Genesys Cloud - Main

 View Only

Sign Up

Back to discussions
Expand all | Collapse all

Knowledge Fabric File Connector API - ingestionStatus: Failed after successful file upload

  • 1.  Knowledge Fabric File Connector API - ingestionStatus: Failed after successful file upload

    Posted 7 hours ago
    Hi community,
     
    I've been working on automating Knowledge Fabric ingestion via the File Connector API (FileUpload source type) and running into a consistent failure in the AI processing stage that I haven't been able to resolve despite thorough troubleshooting. Sharing here in case others have hit this or anyone from the product team can shed light.
     
    **Setup**
    - Region: US East (api.mypurecloud.com)
    - Source type: FileUpload, triggerType: Manual
    - Org has AI Experience tokens confirmed provisioned
     
    **What works fine**
    The first three steps of the sync flow complete without issues every time:
    1. POST /api/v2/knowledge/sources → source created, status Active
    2. POST .../synchronizations → synchronizationId returned, status InProgress
    3. POST .../synchronizations/{id}/uploads + PUT to presigned S3 URL → HTTP 200, ETag confirmed matching local MD5 checksum on every attempt
    4. PATCH .../synchronizations/{id} with {"status": "Completed"} → status moves to PostProcessing
     
    **Where it fails**
    After the PATCH, ingestionStatus consistently returns Failed. The Knowledge Activity panel in the UI shows the sync as Failed but the Sync Details summary reads "1 file synced successfully" - which suggests the failure is happening specifically in the post-transfer AI processing stage (chunking/embeddings), not in the file transfer itself.
     
    **Reproduced across three independent attempts:**
    - Attempt 1: .md file, 4,667 bytes, Content-Type: text/markdown
    - Attempt 2: .txt file, 356 bytes, Content-Type: text/plain (minimal content, three plain sentences)
    - Attempt 3: brand new source created from scratch + same minimal .txt file
     
    All three show the same pattern. No error message or failure reason is exposed in either the API response or the UI detail panel beyond the Failed badge.
     
    **What I've ruled out**
    - File integrity: ETag matches local MD5 in all three cases
    - Encoding: UTF-8 valid, no BOM
    - File format/content: both .md and .txt fail identically
    - The source object itself: same result with a fresh source
    - Token provisioning: confirmed with org admin
    - Missing steps: the four-step flow matches the documented FAQ exactly
     
    **Questions for the community / product team:**
    1. Is there a known issue with FileUpload ingestion in US East that could explain this?
    2. Is there any additional configuration or provisioning step required for FileUpload sources that isn't reflected in the public documentation?
    3. Is there a way to expose more detail about *why* ingestionStatus fails - even just an error code - from the API or the UI?
     
    Happy to share source IDs and synchronization IDs directly with the product team if that helps pull internal logs.
     
    Thanks!

    #AIConfiguration
    #API/Integrations

    ------------------------------
    Ricardo Solano
    ------------------------------


  • 2.  RE: Knowledge Fabric File Connector API - ingestionStatus: Failed after successful file upload

    Posted 7 hours ago

    Hi, @Ricardo Solano

    Based on what you described, it really sounds like the file transfer part is working and the failure is happening after the synchronization is marked as completed, during the Knowledge Fabric processing/indexing stage.

    Since you already tested different files, different formats, a fresh source, valid UTF-8, and confirmed the S3 upload/ETag, I would not keep troubleshooting only the upload step. At this point, I would check whether the uploaded document appears with any document-level status or metadata in the Knowledge Fabric APIs, if that capability is available in your org.

    I would also open a Customer Care case and include the source ID, synchronization ID, upload filename, timestamps, region, and any request/correlation IDs you captured. The "1 file synced successfully" message together with ingestionStatus = Failed is exactly the kind of scenario where backend logs will probably be needed, because the UI/API does not appear to expose enough failure detail.

    One thing I would also validate is whether all required Knowledge Fabric/AI permissions and subscriptions are active for the user or OAuth client running the API process, not only whether the org has AI Experience tokens provisioned.

    In short, your process seems aligned with the documented four-step flow. If the failure happens consistently only in PostProcessing with minimal valid files, I would treat it as either a provisioning/permission issue specific to Knowledge Fabric processing or something Genesys needs to investigate from the backend side.



    ------------------------------
    Arthur Pereira Reinoldes
    ------------------------------



Related Content