Vertex AI Search or Document AI

[MB0]

Hi,

I want to query DOCX/JPG documents to check if they contain specific sections of text.   We want to use out-of-the-box RAG and hence thinking of using Vertex AI Search and PaLM2.    Do we then still require Document AI to extract text and feed to Vertex AI Search or that's already taken care of by Vertex AI Search?    Also, in this use case, PaLM2 is better or Gemini models?

Thank you.

[caryna]

Hi @MB0,

Welcome to Google Cloud Community!

Vertex AI can handle text extraction from certain document types. However, it’s recommended to use Document AI for higher accuracy and better support.

As for model selection, PaLM2 is considered as the most powerful general-purpose LLM model at Google. However, model users are advised to migrate to Gemini models as it delivers significant performance improvement over old models. 

Here’s a comparison of the two models:

• PaLM2: it has better multilingual and reasoning capabilities.
• Gemini: is better with Q&A style prompts in some scenarios.

Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.

[MB0]

@caryna thank you for your response.   SInce we do plan to use Gemini 2.5 models (Flash lite/Pro) which are multi-modal and can handle Q&A on entire document, why is Document AI then necessary?  Our documents are small - company policy docs etc. which do not require chunking/embedding.    Even Gemini chat app suggests me the same.