Web Scraping AI AGENT, that absolutely works 😍
Vložit
- čas přidán 8. 05. 2024
- ScrapeGraphAI is a web scraping python library that uses LLM and direct graph logic to create scraping pipelines for websites, documents and XML files. Just say which information you want to extract and the library will do it for you!
🔗 Links 🔗
Scrape Graph AI
github.com/VinciGit00/Scrapeg...
Code used in the video - github.com/amrrs/scrapegraph-...
❤️ If you want to support the channel ❤️
Support here:
Patreon - / 1littlecoder
Ko-Fi - ko-fi.com/1littlecoder
🧭 Follow me on 🧭
Twitter - / 1littlecoder
Linkedin - / amrrs - Věda a technologie
this ScrapegraphAI tool is the most interesting scraping tool I've tested so far
"poor" Love you brother! Right there with you. Great video. Been trying and failing to get a scraper with java support. Cheers!
Someone noticed it :D
Really great video, thank you. I would be interested in seeing more videos about ScrapeGraphAI.
Awesome we need more practical session with code like this.
I like all your videos , keep rocking bro
Great video! Thank you!
Next time it will also need a visual model to solve capchas because website administrators will be protecting their precious content from scraping :)
Haha
thanks for the heads up at 6:00 .
worked when using that version only
I find this live example pretty useful for general purpose, I can think of multiple ways I could use this for one off PoCs
Glad it was helpful!
There is a playwright function that bypasses the irrelevant resources so the scraping becomes faster
It’s a spider, not an octopus. Spiders crawl on webs.
What is an octopus? Which crawls API's or do datamining
Amazing ! If my PSU wasn’t dead I wouldn’t be sleeping for days
We need more tutorials of practical live examples of llm especially rag and fine tuning
It depends on the llm used and questions you pose it. It can often not generate json and the library isnt best suited for iteration through a collection of sites
If your LLM gives you an article you can't find, my first assumption is that it made it up. While this is an interesting use case, it's going to likely take very precise prompt engineering to not get hallucinated outputs.
No, it's my bad. After the video I reviewed the web page. In fact, I added the screenshot in the video. It was inside the carousel
Could you please do more videos on this. Like trying to use it on more educational content with equations used using mathjax and katex
i luv it❤
thanks! 👍
Watching this before youtube gets upset again. 😉
Honestly, I was actually scared before uploading this, but let's see!
@1littlecoder Hopefully all is well.
Ok, now that's a good useage of ai model
thanks for the tutorial, please make more tutorial for this ScrapeGraphAI, can you make one for scraping the website that has antibot or credential (require login)
great video....
Glad you enjoyed it
Is the llm there to convert the raw html to structured data? Then it saves to rag and you can query the data with another llm to analyse? I need to scrape homepages from 10k sites tostructured data into rag db to ask The sites questions, can it be setup todo many sites like an automated agent, or can it be used as a tool or function call in an agent framework like crew ai? that video would be cool
RAG: Ritual Augmented Generation 😂
🕯️🕷️🕯️
🥇
Need proper pdf parsing ai that I can run on a cloud server without gpu. Extracting text, tables and images and arranging it in a db based on a prompt that puts each data in the right table. That will be amazing if you can find something like that.
reddit being called front page of internet is like... no please
What am i missing.... error running the async cell?
does it work on google colab?
That logo is a spider 🕸️🕷️
I fail to see why rag is needed when the library can simply be asked to identify the html path/element that contains the content, and then extract the html from that with cheerio
I tried to install the scrapegraphai but I'm getting stuck in the yahoo search dependency which breaks the execution and return attribute error.
it is the same error or not here: I'm very beginer level if someone know how to fix mine please leave the comment
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
Cell In[25], line 17
3 graph_config = {
4 "llm": {
5 "model": "ollama/mistral",
(...)
13 }
14 }
16 # Instantiate the SmartScraperGraph class
---> 17 smart_scraper_graph = SmartScraperGraph(
18 prompt="List me all the articles",
19 source="news.ycombinator.com",
20 config=graph_config
21 )
23 # Run the smart scraper graph
24 result = smart_scraper_graph.run()
File ~\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.11_qbz5n2kfra8p0\LocalCache\local-packages\Python311\site-packages\scrapegraphai\graphs\smart_scraper_graph.py:47, in SmartScraperGraph.__init__(self, prompt, source, config)
46 def __init__(self, prompt: str, source: str, config: dict):
---> 47 super().__init__(prompt, config, source)
49 self.input_key = "url" if source.startswith("http") else "local_dir"
File ~\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.11_qbz5n2kfra8p0\LocalCache\local-packages\Python311\site-packages\scrapegraphai\graphs\abstract_graph.py:49, in AbstractGraph.__init__(self, prompt, config, source)
47 self.config = config
...
--> 227 params = self.llm_model._lc_kwargs
228 # remove streaming and temperature
229 params.pop("streaming", None)
AttributeError: 'Ollama' object has no attribute '_lc_kwargs'
Output is truncated. View as a scrollable element or open in a text editor. Adjust cell output settings...
wondering when somebody will integrate `undetected-chrome` to it.
So, this is impossible to run in Colab? I like to automate many of my tasks using Github actions.
You can run on colab. But you'd need openai keys
it still have the risk of being blocked, it's just a way of parsing
Will it work to scrape linkedIn jobs?
another question , what if we only want to scrape and not emmbed anything ?
I think in those cases you can probably use a conventional libraries I guess but that's a good question there are different classes within this library that might let it do
@@1littlecoder
from scrapegraphai.graphs import BaseGraph
from scrapegraphai.nodes import FetchNode, ParseNode,generate_answer_node
graph = BaseGraph(
nodes={
fetch_node,
parse_node,
},
edges={
(fetch_node, parse_node),
(parse_node, generate_answer_node),
},
entry_point=fetch_node
) .. i dont have time to try it now cause im at work :))
scrape Facebook please! - I need to do the most boring thing for work, I tried to program a scrapper but FB makes it very hard, I was only partially successful (expecially grabbing the post date). This method looks very exciting :)
What was ur requirements?
@@webhosting7062 I write a daily report, based on the new posts in various FB groups .. but FB doesn't put posts in the correct order (also, pinned posts up the top will be old posts) .. so i need to check the date, but, FB obfuscates the date like a MF .. i wasn't able to figure it out with selenium.
so, requirements are .. 'get the latest (less than ~24hr old posts) from a FB group.
BTW, that's a spider in the logo. It's a spider that lives in the World Wide Web 😅
How did I not even think about it?😭😭😭
@@1littlecoder :)
🤣 I enjoy your sense of humor. Thank you. You are RICH in kindness and intelligence. That’s almost as good as money…. Money only buys limited amounts of happiness.
Your videos are very helpful and informative. I’ll pay you to help me figure a couple things out. What’s your contact?
Thank you 1littlecoder@gmail.com is my email
Only is own-lee
Not one-lee
Btw great video
😭 will try to fix it!
looks something like spider (scrape/crawl) + bone (GET/fetch) + document | parse ( HTML) ???
Plus RAG, yes!
This is great, thanks for sharing, Can you share your python version as i am getting an error when running the same code
I guess mine is 3.9ish
Would it be possible to explain the whole thing to someone who has nothing to do with programming? I was able to install everything but I can't do anything with the code from github...
Would be great :) Thanks for the video! Very interesting but unfortunately not feasible for me.
(I'm on Linux)
Do you want me to show how to run the code from GitHub? Will it be helpful
Yeah! At least in a way that's easier to understand. I don't know anything about code, so I need things to be clear and simple.
Thanks!:)
I don't understand that for web scrapping why do I have to install so much of other dependencies like ollama etc. I mean it is just a simple webscraping why make the thinks complex? Still for the complex task a complex prompt needs to be given.
ollama is just a frmework to run LLMs locally, so it downloads the model insted of using an API and connecting to server
If you just want scraping, don't bother with this.
However, if you want scraping + RAG, with LLM integration, then use this. But it's not without it's issues
can it do heavy JavaScript sites ? :))
I've not tried it! it'd be a good opportunity to try that, especially given it uses Playwright!
@@1littlecoder ill tell ya, i tried and it fails miserably :)) , if you have better luck let us know man
@@adriangpuiu ah that's bad. Which website was it ?
@@1littlecoder the user replyes are incapsulated in a JS response from what i noticed, maybe they have an api or soething , i was just unable to figure it out . YET ...
@@1littlecoder its the appian discussion forum
I think Chrome extensions are best.
What about site build with jquery.. Does it works for that too?
I have not tried it . Someone else in the comments said it might not very good.
I like all your videos , keep rocking bro
Thank you so much 😀