How to connect the json script obtained from chatgpt as index in Cognitive search

[suma-sai-paluri]

Please provide us with the following information:

This issue is for a: (mark with an x)

- [ ] bug report -> please search issues before submitting
- [ ] feature request
- [ ] documentation issue or request
- [ ] regression (a behavior that used to work and stopped in a new release)

Minimal steps to reproduce

Hi @pamelafox can you please help me with the solution. I’m implementing a chat system using my company data which are offer documents (each ranging 15-25 pages long) made to different companies. I would like to perform numeric searches, date searches such as “How many offers are made to the company called aaaa in the past five years” or “Provide me the list of all the offers greater than 10000 euros”. I don’t have a specific file which has this kind of data written in it.

Inorder to achieve this i have tried extracting the related elements in the system message : Reply only in JSON format with the following structure: {"amount": ,"company_name": , "date_of_the_offer":, "duration": } for chat gpt 35 turbo and saved the obtained results in the page text which is obtained during the get document text phase in predocs.py. Later created added the above elements in the schema of the search index and made filterable ,searchable etc.
when i tried checking the results of the index in cognitive search site. The text retrieved for the above elements after the search is incorrect. The keys are pointing to the wrong text.

Please provide with any sources that could help me.

Thanks! We'll be in touch soon.

[github-actions]

This issue is stale because it has been open 60 days with no activity. Remove stale label or comment or this issue will be closed.

[pamelafox]

Thanks for raising this. The sample doesn’t auto-map JSON keys into separate fields. To enable numeric/date filtering, you’ll want to add dedicated fields in app/backend/prepdocslib/searchmanager.py (e.g., amount as Double, date_of_the_offer as DateTimeOffset, duration as Int32, company_name as String with filterable/sortable as needed), update the ingestion to populate those fields (instead of embedding the JSON into content), and recreate the index. The docs in docs/customization.md cover how to modify the index and reindex.

Since this is an older thread and not a bug in the sample, I’m going to close it. If you’re still running into this, please open a new issue with your field definitions, the update_content payload you’re sending, and the query/filter you’re using in the portal, and I’ll take a look.