Get AI To Do Your Work! FOR FREE!! (Crew AI)

Crew AI is an amazing framework that allows you to create AI Agents, and assign tasks the agents will then perform. To each agent we assign a Tool that we can either choose from the langchain library or custom develop.
The entire process is sequential, and the agents can delegate tasks to each other and "discuss" ideas between them.
This is all done using a bit of python and a few extra modules like a KZbin api, reddit api and google trends. We can choose to either use OpenAI API or a local llm, however the results with local llm is quite poor, so chatgpt-4-turbo is preferred!
Hope you folks enjoy the video! All code is on my github!
#crewai #aiagents
Timestamps:
0:00 Intro
1:11 AI Agents We need
2:45 Crew AI
3:28 Create a python environment
3:55 Creating the Agents
5:11 Creating the Tasks
8:23 Tools
9:36 KZbin Trends Search Tool
11:56 Google Trends Search Tool
12:57 Reddit Scraper Tool
17:56 Langsmith
18:40 THE RESULTS
Related Videos: • Run your own PRIVATE A...
Code: github.com/fmiguelmmartins/cr...
Medium Article: / create-an-ai-team-to-m...
Sources:
CrewAI: www.crewai.com/
OpenAI API: openai.com/
Serper API: serper.dev/
Langsmith: smith.langchain.com/
Reddit API Key: / apps
KZbin API: developers.google.com/youtube/v3
Anaconda:
Instagram: / fx2mx3
Facebook Page: / fx2mx3

[@cybercdh]

You’re better than Network Chuck. Love this content, so glad I found your channel.

[@DigitalMirrorComputing]

Very kind of you, thank you mate! :)

[@TheDropForged]

Your tutorials are the best. Detailed and easy to understand.

[@DigitalMirrorComputing]

thanks for the support mate! :)

[@DesignbuildsolutionsLLC]

This is fantastic! Now, if I only had the time to do it for my own channel

[@toenytv7946]

As a KZbin content creator. This is conceptually very smart video to have made. Much appreciate all your hard work and time you are take to enlighten us with your understanding of ai in such a great use case. Hope I got and get it right. Cheers. Keep up the great works in the AI realm. Great ending too.

[@DigitalMirrorComputing]

wow mate, thank you so much for your kind words! Thanks for watching and for the support! :)

[@toenytv7946]

No thank you and just keep doing what your doing those that matter will support you those that don’t their loss. 🫵got this

[@gmssg6898]

Thank you so much!!!

[@DigitalMirrorComputing]

Glad you enjoyed it mate! :)

[@OfficeArcade]

Oh man, I'm totally trying this for my channel.

[@DigitalMirrorComputing]

It's super cool!! :D

[@RICHARDSON143]

❤

[@JacquesvanWyk]

SO how different would the answers be using local LLM to openai? Is it big difference in quality? I know it will be slower.

[@DigitalMirrorComputing]

It's a big difference mate. I actually did use local llm and had that bit recorded, but ended up cutting it off the final edit. The issue with local llm's is that they are not "large" enough or at least I can't run them on my setup, to yield good results. I did get some promising results from Nous Hermes2 and Wizardllm but waaaaaay to slow! Thanks for watching !! :)

[@JacquesvanWyk]

@@DigitalMirrorComputing I guess Groq for now. I am new to this but was thinking to use maybe cheaper models for not complicated work and for others Groq or GPT4. But I guess with time and experience we will see what works best.

[@DigitalMirrorComputing]

@@JacquesvanWyk The only limitation you will encounter with GPT-4 are tokens. Depending on how much information your agents send to the LLM. GPT-4 has a limit for 8K tokens, and if you scrape reddit you can very easily that number. But you could also tame that data and treat it, so that you don't incur into higher cost. I just sent everything lol, hence why I had to go for turbo. But removing emojis and filter the data is a good start! Good luck mate!

[@user-pt3yf5uv8b]

I would love to see this done with local llms. I can afford using chatgpt 4. I tried it out and it ate $12 just with 2 runs 🙄

[@RalphKrausse]

Korn!

[@DigitalMirrorComputing]

hell's yeah!!! :D

[@vivaldi0]

Always go with open source, as I just left openAI, due to .....well.....let's say an AI that has 'wokeness' built into it, and I cannot have that influence my project. Bias from Seattle and the Northwest must discontinue. They will no longer get my 20/month. I will run it locally. I have a mighty powerful computer as most, so I would love to see the open source route, before the openAI route.

[@josephpark3949]

You said free in your title, and then said this is all paid for near the end and didnt say how to do it for free and instead just talked about answering calls. You used chatgpt 4 which is not free. How do you do it for free?

[@DigitalMirrorComputing]

Oh shit you do have a point there! you can use ollama local AI instead of GPT 4! When I said Free I was thinking about not hiring someone to do the job! Sorry if I mislead you, it was unintentional! :)

[@josephpark3949]

@@DigitalMirrorComputing thanks for the quick reply! I'll do that. Thanks