CrewAI LLaMa 3: How I created AI Agents to Automate News Agency

[@cgtinc4868]

Great job. Can you make another video of how to get the info to the database in the first place? thanks

[@d.d.z.]

Thank you Mervin, I'm excited to see more case uses. 🎉

[@saintsscholars8231]

Might be able to create a bet prediction algorithm that uses historical data to give predictions using similar data. Might need to include some additional tools for analysis however

[@JeffreyMerilo]

Why aren't there other implementations different from the examples from CrewAI?

[@MuneebMomin]

Amazing video! Man I’m a fan I think you were the first one to always post videos. I know Llama 3 you posted in 3hrs i think 😂 can you share what code assist you use? Really curious about it!

[@luismoriguerra669]

amazing simple example very clear, thanks!

[@Yakibackk]

Underrated channel ❤️

[@HyperUpscale]

I love your videos! I still can't understand... how can you create a project, test, demonstrate and create a video faster than I can follow and test... 🤯

[@YonatanSetbon]

if you run in cmd, pip install crewai[tools] flask requests (without the ' ' )

[@nigeldogg]

When defining agents in a yaml file in CrewAI, do you define tools the same way you would in a python file? I’ve been running into a KeyError…

[@swapnil0402]

Hi, may I ask you one help on this agent. In your tools , you created a functions and it has a team name as an argument. But inside task you did not send the argument. Does the Agent automatically figures out the argument from the prompt?

[@SamiSabirIdrissi]

another one! awesome!

[@spacecolumn5162]

thanks, it is very illustrative!

[@bertobertoberto3]

This is a nice way to ground results, would be nice to have another tool to lookup the same stuff on the internet, otherwise there’s really no details about the game and all of that is just hallucinated

[@MadeInJack]

I would suggest you to add a light source in front of you to improve your video quality :)

[@1242elena]

Question ⁉️ why use a score function call instead of using serper/google api which would provide that info plus more context

[@JoeSmith-kn5wo]

Because you have no clue what serper is going to return to you. You when create your own API with predictable data return to you get better results.

[@1242elena]

@@JoeSmith-kn5wo true 😂

[@ApurvaPragya]

in windows use set instead of export

[@watchdog163]

This sucks. What you really want is an agent that breaks down topics and creates its own agents and tasks.

[@NetZeroEarth]

Merv can you provide the flask api code and database please

[@1242elena]

Fantastic, this a major improvement. Llama 3 is powerful, fast, and free with Groq! That plus CrewAI provides a truly free, capable, and fast multi agent workflow! finally! It's great to see open source finally take the lead. No more money wasted on gpt-4!

[@cudaking777]

This is really greate it gaves alot of ideas and inspiration. Do you plan for more complex project having multiple agents with voice recognition , pulling the user portfolio and prefrence as well any history chat to adopt the AI assistant to support the user in best way. With capabilities of TTS and STT and access LLM both locally or function calling API the trick to to have a model selector for the right TTS or STT model as will the right llm model based on preformance matrix based on the task.

[@cudaking777]

The assistant will have multiple skills and tools and agent orchestration as well multi-agent system this is including research , project manager , task manager, coder. Al theses agents work to togther to cover any possible requests from the user. It also can yours sentiment analysis to understand the user intent from query and start activating and using these tools to have a faster response and also to have a more human interaction based on the mood and the behaviour of the user.

[@free_thinker4958]

@@cudaking777 we can use memory in crewai agents and the crew itself to have the short term memory , long term memory ,entity memory and contextual memory is the one that will combine all the previous 3 memories and pass it as future context to the llm with the iser query ...that is a good starting point to have an assistant that is built on top of agents and that will be personalized and self improve over time

[@yazanrisheh5127]

Keep doing more CrewAI videos please. Could you figure out a way where we can use GPT4 for best output but also somehow save costs? GPT4 with CrewAI is too expensive

[@free_thinker4958]

Use prompt engineering with only open source models

[@basilbrush7878]

Mixtral 8x7b and 8x22b perform just as well as GPT-4, particularly using the SerperAPI. I wouldn't bother with the expensive GPT-4 if you want to save money

[@yazanrisheh5127]

@@basilbrush7878 Thanks for the tips!

[@rastinder]

can you make another video for web scraping using lama 3 and cew ai and undetectable driver?? or at least tell if it is practical

[@Copa20777]

Thank you sir.. keep growing your channel with such relevant and practical videos, subscribed

[@aventurileluipetre]

where is llama 3?

[@AjirogheneSunny]

Hi, can we build an app around this, what do you think? I🎉