---
title: "Let me Google that for you"
video_id: ie_S4Y7DiDU
url: https://www.youtube.com/watch?v=ie_S4Y7DiDU
workflow_page: https://workflows-of-ai.com/workflow/let-me-google-you
language: en
captions: generated
---

0:01 hey there um Marcus here so uh I'm going
0:04 to talk about the workflow today that I
0:07 already tried to record actually uh
0:10 something went wrong with the recording
0:11 so I re-record it
0:13 now um and it's uh about uh The Sur
0:18 plugin so search engine results page so
0:21 basically a plugin that can Google for
0:24 you which makes it possible to do a lot
0:26 of research uh fact checking stuff like
0:31 that the workflow I show today is very
0:34 simple it's just going
0:35 to uh basically analyze a
0:39 question and then search for actual
0:42 answer so not
0:43 ask uh open AI the question right away
0:46 because open AI as everyone know can
0:50 hallucinate so it needs context to be
0:53 trustworthy and in this case we're going
0:56 to create the uh context via search
1:01 as always what you see in front of you
1:02 this is trole open source
1:05 free uh you can download it afterwards
1:09 um download AI interpolator modules that
1:12 I'm going to show and set this up as
1:15 easy as it looks
1:17 here so I'm going to start by logging
1:19 into the
1:21 system um and then create that content
1:25 type that we call
1:28 research as always if you're going to do
1:32 this as a real product or use big data
1:35 or so on you should use Uh custom entity
1:39 types but in this case I just want to
1:41 show how you can graphically via or
1:44 sorry via the GUI set up everything so
1:47 instead of title we just call this a
1:49 question that's going to be the question
1:51 we want answered
1:54 basically um there comes a body field
1:57 with druple by default in the nodes the
2:02 content types so we're going to delete
2:04 that that's more for CMS
2:06 right uh we're going to start by adding
2:10 the first
2:11 field which is going to be a text plane
2:14 and this is going to be the search
2:17 word
2:19 and I it's just going to be one search
2:22 word right so that's going to be the
2:23 search word we use in Google to find
2:26 information that might answer the
2:28 question or
2:30 so because we don't want to fill out
2:32 this uh manually because we might not
2:35 know what we search for we're going to
2:37 use open AI for this so we're going to
2:40 use the open AI
2:41 string that takes a string and does
2:44 something on that
2:45 string uh and generates a
2:48 string uh we use base mode and we use
2:51 the question as the base field
2:55 so then we create a prompt here then
2:57 based on the
3:01 follow in
3:04 question create a search
3:09 phrase for Google for
3:14 finding results that might answer
3:19 it
3:20 question and then we're going to copy
3:23 context because that's the context of um
3:27 basically the question we're ask asking
3:29 right
3:31 uh we're going to run this in batch so
3:34 we uh see the graphical output take gp4
3:38 it's a little bit better quality save
3:41 this and then comes then the Ser field
3:44 and what we need as the field to fill
3:46 out we need
3:50 links uh we call these search
3:55 results and this can be unlimited so or
3:59 or you can set a fixed number basically
4:01 these are the amount of search results
4:03 you want back that can be defined here
4:05 in the inator then it's just going to be
4:08 external links no link text is needed
4:11 and we enable the AI interpolator here
4:13 as well and the rule we want to run is
4:16 Ser to
4:18 link
4:20 and we say I don't want it to take too
4:23 long now I'm just going to ask some
4:24 simple questions I just want three
4:26 search results because of course it
4:28 needs to scrape every page
4:30 it need to analyze every page in open Ai
4:33 and so on so it takes time but of course
4:35 if what you're researching is complex
4:38 you should try to get as many search
4:39 results as possible you can do search of
4:41 that so you can get 10 pages in batches
4:45 then start from page uh from 11th
4:48 results start from 21st results and so
4:51 on you can search on use you can choose
4:55 which Google to search Which
4:57 languages uh you search with Which
4:59 languages you want to have in the search
5:03 result and which country you're
5:05 searching
5:06 from
5:08 um we're going to do this on the base
5:10 mode and we're going to use the search
5:12 word that we generate then it's
5:14 important here that we set the higher
5:16 interpolator W uh than the initial query
5:22 because we need the search word before
5:23 we can search right and we said patch
5:26 again after we have done that we need to
5:29 to get the actual content from this so
5:32 we're going to use text format long and
5:35 we call that website
5:37 content uh that's the content we get
5:39 back from the links these also have to
5:42 be well as many as the links
5:44 are allow full HTML enable Ai
5:49 interpolator and in this case we're
5:51 going to use the scraping Bol crawler
5:53 Rule and we're going to use this from
5:56 the search results so the Link's there
5:58 basically
6:00 uh we can get this from USA we're going
6:02 to use chromes or use a real
6:05 browser uh we use readability mode so we
6:08 say we just want the content that is the
6:10 actual
6:11 article and we have to set this to an
6:14 even higher weight then because we have
6:17 to make sure that uh the links exist
6:20 before we can start
6:22 scraping and then basically comes the
6:26 last thing we need which is a text plan
6:29 long that's going to be the
6:33 answer and this is the answer to the
6:36 initial question based on the scraped
6:40 content right so we're going to ask open
6:43 AI to an answer the question based on
6:47 the scraped content we're going to use
6:48 advanced mode for
6:50 this uh and how we write this then based
6:56 on the context
7:00 answer the following question in
7:06 around
7:08 three
7:11 paragraphs
7:13 uh yeah three
7:16 [Music]
7:18 paragraphs
7:20 question so what you can do with uh the
7:24 advanced mode with tokens then is of
7:26 course that you
7:27 can you can basically
7:30 uh use multiple inputs which is what we
7:33 need in this
7:36 case so we find this under
7:39 nodes and the question was on the title
7:41 that's why you don't see question here
7:44 so we take node title for
7:47 this and
7:49 then the
7:52 context which is then the
7:56 website content this was called
8:00 uh and this is the prompt we're going to
8:03 use uh we have to set an even higher
8:07 interpolator weight we do batch and we
8:10 take a bigger model in this
8:15 case because we don't know how big the
8:17 context is going to be so we use a
8:19 16k um model so we make sure it fits
8:23 inside the
8:25 context there is one thing I should
8:27 probably do here as well since we're
8:29 asking for paragraphs it might actually
8:32 generate multiple answers and split up
8:34 each paragraph As one so this is prompt
8:38 engineering of course you could write
8:40 three paragraphs in one result set or
8:42 something and it would probably work so
8:45 you would have to find the best prompt
8:48 possible there if you want it in just
8:49 one field but I just do multiple Fields
8:53 just to make sure that uh you will see
8:57 the result at least
9:00 uh and that's
9:02 it I'm going to do another thing I'm
9:04 going to hide the website content from
9:06 the manage display so the end output uh
9:11 because that's going to take up so big
9:14 space that I have to scroll and you
9:16 might not know what the actual answer is
9:19 so then we add the
9:22 research uh and I have a question here
9:25 that's very interesting what is Link's
9:27 personality in the Zelda games
9:30 uh and you see all the other fields
9:33 here those are the fields now that's
9:35 going to be outo fi basically so I click
9:40 save uh this is going to take a little
9:43 bit of a time um because now basically
9:47 it is generating a search
9:49 word searching for this to get the links
9:52 scraping those links with a real
9:55 browser uh and in the end asking open a
9:59 I then to answer the question based on
10:01 the content inside those
10:06 links oh let's see it didn't actually
10:10 answer it let's just try to resave and
10:12 it might
10:15 work
10:16 [Music]
10:20 um so maybe I set up the order wrong
10:23 let's see yeah here we go so here we
10:26 have an answer then links personel in
10:28 the selda Gam where it's depending on
10:30 the specific game yeah you can see the
10:33 answer here
10:34 basically if we click edit you can also
10:37 then see what you saw before that
10:38 there's
10:40 actually uh script content and this
10:43 content is then what went into the
10:48 question uh to open AI so to make sure
10:52 basically that it has a context to
10:54 answer on meaning that it can answer
10:58 very well I'm not going to say
11:00 truthfully because of course you should
11:02 not just trust all the links you find in
11:05 Google but the probability of it
11:08 answering is uh correctly is a lot
11:11 higher at
11:12 least uh but as I said this is a very
11:15 simple workflow I showed
11:17 now um if you just flex your creativity
11:21 muscles I'm sure you can think of so
11:23 many more use cases for
11:26 this so yeah thank you for watching