• SCHEMA
  • SOURCE
  • VECTOR
  • HUMAN

How to debug RAG retrieval pipelines with your AI

Connect your AI to Azure AI Search once. From then on, your agent can audit your index structures and data connections directly.

Ask AI about this page

Short answer

How can I use my AI to debug RAG retrieval?

Your AI pulls your index schemas and data source configurations to find mismatches in vector fields or broken data paths. You get a clear list of configuration errors to fix.

Audit outcomes

Where your findings land.

The AI categorizes its findings based on where the configuration error lives.

SCHEMA

Index structure errors

The AI identifies missing vector fields or incorrect dimensions. You get a report on how the index schema deviates from your RAG needs.

SOURCE

Data source mismatches

The AI checks if your data sources are correctly mapped. It flags any connection issues between your raw data and the search index.

VECTOR

Embedding field issues

The AI verifies that vector fields are properly configured for similarity searches. It highlights fields that lack the necessary vector properties.

HUMAN

Manual fixes required

The AI identifies specific settings that require you to manually update the Azure portal. It tells you exactly which parameters to change.

The workflow

What your AI does when the request arrives.

The AI moves through your Azure environment to pull the metadata needed for the audit.

  1. Inventory indexes

    The AI scans your environment to find all available search indexes. This ensures it is looking at the right targets for your RAG pipeline.

    list_indexes

  2. Inspect specific index

    The AI pulls the full schema for a specific index. It looks at field types, dimensions, and vector configurations.

    get_index

  3. Verify data sources

    The AI checks the underlying data sources to ensure they match the index requirements. It looks for connectivity and mapping errors.

    list_datasources

  4. Identify mismatches

    The AI compares the retrieved schema against your intended RAG setup. It flags any field that will cause retrieval to fail.

    get_index

Try it

Copy these to start.

Use these prompts to begin your debugging session.

Starting points

Replace the index names and data source names with your actual Azure resource identifiers.

Azure AI Search Connector

You're all set. Choose your MCP client and follow the setup instructions.

Connector linkhttps://edge.vinkius.com/vk_preview_UGnljcexY2MC0gUzqrUAvXNWXI0LuEzLwo7dKO1k/mcp

Claude Desktop

Follow the steps below to connect in seconds.

  1. 1In Claude Desktop, open Settings → Connectors.
  2. 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
  3. 3Click Add and start a new chat — Azure AI Search capabilities are ready to use.
Configuration · claude_desktop_config.jsonCopy
{
  "mcpServers": {
    "azure-ai-search-mcp": {
      "url": "https://edge.vinkius.com/vk_preview_UGnljcexY2MC0gUzqrUAvXNWXI0LuEzLwo7dKO1k/mcp"
    }
  }
}
Copy into chat04

  • List all my indexes and tell me which ones have vector fields enabled.

  • Get the schema for the 'products-index' and check if the vector dimensions match a 1536 embedding model.

  • List my data sources and verify if the 'blob-storage-source' is correctly linked to my current index.

  • Check the 'customer-docs' index schema for any missing fields that might break my RAG retrieval.

  • Claude
  • ChatGPT
  • Cursor
  • VS Code
  • Windsurf
  • Claude Code
  • JetBrains
  • Cline

Start here

Connect Azure AI Search once, then ask.

You only need to provide your credentials once. The connection is encrypted, and you can immediately start asking your AI to audit your search setup.

Connect Azure AI Search to your AI

FAQ

How this task behaves.

  • 01

    Can the AI fix my index schema automatically?

    No. The AI can only read and inspect your configuration. You must manually apply any changes in the Azure portal.

  • 02

    Can the AI delete my indexes?

    No. The tools provided only allow the AI to list and get information. It cannot delete or modify any data or structures.

  • 03

    Does the AI need admin permissions?

    The AI needs permissions to read the index and data source metadata. It does not need write or delete permissions.

  • 04

    Can the AI check my actual document content?

    No. The AI only inspects the index schema and the data source configurations, not the individual documents stored within them.

  • 05

    How does the AI know if my vector field is wrong?

    The AI compares the field properties it finds via the index schema against the specific vector requirements you describe in your prompt.

  • More questions about Azure AI Search? The Connector page answers them. See everything the Azure AI Search Connector can do

Connect Azure AI Search to Claude, Cursor, ChatGPT & more