Document Intelligence Cold Start

Question

Document Intelligence Cold Start

Casey 65

The cold start seems to range from 15-25 seconds for standard layout OCR.

If I ping it to stay warm, how long will it stay warm for?
Are there any dedicated service plans I could use so it's always warm?

Thanks

dupammi 8,615 Reputation points Microsoft External Staff

2024-05-08T08:10:22.8733333+00:00

Hi @Casey Clifton

Thank you for using the Microsoft Q&A platform.

Regarding your first question, when you ping the Document Intelligence service, it will stay warm for a certain amount of time depending on various factors such as the service's usage patterns, traffic, the number of requests, resource allocation and the service's configuration. Unfortunately, there is no fixed time limit for how long the service stays warm after being pinged. However, you can optimize the service's performance by using the right configuration and settings.

Regarding your second question, there are no dedicated service plans that can keep the Document Intelligence service always warm. However, you can use Azure Functions and try if it can keep the service warm by scheduling a function to ping the service at regular intervals. This might ensure that the service stays warm and ready to process requests.

I hope you understand. Thank you.
dupammi 8,615 Reputation points Microsoft External Staff

2024-05-09T03:21:31.8733333+00:00

Hi @Casey Clifton

We haven’t heard from you on the last response and was just checking back to see if you got a chance to check my above suggestions. Thank you.
Casey 65 Reputation points

2024-05-10T00:36:30.9833333+00:00

Hi @dupammi thanks for your response. FYI we may switch to an alternative OCR approach given this response time uncertainty so a better solution would be valuable in our case.

1 answer

Your answer

dupammi 8,615 Reputation points Microsoft External Staff

2024-05-08T08:10:22.8733333+00:00

Hi @Casey Clifton

Thank you for using the Microsoft Q&A platform.

Regarding your first question, when you ping the Document Intelligence service, it will stay warm for a certain amount of time depending on various factors such as the service's usage patterns, traffic, the number of requests, resource allocation and the service's configuration. Unfortunately, there is no fixed time limit for how long the service stays warm after being pinged. However, you can optimize the service's performance by using the right configuration and settings.

Regarding your second question, there are no dedicated service plans that can keep the Document Intelligence service always warm. However, you can use Azure Functions and try if it can keep the service warm by scheduling a function to ping the service at regular intervals. This might ensure that the service stays warm and ready to process requests.

I hope you understand. Thank you.
dupammi 8,615 Reputation points Microsoft External Staff

2024-05-09T03:21:31.8733333+00:00

Hi @Casey Clifton

We haven’t heard from you on the last response and was just checking back to see if you got a chance to check my above suggestions. Thank you.
Casey 65 Reputation points

2024-05-10T00:36:30.9833333+00:00

Hi @dupammi thanks for your response. FYI we may switch to an alternative OCR approach given this response time uncertainty so a better solution would be valuable in our case.

Answer 1

dupammi 8,615 Microsoft External Staff

Hi @Casey Clifton

Thank you for sharing your feedback.

I understand the challenges you've faced with the cold start response time of Document Intelligence service's and your exploration of alternative OCR solutions.

Should you seek alternatives, I suggest considering Azure Cognitive Services' Computer Vision OCR API. This API offers OCR capabilities for diverse image types, including handwritten text, and boasts extensive language support.

You can find more information about the Computer Vision OCR API in the following Azure documentation:

OCR for images - Computer Vision

I hope this information helps.

Casey 65 Reputation points

2024-05-17T02:17:55.75+00:00

Does this API have improved speed or stability of response times?
dupammi 8,615 Reputation points Microsoft External Staff

2024-05-17T02:25:24.2633333+00:00

Hi

Thank you for your follow-up question.

While the performance of the Computer Vision OCR API may vary depending on various factors such as the size and complexity of the image, the API is optimized for speed and stability of response times. Additionally, the API offers a range of features and settings that can be customized to optimize performance for specific use cases.

I hope this information helps.

Share via

Document Intelligence Cold Start

1 answer

Your answer