Add support for GPT-4v

[LittleLittleCloud]

Please provide us with the following information:

Feature overview

• generate image embedding and create indexing for pdf files

  • split document into pages and save as PNG in blob storage
  • generate image embedding for each page using vectorize-image-api
  • add image embedding field to search index

• Update ReadRetrievalRead service to generate answers using GPT-4V

  • When retrieving, create query embedding using vectorizeText api and get top-k most related pngs from blob storage
  • Then send retrieved pngs along with query to GPT-4V to generate response.

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

[Coruscate5]

Just as a cheering section, I'm actively looking forward to this feature - I've been testing GPT4v but we've mostly been focused on data syncing optimization in AI Search, so excited to see some samples on it!

[luisquintanilla]

TODO:

• Document azd env flags:

azd env set USE_AOAI false
azd env set USE_VISION true
azd env set OPENAI_API_KEY <YOUR-OPENAI-KEY>
azd env set OPENAI_CHATGPT_DEPLOYMENT gpt-4-vision-preview