Power Platform Community Forum Thread Details

I’m building a Copilot Studio agent in Teams where users upload DOCX/PDF files for analysis. AI Builder is proving limiting due to file size, speed, and lack of control over chunking/processing. Are there good alternatives like Azure Logic Apps, Azure AI Search (RAG), with files stored temporarily (e.g., SharePoint, Blob Storage).

My questions:

- What’s the recommended architecture for handling multiple large uploaded documents in Copilot Studio?

- Should document ingestion be fully externalized outside Power Platform?

Appreciate any guidance or reference architectures.

Categories:

General topics

Recommended Architecture for Large Document Handling in Copilot Studio
1. Externalize Document Ingestion and Processing:

Use Azure Blob Storage or SharePoint for temporary file storage.
Trigger an Azure Logic App or Power Automate flow when a user uploads a file in Teams.
The flow uploads the file to Blob/SharePoint and then calls an Azure Function or API for processing.

2. Use Azure AI Search (RAG) for Retrieval:

Index documents in Azure AI Search for scalable retrieval-augmented generation (RAG).
Your Copilot Studio agent queries Azure AI Search for relevant chunks/snippets.

3. Return Results to Copilot Studio:

Processed results or summaries are sent back to the agent via API or connector.

Should Document Ingestion Be Externalized?
Yes.
For large files and advanced processing, externalizing ingestion and chunking (outside Power Platform/AI Builder) gives you more control, scalability, and speed.

Reference Architecture Steps:

User uploads file in Teams → Power Automate/Logic App → Blob Storage/SharePoint.
Logic App/Azure Function processes file, chunks, and indexes in Azure AI Search.
Copilot Studio agent queries Azure AI Search for answers.
Results returned to Teams chat.