Agent not retrieving field from Azure Search Service index

[jennguye]

Data source, Indexer, and Index was configured for Azure AI Search to index SharePoint. Fields in Index are a meta data and custom fields (mapped).
Agent in Foundry retrieves the information and response correctly but the URL to the document is empty. 'Knowledge' is set to use Azure AI Search.

the link routes to the playground and not the location where the document is in SharePoint.

{
name: "azure_ai_search"
context: {
trace_id: "thread_df3YPi2cPj1cNVwXxZUdnBQa"
span_id: "call_zbL6HxV6SVMfnfHdo9fvJ3bT"
thread_id: "thread_df3YPi2cPj1cNVwXxZUdnBQa"
}
kind: "Tool"
parent_id: "step_AqhiJZzJ8drHDFh3h9jVF6fb"
start_time: "2025-07-16T13:38:31.000Z"
end_time: "2025-07-16T13:38:32.000Z"
status: {
status_code: "Ok"
description: "completed"
}
attributes: {
span_type: "Tool"
id: "call_zbL6HxV6SVMfnfHdo9fvJ3bT"
type: "azure_ai_search"
azure_ai_search: {
input: ""password management standard""
output: "{'summary': 'Retrieved 5 documents.', 'metadata': {'urls': ['doc_0', 'doc_1', 'doc_2', 'doc_3', 'doc_4'], 'titles': ['doc_0', 'doc_1', 'doc_2', 'doc_3', 'doc_4'], 'get_urls': ['https://spo-searchservice.search.windows.net//indexes/spo-index/docs/Y2dpZmVkY29ycC5zaGFyZXBvaW50LmNvbSxmYmYzMDVkOC1lZW...
view all
"
}
output: "{'summary': 'Retrieved 5 documents.', 'metadata': {'urls': ['doc_0', 'doc_1', 'doc_2', 'doc_3', 'doc_4'], 'titles': ['doc_0', 'doc_1', 'doc_2', 'doc_3', 'doc_4'], 'get_urls': ['https://spo-searchservice.search.windows.net//indexes/spo-index/docs/Y2dpZmVkY29ycC5zaGFyZXBvaW50LmNvbSxmYmYzMDVkOC1lZW...
view all
"
}
}

Copilot is setup to retrieve the same data from the same index. Agent responded correctly with a working link to document.

Im thinking that the agent is not able to pull metadata and only custom fields/columns from SharePoint.

Any idea on why the agent isn't able to retrieve the links to the document? in the index, the field set to retrieve the url is 'meta_spo_item_weburi'.

[leestott]

Thanks for the comprehensive write up metadata_spo_item_weburi is designed to hold the SharePoint document URL, and when configured properly, it should allow agents (like Foundry or Copilot) to surface clickable links in responses.

Can you check the following:

Field Mapping in Foundry Agent Configuration

• Ensure the agent is explicitly mapping metadata_spo_item_weburi to the expected output field.
• If you're using a custom schema or transformation logic, confirm that the field isn’t being dropped or renamed during enrichment or post-processing.

Index Field Definition

• Double-check that metadata_spo_item_weburi is defined in the index with the correct type:

  {
    "name": "metadata_spo_item_weburi",
    "type": "Edm.String",
    "searchable": false,
    "filterable": false,
    "sortable": false,
    "facetable": false
  }

• If it’s missing or misconfigured, the agent may not be able to retrieve it even if the indexer populates it.

Skillset or Indexer Output

• If you're using a skillset, confirm that metadata_spo_item_weburi is passed through to the index.
• You can inspect the indexer’s output mapping to verify that the field is included in the final document structure.

Agent Response Template

• Some agents require explicit configuration to include metadata fields in the response.
• If the agent is using a templated response or citation format, make sure it references metadata_spo_item_weburi as the source for the link.

Playground vs. Production Behavior

• The playground may default to internal document IDs or preview links unless explicitly told to use metadata_spo_item_weburi.

Debugging Tips
Try querying the index directly via REST API or Azure Search Explorer and inspect the returned JSON. If metadata_spo_item_weburi is present and populated, the issue is likely in the agent’s response logic not the index itself.

Here is a quick diagnostic blueprint to resolve the meta_spo_item_weburi field, as Foundry agent isn't surfacing it in the final response.

Direct Index Query (via REST or Search Explorer)

Use the Azure Search REST API or Search Explorer to issue a manual query:

GET https://<search_service>.search.windows.net/indexes/<index_name>/docs?api-version=2023-07-01-Preview&search="password management standard"

Include these headers:

Content-Type: application/json
api-key: <your-admin-or-query-key>

Inspect the returned JSON

• Presence of meta_spo_item_weburi
• Whether it's populated and in plain string format
• Is it showing actual SharePoint URLs or placeholders?

Field Mapping in Foundry Agent

Review your Foundry agent’s blueprint/config (YAML or UI config, depending on how you’re deploying):

output_fields:
  - title
  - summary
  - meta_spo_item_weburi  # ← Ensure this is explicitly listed

If your agent uses response templating, confirm that the link field is referenced:

{{meta_spo_item_weburi}}

If using response citations or markdown generation, make sure the URL isn’t being dropped due to null handling or empty string filtering.

if you using Copilot Studio Compare Copilot vs. Foundry Retrieval Logic

Copilot likely uses something akin to:

document.url = metadata["meta_spo_item_weburi"]

Make sure Foundry isn’t relying solely on internal get_url fields, which route to Azure Search’s doc explorer (like https://spo-searchservice.search.windows.net/...) rather than SharePoint. If you're seeing those doc_0 style this could be the issue.

Optional Debug Extension

You could tweak your Foundry agent temporarily to log the retrieved document object pre-render:

print(json.dumps(doc, indent=2))

Just to confirm the field is present before the templating starts.

[jennguye]

Just want to provide an update on this and that I'm still going through your response trying to get it to work (Thank you for providing those details). I was able to get the URL to display (and clickable) in the chat playground in Foundry after recreating the data source for a deployment. the custom fields were not selected when the data source was created the first time.

URL is not displaying in the Agent Playground still. Working on this, will provide another update soon.

[jennguye]

From Index in Azure

{
"name": "metadata_spo_item_weburi",
"type": "Edm.String",
"searchable": false,
"filterable": false,
"retrievable": true,
"stored": true,
"sortable": false,
"facetable": false,
"key": false,
"synonymMaps": []
},

Indexer Output:

Data Source created for a deployment in Foundry - Chat Playground:

Agent Playground:

Skillset is not being used

Included 'metadata_spo_item_weburi' to the instructions:
You are an agent that retrieves information and documents from an index. You have access to the index named 'spo-index' and you can use Azure AI Search to retrieve document information. Provide link to the document metadata_spo_item_weburi, the name of the document, and a summary of the document.

• This did include a link but it's a link that is mentioned in the document. (Looks like we're half way there!)

1 - link to the AI Foundry playground..
2 - link to another internal resource page
3 - shows the name of the pdf but not clickable

No template is being used.

just this is setup: