Reuse in Copilot Studio

[cforce]

Currently, the azure-search-openai-demo provides a web UI and backend for Retrieval-Augmented Generation (RAG) with Azure Cognitive Search and Azure OpenAI. While this is excellent for demos and standalone use, it lacks an easily consumable REST API that external tools or agents can call directly.

Many organizations want to integrate the demo’s functionality (chat + RAG over indexed documents) into their own environments, such as Microsoft Copilot Studio, Teams apps, or other agent frameworks.

Copilot Studio supports Tools (REST API connectors, custom connectors, etc.) that can call external APIs. If the demo exposed a clean REST API (e.g., /chat endpoint that accepts a query and returns an answer with citations), it would be easy to plug it into Copilot Studio with minimal effort.

This avoids duplicating the RAG logic in multiple places and makes the project more useful as a service backend, not just a sample frontend.

Feature Request:

What is the advice on how Microsoft Copilot Studio supports adding this REST APIs as tools, so this feature would immediately enable integrations into Copilot agents.

Add an optional REST API layer to the project that exposes the core functionality:
Input: user query text, conversation ID (optional), and parameters (temperature, top_k, etc.).
Output: model-generated answer, source documents used, conversation context.
Ensure the API is secured (e.g., Azure AD auth, API keys).
Provide sample requests (cURL/Postman) and documentation for integration.
Optionally, add an example showing how to connect this API to Copilot Studio via a REST API tool.

Advantage
Enables low-maintenance integration of the sample with Microsoft Copilot Studio and other agent frameworks.
Allows enterprises to re-use the project as a backend service without modifying its UI.
Broadens adoption of the sample beyond demos, making it practical for production integrations.

Directly connecting Copilot Studio to Azure AI Search and Azure OpenAI (already possible). However, i prefer the ready-made orchestration and response formatting that this demo provides. Also access rights enformcent build on top of categories and oids.
Maintaining a fork of the repo with custom APIs (higher maintenance burden).