Sean-HynesQ: Looks like a great product thank you!
In the past, I have gotten from OpenAI , "I cannot extract or browse the web". DO certain LLM libraries don't allow this?
Also, Do certain Functions NOT Need to go to a External LLM? like Extract Email, that could be a internal Regex Script hopefully.
Yaro_DocGPT_AI
May 15, 2024A: Hi Sean!
You can use =GPT_EXTRACT() function for extracting some data from data :)
Or use customer prompt like =GPT/Mistral/Gemini/Claude("Extract email form the text:", A1)
- For accessing to the WEB, you can use following built-in functions in current add-on:
WEB SCRAPPING:
=WEB_URL("https://docgpt.ai")
=WEB_TITLE("https://docgpt.ai")
=WEB_DESCRIPTION("https://docgpt.ai")
=WEB_H1("https://docgpt.ai")
=WEB_H2("https://docgpt.ai/")
=WEB_HEADINGS("https://docgpt.ai")
=WEB_PARAGRAPHS("https://docgpt.ai")
=WEB_URLS_DATA(A37:A39)
About web scraping, which model can/should I choose? Or do you scrape and then pass it on to gpt/Mistral/Gemini/Claude?
"About web scraping, which model can/should I choose? Or do you scrape and then pass it on to gpt/Mistral/Gemini/Claude?"
Web scrapping does not use any AI model. It makes direct GET requests to the links. it supports query params and any long of uri.
You can make using the follow flow for example - i want to rephrase competitors titles, descriptions or h1,2 tags:
1. Make =WEB_TITLE("https://docgpt.ai")
2. Click on button on the sidebar - Replace selected function (and select current cell with result)
3. Thits button will replace your function with the text result
4. After that in a new cell write =GPT("rephrase title for better click rate", A1). Where A1 is text result of prev fetched title
Thanks Yaro for the reply on "Web scrapping does not use any AI model."
I just do not really understand WHO scrapes the website. Is this a inbuilt function of Google Sheets, or is it you/a function of DocGPT (if yes any limits?), or takes scraping place from my IP adress? Also are the any options like residential proxies offered or recommended if the scraping is blocked (which may happen if you scrape hundreds of webpages in short time).
"I just do not really understand WHO scrapes the website. Is this a inbuilt function of Google Sheets, or is it you/a function of DocGPT (if yes any limits?)"
I have implemented all functions via native code for Add-Ons to provided the best possible performance.
So there is no custom "middle server".
This Add-On functions are executed in cloud of Google. which could use some period of time the same external ip and often changed.
To have as diff as possible, just refresh the page with spreadsheet as often as possible. ( it will rebalance your requests to diff server "pods" )
There are limitations only for amount calling functions for Google Sheets/ per account. I do not remember how much, but if you are not going to ddos someone, you should not to reach them :)