AI Provider - Azure Knowledge Base

Connects Drupal to Azure AI Search Knowledge Bases for agentic multi-source RAG, powering AI Chatbots and AI Agents with grounded answers and citations from search indexes, PDFs, and web sources.

The AI Provider - Azure Knowledge Base module bridges Drupal with Azure AI Search's Agentic Retrieval service. Instead of searching a single index and returning a list of links, an Azure Knowledge Base acts as a self-contained question-answering engine: it plans queries, searches across multiple enterprise repositories (your Drupal content, PDF documents, web pages, and SharePoint libraries), and uses an LLM to synthesize an accurate, grounded answer with clickable citations.

Features

Basic Functionality:

• Exposes Azure AI Search Knowledge Bases as native ChatInterface models inside the Drupal AI ecosystem.
• Powers user-facing chat blocks via the AI Chatbot (DeepChat) processor.
• Provides an executable tool for AI Agents and Assistants via Drupal AI Function Calling.

Unique Capabilities:

• Multi-Source Agentic Retrieval: Queries multiple distinct knowledge sources in a single request:
  • searchIndex: Drupal Search API indexes (via Search API Azure AI Search) or external indexes.
  • azureBlob: Unstructured documents (PDF, Word, Markdown, CSV) stored in Azure Blob containers.
  • web: Authoritative external documentation and web pages.
  • remoteSharePoint & indexedSharePoint: Microsoft 365 / SharePoint sites.
  • indexedOneLake: Microsoft Fabric / OneLake data lakes.
• Dual Connection Modes: Connect either by auto-detecting an existing Search API Azure AI Search server (search_api_aais), or connect directly with an Azure AI Search endpoint URL and API key from the Key module.
• Automatic Entity Citations: Resolves citations back to canonical Drupal URLs (e.g. /node/123) using Search API item IDs or title matching.
• Two-Way AI Prompt Library Sync: Manage Knowledge Base retrieval and answer instructions directly from Drupal's AI Prompt Library with push/pull synchronization.
• Runtime State API Toggle: Instantly enable or disable the chatbot per environment (e.g. dev, staging, prod) without modifying exported configuration.
• OData Filter Add-Ons: Filter search index results dynamically using standard OData filter expressions (e.g., status eq true).

Use Cases:

• Enterprise Knowledge Bots: Add an on-site chatbot that answers complex customer questions using both your Drupal site content and internal PDF reference manuals.
• Autonomous AI Agents: Give Drupal AI Agents a grounded retrieval tool so they can look up factual information before generating content or summarizing topics.

Post-Installation

1. Go to Administration > Configuration > AI > AI Providers > Azure Knowledge Base Provider (/admin/config/ai/providers/azure-kb).
2. Select your Connection Mode:
   • Search API Server: Select an existing Azure AI Search server configured via search_api_aais.
   • Direct Connection: Enter your Azure AI Search endpoint URL (e.g., https://my-service.search.windows.net) and select your API key configured in the Key module.
3. Select your default Knowledge Base and the Knowledge Sources you want to query.
4. Click Save & Test Retrieval to confirm connectivity with Azure.
5. (Optional) Place a chatbot: Navigate to Structure > Block layout, place an AI Chatbot (DeepChat) block in any region, and select Azure Knowledge Base (Agentic Retrieval) as the Chat Processor.

Additional Requirements

• Required Contributed Modules:
  • AI (drupal/ai): Core AI abstraction layer and prompt library.
  • Key (drupal/key): Secure storage for your Azure AI Search API key.
• Azure Service: An active Azure AI Search service with Knowledge Base preview capabilities (REST API version 2025-11-01-preview or later).

Recommended modules/libraries

• AI Chatbot (drupal/ai_chatbot): Bundled with the AI module; provides the DeepChat web component block UI.
• Search API Azure AI Search (drupal/search_api_aais): Allows reusing Search API server credentials and automatically linking citations to published Drupal entities.
• AI Assistant API (drupal/ai_assistant_api): Enables autonomous AI agents to call the Knowledge Base via Function Calling tools.

Similar projects

• Search API Azure AI Search: Focuses on indexing Drupal content into Azure AI Search for traditional keyword, vector, and hybrid facet search. AI Provider - Azure Knowledge Base complements it by adding agentic retrieval across non-Drupal sources (Blob storage, Web, SharePoint) and conversational answer synthesis.
• AI Search (drupal/ai_search): Provides RAG using locally vector-indexed Drupal content. AI Provider - Azure Knowledge Base offloads embedding generation, chunking, retrieval planning, and answer synthesis entirely to Azure's managed Knowledge Base infrastructure.
• Standard AI Providers (OpenAI, Anthropic): Generate answers using general model training weights. This module grounds all answers strictly in your curated organizational knowledge sources with verified reference citations.

Community Documentation

• Refer to the module's README.md for developer architecture and API examples.
• Microsoft Azure AI Search Knowledge Base Documentation
• Drupal AI Module Documentation