Unable to Access PDF Content in Data Cloud Data Explorer Using Azure Blob Storage and UDLO

Rutuja Gurule 0 Reputation points
2026-10-23T10:59:15.8766667+00:00

Subject: Unable to Access PDF Content in Data Cloud Data Explorer Using Azure Blob Storage and UDLO

Product: Salesforce Data Cloud Issue Type: Unstructured Data Ingestion / Search Index / Retriever Configuration

Description

We are implementing an integration to ingest PDF files from Dropbox into Salesforce Data Cloud through Azure Blob Storage.

We have configured the Azure Blob Storage connection, created the Unstructured Data Lake Object (UDLO) and Unstructured Data Model Object (UDMO), and configured the search index.

We have also implemented Azure Event Grid notifications and an Azure Function to notify Data Cloud when files are created or deleted in Azure Blob Storage. The ingestion API returns accepted: true, indicating that Data Cloud has accepted the notification.

However, we are unable to verify that the actual PDF content has been processed and made available for retrieval through Agentforce.

The UDLO displays the PDF file metadata, and the search index status is Ready. However, we need clarification on how to access and validate the extracted PDF content and generated chunks in Data Cloud.

We found a similar issue discussed in the Salesforce Trailblazer Community: Not able to ingest the Unstructured (PDF) data from Azure.

Current Configuration

  1. Azure Blob Storage is configured as the source for PDF files.

The Azure connection and SAS authentication are configured.

A UDLO is created with the directory path and PDF file pattern (*.pdf).

The UDLO is mapped to a UDMO.

A search index is configured on the UDMO.

Azure Event Grid is configured to notify Data Cloud about file creation and deletion events.

An Azure Function sends notifications to the Data Cloud unstructured ingestion API.

The API returns accepted: true.

PDF file metadata is visible in the UDLO, and the search index status is Ready.

Issue

Although the PDF files are listed in the UDLO and the search index shows a Ready status, we need to confirm whether the actual PDF text has been extracted, processed into searchable chunks, and made available through the configured retriever.

We also need guidance on how to validate the generated chunks and retrieve the PDF content through Agentforce.

Assistance Required

Could you please help us clarify the following?

What is the expected processing workflow for extracting PDF content from Azure Blob Storage when using a UDLO and UDMO?

How can we verify that the PDF text has been extracted and that the corresponding chunk objects have been generated successfully?

Is it expected that unstructured PDF content will not appear in Data Explorer in the same way as structured CSV data?

If the search index status is Ready but the PDF content is not retrievable, what additional configurations or troubleshooting steps should we perform?

How should we configure and validate the search index and retriever so that Agentforce can retrieve relevant PDF content and generate answers based on the documents?

Are there any additional permissions, file notification requirements, ingestion steps, or processing delays that could prevent PDF content from becoming searchable?

Expected Outcome

We would like to confirm that PDF content is being processed correctly in Data Cloud and obtain the recommended configuration and troubleshooting steps required to retrieve relevant PDF passages through Agentforce.

Please provide guidance on the expected behavior of UDLO, UDMO, search index processing, generated chunks, and retriever configuration for unstructured PDF data stored in Azure Blob Storage.Case Subject: Unable to Access PDF Content in Data Cloud Data Explorer Using Azure Blob Storage and UDLO

Product: Salesforce Data Cloud
Issue Type: Unstructured Data Ingestion / Search Index / Retriever Configuration

Description

We are implementing an integration to ingest PDF files from Dropbox into Salesforce Data Cloud through Azure Blob Storage.

We have configured the Azure Blob Storage connection, created the Unstructured Data Lake Object (UDLO) and Unstructured Data Model Object (UDMO), and configured the search index.

We have also implemented Azure Event Grid notifications and an Azure Function to notify Data Cloud when files are created or deleted in Azure Blob Storage. The ingestion API returns accepted: true, indicating that Data Cloud has accepted the notification.

However, we are unable to verify that the actual PDF content has been processed and made available for retrieval through Agentforce.

The UDLO displays the PDF file metadata, and the search index status is Ready. However, we need clarification on how to access and validate the extracted PDF content and generated chunks in Data Cloud.

We found a similar issue discussed in the Salesforce Trailblazer Community: Not able to ingest the Unstructured (PDF) data from Azure.

Current Configuration

Azure Blob Storage is configured as the source for PDF files.

The Azure connection and SAS authentication are configured.

A UDLO is created with the directory path and PDF file pattern (*.pdf).

The UDLO is mapped to a UDMO.

A search index is configured on the UDMO.

Azure Event Grid is configured to notify Data Cloud about file creation and deletion events.

An Azure Function sends notifications to the Data Cloud unstructured ingestion API.

The API returns accepted: true.

PDF file metadata is visible in the UDLO, and the search index status is Ready.

Issue

Although the PDF files are listed in the UDLO and the search index shows a Ready status, we need to confirm whether the actual PDF text has been extracted, processed into searchable chunks, and made available through the configured retriever.

We also need guidance on how to validate the generated chunks and retrieve the PDF content through Agentforce.

Assistance Required

Could you please help us clarify the following?

What is the expected processing workflow for extracting PDF content from Azure Blob Storage when using a UDLO and UDMO?

How can we verify that the PDF text has been extracted and that the corresponding chunk objects have been generated successfully?

Is it expected that unstructured PDF content will not appear in Data Explorer in the same way as structured CSV data?

If the search index status is Ready but the PDF content is not retrievable, what additional configurations or troubleshooting steps should we perform?

How should we configure and validate the search index and retriever so that Agentforce can retrieve relevant PDF content and generate answers based on the documents?

Are there any additional permissions, file notification requirements, ingestion steps, or processing delays that could prevent PDF content from becoming searchable?

Expected Outcome

We would like to confirm that PDF content is being processed correctly in Data Cloud and obtain the recommended configuration and troubleshooting steps required to retrieve relevant PDF passages through Agentforce.

Please provide guidance on the expected behavior of UDLO, UDMO, search index processing, generated chunks, and retriever configuration for unstructured PDF data stored in Azure Blob Storage.

Azure Blob Storage
Azure Blob Storage

An Azure service that stores unstructured data in the cloud as blobs.

0 comments

1 answer

Sort by: Most helpful
Most helpful Newest Oldest

  1. AI answer

    2026-10-23T11:00:55.3066667+00:00

    Validate the PDF through a direct retriever test rather than expecting its extracted text to appear as structured rows in Data Explorer; the UDLO represents the source files, while the search index parses and chunks UDMO content for retrieval.

    Expected processing boundary

    The expected sequence is:

    1. Azure Blob Storage contains the PDF.
    2. Event Grid and the Azure Function notify Data Cloud of the file event.
    3. The UDLO references the unstructured file and maps it to a UDMO.
    4. Content harmonization processes the supported PDF content.
    5. The search index parses and chunks the UDMO content.
    6. A retriever searches those indexed chunks.
    7. Agentforce invokes that retriever and uses the returned passages.

    Salesforce supports PDF content harmonization from UDMOs ^1^, and documents the UDLO-to-UDMO model separately ^2^.

    Therefore, seeing the file’s metadata in the UDLO is not equivalent to seeing extracted PDF text. Data Explorer should not be treated as a CSV-style view of the PDF’s paragraphs or generated chunks.

    The available Salesforce documentation does not define accepted: true as proof that Data Cloud subsequently read, parsed, chunked, and indexed the file. Treat it only as evidence that the notification request was accepted for downstream handling.

    Validate indexing and retrieval

    Use one small, non-password-protected, text-based PDF containing a unique phrase such as UDLO-VALIDATION-2026-A7F9. This isolates parsing and retrieval from OCR, size, and ambiguous-query issues.

    1. Upload the PDF under the exact container, directory, and *.pdf path configured in the UDLO.
    2. Wait for its metadata to appear in the UDLO.
    3. Open the search index configuration and verify:
      • The indexed source is the intended UDMO.
      • Parsing and chunking are configured for the document content.
      • Any selected content/body field is searchable.
      • Processing details show no failure for the test document.
    4. Open the retriever associated with that index and use its available preview/test control to search for the unique phrase.
    5. A successful result must return a passage from the test PDF and a source identifier linking it to that document.
    6. Check retriever prefilters, result limits, and returned fields—especially SourceRecordId—if the document is indexed but omitted from results ^3^.
    7. Configure Agentforce to use that same retriever, then ask for the unique phrase without adding restrictive filters.

    Salesforce documents search-index creation, parsing/chunking, and creation of the retriever used by Agentforce ^4^.

    There is no documented, universally available Data Explorer path in the supplied product guidance for browsing generated chunks directly. If the org does not expose a chunk preview, validate chunk generation through retriever results and index processing/error details instead. A Ready index configuration alone should not be used as evidence that this particular PDF produced searchable chunks.

    Verify Azure can supply the file content

    Because the metadata is already visible, do not recreate the connection. Verify the existing SAS can perform both required operations:

    • List (l) on the container, needed to enumerate blobs ^5^.
    • Read (r) on the container/blob, needed to read blob content and metadata ^6^.
    • The SAS has not expired and its resource scope includes the configured container/path.
    • Storage firewall or private-endpoint rules permit the Salesforce connection to reach the blob endpoint.

    Prefer a user-delegation SAS where the Salesforce connector supports it; Microsoft recommends it over service/account SAS for stronger revocation and credential handling ^7^.

    Verify notification delivery separately

    In Azure Portal, open the storage account’s Event Grid subscription and check:

    1. Filters: subjectBeginsWith must match the container/folder, while subjectEndsWith can be .pdf. Blob event subjects use /blobServices/default/containers/<container>/blobs/<blob> ^8^.
    2. Metrics: compare matched, delivered, delivery-failed, dropped, and dead-lettered event counts.
    3. Dead-lettering: enable a dead-letter container if it is not configured. Event Grid otherwise can drop events after retry or TTL limits ^9^.
    4. Ensure BlobCreated is emitted only after the upload is committed, particularly if Dropbox integration uploads with multiple block operations.

    If Azure delivery, SAS read access, and the isolated retriever test all succeed but Agentforce still returns no passage, collect the test blob path, event ID/time, Function response, UDLO record, index processing details, retriever query/results, and Agentforce trace for Salesforce Support. That evidence identifies whether failure occurs during harmonization, chunk indexing, retrieval, or Agentforce invocation.

    References

    1. Unstructured Data File Formats and Connectors - Salesforce
    2. Unstructured Data in Data 360 | Salesforce Help
    3. Agentforce: Agent Unable to Retrieve or Return Expected Knowledge ...
    4. Step 3 - Create Search Index and Retriever | Ground Agentforce on Your ...
    5. List Blobs
    6. Create a service SAS
    7. Secure your Azure Blob Storage
    8. Reacting to Blob storage events
    9. Event Grid message delivery and retry
    AI-generated content may be incorrect. Read our transparency notes for more information.

    Was this answer helpful?

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.