Thanks for the nice video and the files. But with ollama and phi3, they don't work well. I tested the QuickFS api separately, it returns the data, but for some reason the agents are clueless about the rest of their tasks and they just repeat what they need to do without doing anything. I guess gpt4 makes the difference.
You could have searched Google for the urls Site:"www...." "?id=" This will give you all urls for doctors pages, from there, scraping data is easy with beautiful soap. Happy scraping
I tried using function calling for invoice data extraction, but when the schema content and description got big I noticed a weird regression where the gpt will return a weird {text:nonsense} instead of the valid schema, for reference I was using gpt 3.5 1106
Hey, I have this concept in mind. An AI which can do anything possible. If it needs the help than it can surf the internet on it's own. Is it possible? e.g if I ask it to create a new blog and earn money for me or hack into a website etc etc (these are just the examples hehe) . A high level AI model which can do anything asked
CrewAI creator here, so cool to see how other people use it! I'd recommend you trying the new version, even simpler to use, agents and tasks can be yaml files now, a bunch of our own tools now that unlocks even more use cases :) Great video, keep it up!
Its way faster to get it your self 😂 AI should help with behavioral analysis and pattern that we human have hard time to identify and act upon if its repeats but no one has ever creates that yet.
you're totally right. could have just been a few api calls like i mentioned, but its alot of fun and i was just trying to learn about crew!! gpt-4 is actually really good at fixing itself when it makes a mistake but it ends up being pretty expensive :( crew is improving and i think with a few smaller models it will eventually be able to do what we're all dreaming of like u mentioned 😍😍
This was great. I really like seeing code on the screen. The tutorial of how to use the product has so much more value than just a review. I see you list a discord. I guess I'll be heading there next. 😊
I am a big fan of crewAI. One of the problems is with how rapidly it is iterating. I'm now on version 11.2 and find myself having to run through my code from just a week or so ago to see if I need to update it with new features and tools that seem to be appearing daily. It will settle down as it matures of course, but it is fun seeing how it improves and I'm glad the developers are so responsive. I liked your examples outputting charts and content through markdown files. I am a big Obsidian fan and as that uses markdown I do like using crewAI to create new folders and files in my Obsidian vault to store the outputs from my Crews.
definitely an issue i encountered, i'll probably have to go back and edit this project to work with new changes as well. looks like they're gearing up for another release so i'm excited for that! ty for tip on Obsidian ive been hearing alot about ti recently i think its time to give it a shot! thx for stopping by!
hey! sorry for late reply. check this out and lmk if this work js.langchain.com/docs/integrations/llms/ollama I have some python examples in this video that would use Ollama or HF ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-U_Sg3Odf1vk.html at 05:06 It is in python but if you search the `javascript langchain` docs u should be able to find a way to do the same in js! lmk if u cant figure out i will try to help. thank u!!
Thank you very much for this! I work in the IT industry and the one way I continue to learn is by doing. I would love to see this whole process completed and follow along.
hi sorry for late reply! were u able to get this working? which version of python and package are u using? happy to help pls link a gist and i'll take a look! thank u!
does copilot do the @docs functionality for packages? that is the best reason to use cursor over copilot at the moment, it uses rag to pull in the latest docs and references that to write your code.
hey! so what you can do is put all of these commands in a "script" in vscode (basically a python file, it'll look like myscript.py) and run it from your terminal. you can probably ask ChatGPT to help you with this but that's the basic idea. lmk if u have any other questions, happy to help