Hi everyone,
I am interested in using Gemini Deep Search in an automated way. As far as I can see, Gemini 1.5 for deep search is only available through chat right now.
Do you have any idea when it could be accessible by API?
Cheers
Hi everyone,
I am interested in using Gemini Deep Search in an automated way. As far as I can see, Gemini 1.5 for deep search is only available through chat right now.
Do you have any idea when it could be accessible by API?
Cheers
Hello,
Welcome to the community!
The Deep Research feature works by:
Searching various websites to gather relevant information.
Summarizing the findings into a concise response.
Providing a grounding citation to ensure accuracy.
It is bit expensive but it worth the money, if you have high stake tasks. It is not available on API However you can create your own Deep Search.
How to Create Clone Deep Research
To generate your own deep research, follow these steps:
Query Classification – Use a classifier model (powered by Gemini Flash) to check if the query actually requires deep research.
Outline Generation – deep research need an action plan outline.
Google Dorks for Search – Use Gemini Flash to generate Google Dorks based on action plan, optimizing search queries for the most relevant results.
Scraping Multiple Pages – Collect data from various sources to ensure a broad perspective.
Bias Removal & Validation – Use a final validator model (using the most capable AI models) to filter out misinformation, ensuring the results are as much as accurate and less biased. And cite all the sources.
This final validation step is most important in removing false or misleading data from web results that LLM already know that may be inaccurate or falsiable.