Issue with Integrated vectorization enabled

[TB-Shyam]

I enabled integrated vectorization in the environment variables. Deleted the old index and run azd provison. I get the below error.

It states that Field mapping specifies target field 'title' that is not present in the index.

I checked the searchmanager.py file and that index schema doesnt have the title field in it.

I have two options:

We have two options:
a) Add a "title" field to the index schema in searchmanager.py
b) Change the field mapping in integratedvectorizerstrategy.py to map to an existing field like "sourcepage" or "sourcefile"

Which way would you recommend me to proceed with?

[Bennzn]

I receive the same error message (seems the last update of the repo?) even when I try to create a fresh new index using prepdocs.py or using azd up. Also with a complete fresh initialized project and start it up from scratch.
Tried on 2 different clients, changed python version ecc.

Or there was some change on the backend of the AI Search service? As there is a HttpResponseError

I compared the relevant py files with an older project that I spun up 2 months ago without a problem, but didn't find some difference

Cheers,
Ben

[Bennzn]

Side fact:
On an existing deployment which was working very well, now the existing Indexer throws the same error.
We did not redeploy or other. It was a "productive" demo.
New the indexer fails always with "Field mapping specifies target field 'title' that is not present in the target index" like during a new / fresh deployment.

So for me it seems there was a change in the backend of AI Search.

EDIT:

It is / was a change on the backend:
https://learn.microsoft.com/en-us/answers/questions/2182611/ai-search-indexer-started-failing-with-field-mappi

When you change in integratedvectorizerstrategy.py line 175:

from:
field_mappings=[FieldMapping(source_field_name="metadata_storage_name", target_field_name="title")],

to:
field_mappings=[FieldMapping(source_field_name="metadata_storage_name", target_field_name="sourcefile")],

The deployment succeed.

Need to check now the impact on existing indexes and change the field mapping

[TB-Shyam]

Thank you so much. In my case I have added the title feild in the index. It works for me in that way

[egor-yudkin]

I believe changing the line 175 in integratedvectorizerstrategy.py as @Bennzn suggested is the right approach.
The rest of the application doesn't use field "title", but uses "sourcefile" in some operations such as updating content in index.

@TB-Shyam you may want to test if the index is being updated properly after your change.

[pamelafox]

I am fixing this in upcoming PR to port to embedding-3 models:
https://github.com/Azure-Samples/azure-search-openai-demo/pull/2470/files#diff-d1555d6523fdd0048963ceb0b11ef106a227cb9838b956eb57e80b1ea5838349R111

But you can also bring in those changes to the mapping. We also added storageUrl as it was easy to do.

[TB-Shyam]

@pamelafox Could you please guide me how I can bring this changes so its fixes the integrated vectorization issue that I am currently facing. Thank you

[cforce]

Please merge the PR with the fix included or at least merge the below quick fix because without app is not runnable (prepdocs fails which is mandatory for app deployment)

from:

field_mappings=[FieldMapping(source_field_name="metadata_storage_name", target_field_name="title")],

to:

field_mappings=[FieldMapping(source_field_name="metadata_storage_name", target_field_name="sourcefile")],