Create an agent that can accept a file (e.g., a document), send it to DeepL for translation via their API, and return the translated file.

Issues I'm facing:

Tool-based approach:
- When I use the built-in tool to connect to DeepL, the agent doesn't seem to recognize the uploaded file properly.
Topic-based approach:
- If I use a topic to handle the file, it does recognize the file, but struggles with extracting or handling the file name correctly.
Using Power Automate Flow:
- I tried creating a flow to handle the translation, but:
  - The agent says the connector isn't properly configured.
  - When I call the flow from a topic, it expects a record instead of a simple file input, which to be frank i don't know what to do here.

Questions:

Has anyone successfully used the DeepL connectors in Copilot Studio or Power Automate?
How did you set them up?
What’s the best practice for handling file uploads and passing them to external APIs like DeepL?
Is there a recommended way to structure the flow so it works seamlessly with Copilot Studio?
Should I build my own Connector ?

Any help, examples, or documentation links would be greatly appreciated!

Thanks in advance!

Categories:

General topics

Hi,

For this scenario, I'd recommend using a Power Automate flow as the integration layer. Copilot Studio can capture the uploaded file, pass the file content and metadata to the flow, and then let the flow handle the DeepL API call, error handling, and return the translated file.
A few observations:

File inputs in Copilot Studio are typically passed as a record/object containing properties such as file name, content, and MIME type, which is why your flow is expecting a record rather than a simple string.
If the DeepL connector isn't recognized by the agent, verify that the connector connection is valid, shared correctly, and available in the same environment as the agent.
For file name issues, inspect the file record being passed from the topic and extract the filename property before calling the flow.
If the standard connector doesn't expose the functionality you need, a custom connector for DeepL can be a good option, especially for document translation endpoints and custom authentication requirements.

My recommended architecture would be: Copilot Studio → Agent Flow/Power Automate → DeepL API → Return translated file to the agent. This tends to be more reliable than calling the external API directly from the agent and gives you better control over file handling and error management.

Thanks & Regards,

Manoj Annavajjala