I’m trying out the Layout 3.1 container version of Document Intelligence’s prebuilt-layout model, with the package

azure-ai-formrecognizer = "3.3.0"

to analyze documents.

This is what the log shows initially, when running the container:

Desktopazure-document-intelligence>docker-compose –env-file “.env.docker” up

[+] Running 1/1

 ✔ Container azure-form-recognizer-layout  Recreated                                                               0.2s 

Attaching to azure-form-recognizer-layout

azure-form-recognizer-layout  | 

azure-form-recognizer-layout  |

azure-form-recognizer-layout  | EULA Notice: Copyright © Microsoft Corporation 2020. This Cognitive Services Container image is made available to you under the terms [https://go.microsoft.com/fwlink/?linkid=2018657] governing your subscription to Microsoft Azure Services (including the Online Services Terms [https://go.microsoft.com/fwlink/?linkid=2018760]). If you do not have a valid Azure subscription, then you may not use this container.

azure-form-recognizer-layout  |

azure-form-recognizer-layout  |

azure-form-recognizer-layout  | Using API key for authentication.

azure-form-recognizer-layout  | Using '/share' for shared mounts.

azure-form-recognizer-layout  | Logging to console.

azure-form-recognizer-layout  | Submitting metering to 'THIS PART IS EDITED OUT'.

azure-form-recognizer-layout  | Is hosting: False

azure-form-recognizer-layout  | Build information: Form Recognizer Layout 2023-07-31  20240203.1-b0c0f5500a46efc1634694ae85c20587fa5d7ab0

azure-form-recognizer-layout  | Logging to console.

azure-form-recognizer-layout  | Using file-based Object storage.

azure-form-recognizer-layout  | Hosting environment: Production

azure-form-recognizer-layout  | Content root path: /app

azure-form-recognizer-layout  | Now listening on: http://[::]:5000

azure-form-recognizer-layout  | Application started. Press Ctrl+C to shut down.

azure-form-recognizer-layout  | WARNING: Logging before InitGoogleLogging() is written to STDERR

azure-form-recognizer-layout  | I20240628 18:24:01.944885    20 TextPipelineDevImpl.cpp:55] Text Pipeline initialize with option: UseModelDelayLoad: 1 UseGPU: 0

azure-form-recognizer-layout  | I20240628 18:24:01.945122    20 EnvVarUtility.cpp:55] Environment variable "ONEOCR_INSTALL_FAILURES_SIGNAL_HANDLER": "0".

When I send a request to analyze a document using begin_analyze_document() , and parse the JSON response, the language data is always empty:

...

"languages": [],

    "pages": [

        {

...

The Docker container’s log shows this after sending a request to analyze a PDF document:

azure-form-recognizer-layout  | StorageTimeToLiveInMinutes is set to 2880 minutes.

azure-form-recognizer-layout  |  Not a TIFF or MDI file, bad magic number 20517 (0x5025).

azure-form-recognizer-layout  | fail: formrecognizerlayout[0]

azure-form-recognizer-layout  |       Failed to track On-Premises Usage buildVersion='1.2.521.0-20240203.1-b0c0f5500a46efc1634694ae85c20587fa5d7ab0' hostId='5143f0834253' serviceName='formrecognizerlayout' requestId='6af7c387-2c81-4634-816f-4d36ba268fd8' tags='billing' namespace='vdi' Namesapce='vdi' billingUnit='1' errorCode='BillingInternalServerError'   

I know that the issue is not with the document because the language data appears for the same model when trying it on Document Intelligence Studio and programmatically using the non-containerized web service.

Any help figuring this out is much appreciated!