Step-by-Step Langtrace + CrewAI Tutorial - Production Agent Stack

[@RunnerProductions]

Yes tutorial and actual use function videos are the best. More value add with actually things to be able to do rather than some channels that just talk hype and news all the time

[@cbarrue]

great content Mathew, as always

[@A_Me_Amy]

That is true It has been a long time since you have talked about revoking your API key

[@morena-jackson]

Great video Matt I love these kind of videos, I follow along from start to finish.

[@MrKasper0001]

Yes I love the tutorial style videos especially involving open source projects that can be ran locally.

[@DavidROliver]

Thank you, Matthew. This tutorial clearly shows how easy it is to implement agentic workflows in production environments. I've created a demo that I will document and share. In this demo, I use an agent to inspect a CSV file for errors. The agent then queries APIs to find the correct data and updates the file accordingly. Finally, it performs an SFTP transfer to another location, which depends on the content of the file. I believe these labor-saving examples really demonstrate the power and usefulness of this technology. While humans can perform many tasks that agents can do, I would argue that they may not do them as effectively over an extended period. Therefore, I believe we should leverage AI to benefit everyone by assigning it the jobs that are challenging for us, rather than those we excel at.

[@rtwg605]

+1 to more tutorials. Long, detailed, too.

[@lyb-livingyourbest124]

Cline is amazing, I built a full AI Blockchain project and it did everything from start to finish. It's so amazing.

[@titombo]

You have it open source? I am curious how ir works

[@lyb-livingyourbest124]

​@@titombo I'll open source it soon

[@joaquincastellanos3411]

Wonderful tutorial, The whole series was very useful, Thanks!. Hopefully you will be able to make more 😉

[@wurstelei1356]

Nice tutorial. Interesting would be, how a local LLaMa3 instruct crew perform on your LLM tests.

[@MetaphoricMinds]

This isn't just a video. It's a historical document.

[@damienrhughes]

Explain?

[@VadimBesedin]

Thanks, looks very useful. But what about security of your data? You doing your research, tweaking your AI/Agent system, getting your experience and insights, and it all gets passed to some third-party service?

[@jimbo2112]

Great vid. Clear and simple. I'll go have a go myself now/ 👍

[@TropicalCoder]

How many agents does it take to screw in a light bulb?

[@leonkowalski6987]

I watched your previous video where you were having issues with environments. I notice here rather than created a new environment, you're just installing everything in base. I tried many times to make it work inside a dedicated conda env, but couldn't get it to work and the summary explanation in the previous video was tricky to follow. Do you know what the issue is, and is this why you are using base? Thanks!

[@tobiasm161]

Have you also planed to go in-depth with LangGraph as well? The more complex way then crewAI

[@RadThings]

What’s the difference between langtrace vs langsmith? Is it just a competitor of langsmith for LLM observability?

[@Its_Zeus_]

yes

[@reachkkarthik]

Langtrace is open source and open telemetry based.

[@matt.stevick]

thx matthew b.

[@muraliytm3316]

I didnt got that selections to select a provider, I have installed the crewai as mentioned in the video, how can I fix that

[@sushmapradeep7509]

I have a crew project where crew is supposed to call Agents with user message. Agents need to extract information from the user message based on given description & expected output in corresponding task. But it is like Agent is totally ignoring those instructions. What am I doing wrong? Can someone help me with this?

[@rinaldspikse3519]

unfortunatelly , whatever i do, langtrace does not run after installation. I am running a conda environment, langtrace-python-sdk is installed, but after runnin crewai i get the error: ModuleNotFoundError: No module named 'langtrace'. nothing seems to help. :(

[@reachkkarthik]

Hi could you add Langtrace as a dependency to the pyproject.toml and try again

[@worldbridgerone]

try a docker env. or a fresh new linux dev machine even running from a usb

[@95jack44]

Go play with Yacana instead. It's only a "pip install yacana" and it does actualy work out of the box! (local LLMs only)

[@stolenhandle]

@@95jack44 Can you please provide me with two or three use cases for Yacana? I see you're a fan but it's not clear why I would use it.

[@chrispharoah]

Make sure you add "langtrace_python_sdk>=3.3" to your pyproject.toml

[@iyerasri]

Simple question guys. Where do you install crewai using pip, (pip install crewai)? is it in the base conda or your system Python3? The reason I am asking this basic question is, when I create a virtual environment using conda or poetry or simple venv, then install crew over there and then when I run crewai install, it creates another virtual environment inside this current virtual environment. Then finally when I run crew run, it bombs with this error message - An error occurred while running the crew: Command '['uv', 'run', 'run_crew']' returned non-zero exit status 1. Any guidance on setting up crew cleanly on your PC/Mac is highly appreciated.

[@fuzzylop]

Hi there! I'm encountering an issue when trying to use crewai alongside the langtrace_python_sdk. Everything works fine until I add the import statement: "from langtrace_python_sdk import langtrace" When I run the crew, I get the following error:"... ModuleNotFoundError: No module named 'langtrace_python_sdk' An error occurred while running the crew: Command '['uv', 'run', 'run_crew']' returned non-zero exit status 1." I've double-checked my environment setup and package installations, but I might be missing something. Has anyone else experienced this, or does anyone have guidance on resolving it? Also, to Matt-great tutorial! It’s been super helpful so far. Thanks in advance for the help!

[@dungus-bi6kd]

I'm so confused What is it used for?

[@hipotures]

Observability. I personally use Langfuse (locally).

[@Ps5GamerUk]

Blowing the bubble up just that little big bigger

[@worldbridgerone]

@@Ps5GamerUk looking at the bubble's air chemistry inside before it pops

[@vpakarinen]

He's explaining and showing that.

[@rijnhartman8549]

you should create Notion pages for these tutorials

[@donutrangerr]

Actually how do you ACTUALLY put into production and waht's the best way, APIs?

[@pseudonym9667]

You keep pushing CrewAI, I feel like I haven't seen good use cases. Knowing how to get it working is great. But what are working examples of where it's actually useful? I just can't see where it's useful; I lack the imagination, help me

[@95jack44]

You can't find any good ideas because crewAI doesn't work that well. Go have fun with Yacana and its agents. You'll have something working in minutes and it will give you ideas! ^^ There's plenty to build!

[@DeianGiNet]

can u have agents with different models? say 4o programmer and claude tester? can you give us a demo? :)

[@lule-ahmed]

Yeah

[@zakariaabderrahmanesadelao3048]

you pass the LLM as an argument when you define your CrewAI agent like the example below : @CrewBase class EduResearchCrew(): """EduResearch crew""" @agent def researcher(self) -> Agent: return Agent( config=self.agents_config['researcher'], llm=llm, verbose=True, tools=[SerperDevTool()] ) @agent def planner(self) -> Agent: return Agent( config=self.agents_config['planner'], llm=llm, verbose=True )

[@A_Me_Amy]

You know how like a kid you asked them like why did they do that or what was going through their mind and they're like I don't know it's kind of a lot like GPT40 you know not actually thinking about things and then it grows up or the child grows up and begins to actually have internal thoughts about what it's doing I really don't feel like there's any difference between AI and how it thinks and humans and how they think... I think the only real difference would be that you know AI gets to think about and be introduced to everything on the internet or that it's trained on really fast whereas essentially you know a human 4-year-old only gets trained on and what its parents allow them to be in front of. I really do genuinely don't see a difference between AI it's consciousness and humans and their consciousness as far as how they think but what they think about is different. Just like a child thinks about things like you know GPT3 and two and like a preteen or teens like four and four oh and then like you know I very young adult is like a one and then like a you know but it's just about the tools that are being built into how you think just like with a child how they think self-reflection considering what was I thinking when I did this thing instead of just doing that thing

[@user---------]

I didn't understand anything and even for what it was?

[@thelalomorales]

the tutorials are getting shorter ......... because the tech is getting better. . . CREWAI for the wiinnn!! .. GO JON JONES!

[@A_Me_Amy]

I just feel like Nvidia rules the world but they're about to get a hell of a lot cheaper due to competition well or not really competition but alternatives and also less you know give me $10,000 for a basic card will make it $5,000 for a car that's twice as powerful in a year and then we'll make it you know $200 eventually for a card that's like 10 times as powerful soon enough I am pretty sure so then that way everybody on their smartphones will have a you know a 10 trillion model or you know

[@MichielvanderBlonk]

Ever heard of punctuation?

[@adamholter1884]

First