It would be really nice to be able to interrupt the replay, and chat with the virtual developers, to push them in the right direction. And then let the rest of it run again. Repeat until you get exactly what you want.
You mean introduce the agents to agile programming? Give them 2 minute sprints, write new stories based on if they are on the right track or not... lol.
I plan in making an app that communicate with a back-end server allowing you to manipulate data in a database while using a graphical interface. Wish me luck!
Thanks for the clear video - this was really helpful for me to get ChatDev up and running. I found most of the games weren't too good but as proof of concept it's just amazing and a very satisfying project to try as an intro to AI. I kept getting Tenacity errors talking about ratelimit which seemed to be an AI issue. As my first time using the API I needed to add funds before the code would run beyond the roleplaying stage - after that they started building out apps and games. They were a bit hit and miss but definitely building games and apps to order which in itself just blows my mind. I know that GPT is doing all the heavy lifting but still an impressive little project to try as someone who was a noncoder. Doing it alongside Chat GPT and just feeding it back any problems with the code and giving it the transcript of this and other videos really helped me get unstuck
How much of them do we have so far? BabyAGI, AutoGPT, GPT Engineer, Aider, Open Interpreter and now this one. And all of them out of the box were only supporting OpenAI GPT models. Looks like it's some group of engineers inside of OpenAI making all those projects just to pump up sales of tokens :)
I guess we're still far from agents doing proper work. I asked it to create a simple webapp that allows you to enter and edit some things, export to csv and call some API. It created a regular TK app, allowed to enter something, didn't add dates as it should, wasn't a webapp, did not do anything towhen analyze is clicked, and basically didn't do about 85% of what it was told.
Same here. Should have created a simple website, but couldn't really achieve it. Just a simple page with navigation. It's not really what I was looking for.
I'm still looking for one of these that works with an off-the-shelf public available LLM like llama 2. I would love it to work all night but I don't want to pay open AI
This is the most difficult repository I've ever worked with in half a century lol. You're video is better than most but this whole process is chalked.. lots of missing details that really REALLY matter when it comes to files, libraries, paths etc... beware before diving into this project lol cause it will not be as easy as any of these "TUTORIALS" show it to be xD.... if I didn't have experience I wouldn't of been able to even get the ENV running lmao
HE is right through. I got it working but this video was clearly rushed. He did not even show his own example working. He provided one that came with the program..@@Levicandoit
Hey, I keep gettting stuck in the API Key part. When I paste the Key it and hit enter in anaconda it says the following: The filename, directory name, or volume label syntax is incorrect. Would really appreciate it if I can get help for this! Thank you!
Hey there, Matthew! Hope you doing fine, Thank you so much for your content and for the help! I'm a real beginner, never touched codes, and with your tips and step by step, i could config ChatDev nicelly, Only one thing, I didn't have git installed, so I needed to pause your video and search on google why I couldn't git clone. Anyway, Now I just need to understand how to put the AI workers to do their jobs 😅, I'll search on your channel and other sources. Thanks again, Wish you the very best!
Hi Matt. II can only get it to do something basic. I cant get to a certain point and then iterate until i get something that us usable. My prompt was 'create a website to upload epub,docx or pdf and allow it to be summarised or analysed by openai'
Awesome video as always Matt! Quick question. What Mac software do you use for the videos. I am an Engineer tasked with creating E-learning video for our devs and I love the way your videos are put together. Thanks in advance, Gregory
Is there a way to get the ChatDev team to prompt the user for files to reference in the code? Or can I put files in a folder that the agents will pick up and use?
Which version of gtp did it use and what are the costs? Is there a way to set limits so they don't get stuck in a loop and accumulate a ton of tokens? I used that gpt assistant extention for chrome and it got stuck looping and used up all my tokens.
I made space invaders with it using the GPT3.5 version and was the same around 13 cents. although I got what a paid for if you get what I mean, tomorrow I'll run it with GPT4
I can't seem to get my api working, It says that the filename, directory name, or volume label Syntex is incorrect Can someone help. I'm very new to any type of coding but I wanted to see if I can do this as it seems interesting.
I found a solution someone posted that seemed to work for me " @benjiq732 1: conda env config vars set OPENAI_API_KEY="PUT KEY HERE" 2: conda deactivate 3: conda activate ChatDev_conda_env"
When trying to enter my Open AI API key i get this response from the console "The filename, directory name, or volume label syntax is incorrect." Anyone able to help me out with this?
Autogen gives you a framework to build your own team, with definitions. It's much more flexible is my understanding. I'll be making a video about it soon.
Hi Matt, great video. I am getting a rate limit error. Any idea how to fix this? I have checked and I have balance in my open AI account but it wont go past the first few steps when i provide the description of my app.
Muchas gracias por el video, muy ilustrativo e intuitivo. Pero sería interesante avisar al comienzo del video que se necesita una cuenta de pago con OpenAi para poder usar estas funciones. Al menos, para los recién llegados a este mundo del IA como yo, que aún no entendemos todo lo relacionado con los tokens y cosa así. Saludos desde Chile
I ran into a couple issues on windows. First, I had to change the python command for the API key to start with SET, not $env:. Later I had issues with your prompt running the visualizer. you had the thing set for python online_log/app.py, but online_log should be visualizer instead. Other than that, it worked a charm.
Hey Matthew, love your content. I aspire to teach myself coding and work on my own opensource LLM integrations. Do you think you could post the basic hardware and software you like to use in your bio or something. I'm using VsCode and Conda and figured out a weird powershell workaround to use conda with it, but I'm uncertain if there is a better way in general.
I use pip as an installer, and pycharm as an ide. But Mathew also made a video on using LM studio, where you can download any ai you want, "that your computer can handle" and work with it like chat gpt.
Hey Mathew, this isn't working for me :( Iike your video but I wish you went more in depth and showed what it looks like when it works and explained how to setup everything. Mine won't get past the first message from the CEO, please explain how to troubleshoot this
Thanks a lot for this tutorial. Since I'm not expert I can't work the Env command for call the API. On windows it says "the syntax of the file name directory name or volume label is incorrect" I know that probably this is a dumb question, but I can't find anything on web.
@@matthew_berman there are so many production AI sites that need you to pay money because they use their openai API under the hood. I would like to be able to use my own, so those projects can keep it practically free since they don't have the API cost. (Same as filling in your key as an environment variable with chatdev) Also, why not get a way to route API calls directly to the chatgpt+ that I am already paying for anyway. Much of the API functionality chatgpt will be able to handle directly. 🤑 None of these public AI sites will go viral if they have to charge separately.
Great videos as usual. Thanks for sharing. It’s a bit clunky…. Type a description of your app in the terminal? Modified it to read a yaml file with all the info…. Ahhh! Much better.
when I get to the step to install the requirements.txt it does not find the file when I look at what was downloaded it is missing? is there another step I can use to get that file or a workaround
Keep getting this error in terminal: incremental_develop=check_bool(self.config["incremental_develop"])) KeyError: 'incremental_develop' any ideas on how to solve?
Looks fun, what about existing large project? Aider can do it, but it's very limited, with gpt3 it's pretty useless, too small window size. I think we need automatic fine tune and to put the codebase into vector store? Any news on this?
For some reason it won't get past the first message the CEO. I'm not sure why. I also noticed that it says my API Key was not found, but I entered it correctly. Not sure why this is happening.
great video! i am going to try it out this weekend! also, I am going to jump into your discord and send you a message. I have something pretty amazing my kiddo and I have been working on to show you!
Cloning into 'ChatDev'... remote: Enumerating objects: 2738, done. remote: Counting objects: 100% (314/314), done. remote: Compressing objects: 100% (240/240), done. error: RPC failed; curl 56 Recv failure: Connection was reset error: 1537 bytes of body are still expected fetch-pack: unexpected disconnect while reading sideband packet fatal: early EOF fatal: fetch-pack: invalid index-pack output [ this is the error i get every single time help pls
Hello. Good morning, can anyone help me please? It may not be the correct site, and if so I apologize. But I'm starting to learn how to program. I'm using MacBook and I only have both times " ". Does anyone know how to add single quotes? After my ESC key there are no single quotes. Can someone help me please? Thank you very much
Na maioria das vezes tanto aspas duplas:"Olá" quanto aspas simples:'Olá' tem o mesmo valor, tipo, faz a mesma coisa e na programação pelo menos, da na mesma na maioria dos casos '-'
Anaconda? - no thanks - its a monster. And it uses its own version of python, sets PATH and messes my system up. Also, I do not have a GPU on my PC so I bet this doesn't work as usual. And Visual Studio code? - nope not on my mini PC. But ChatDev does look cool from the video - good job devs.
Hey, you left us hanging, I wanted to see the todo app it came up with. 😃
Yeah I was gonna ask the same question lol. Maybe he will upload a new version
Same
Probably because it was a really disappointing result
@mehdim8808 yeah everything is "super extraordinary amazing game changer revolutionary news" here...
@@mehdim8808 ...or it was so good he's gonna sell it.
I am waiting for your tutorial for installing ChatDev, and here it is🥰 Thanks a lot!
You got it!
It would be really nice to be able to interrupt the replay, and chat with the virtual developers, to push them in the right direction. And then let the rest of it run again.
Repeat until you get exactly what you want.
You can actually use a 'human mode', act as a reviewer and guide the team
You mean introduce the agents to agile programming? Give them 2 minute sprints, write new stories based on if they are on the right track or not... lol.
@@rfphill Yes, I'm going to pretend that's exactly what I meant.
Glad I thought of it. 🙃
@@jtjames79 Project manager human mode unlocked lmao
Well, technically you can write the prompt as long as you want right?
Has anyone ever used these to actually build anything? There was a lot of hype around AutoGPT but that was useless too.
True😂
I've been asking this question. Everyone I see tutorials they are obsessed with creating flappy birds😅.... That's not gonna help me in anyway😅
I plan in making an app that communicate with a back-end server allowing you to manipulate data in a database while using a graphical interface. Wish me luck!
@@anthonyjobey8821 I do the same. it works pretty well.
@@anthonyjobey8821sounds very useful be sure to let us know how it goes
Thanks for the clear video - this was really helpful for me to get ChatDev up and running. I found most of the games weren't too good but as proof of concept it's just amazing and a very satisfying project to try as an intro to AI.
I kept getting Tenacity errors talking about ratelimit which seemed to be an AI issue. As my first time using the API I needed to add funds before the code would run beyond the roleplaying stage - after that they started building out apps and games. They were a bit hit and miss but definitely building games and apps to order which in itself just blows my mind. I know that GPT is doing all the heavy lifting but still an impressive little project to try as someone who was a noncoder. Doing it alongside Chat GPT and just feeding it back any problems with the code and giving it the transcript of this and other videos really helped me get unstuck
Very cool :) As always I’m looking forward to the day when we get an AGI or ASI version of this that can make a first class MMORPG :) :)
I'm waiting for your video on Mistral 7B. Your videos are great! Thanks a lot!
Coming soon!
delivers much higher quality with the same prompt than MetaGPT
It would be nice, if you could give them an existing code project, that they can start to build on, that they can refactor and extend.
I am working on that myself, trying with Docker and memAI
How much of them do we have so far? BabyAGI, AutoGPT, GPT Engineer, Aider, Open Interpreter and now this one. And all of them out of the box were only supporting OpenAI GPT models. Looks like it's some group of engineers inside of OpenAI making all those projects just to pump up sales of tokens :)
Exactly! I was also thinking the same!!!😅 It's a scam
Its open source, modify it to use another api. Better yet, use cgpt to modify it to use a different cheaper gpt
Great tutorial. Thank you
Glad it was helpful!
Can you bypass the OpenAI key and use a local machine?
yea i did , its very straight forward but local llm very rarely converge to anything.
I guess we're still far from agents doing proper work. I asked it to create a simple webapp that allows you to enter and edit some things, export to csv and call some API. It created a regular TK app, allowed to enter something, didn't add dates as it should, wasn't a webapp, did not do anything towhen analyze is clicked, and basically didn't do about 85% of what it was told.
Same here. Should have created a simple website, but couldn't really achieve it. Just a simple page with navigation. It's not really what I was looking for.
Hey, am really loving your channel and videos. Keep up the good work. Aman
Can you make a video on AutoGen? It's also a multi-agent framework.
Yes!
I'm still looking for one of these that works with an off-the-shelf public available LLM like llama 2. I would love it to work all night but I don't want to pay open AI
That is what I would like to
@@scitechtalktv9742 it can get expensive particularly with these open-ended repetition quarries to the AI
This is the most difficult repository I've ever worked with in half a century lol. You're video is better than most but this whole process is chalked.. lots of missing details that really REALLY matter when it comes to files, libraries, paths etc... beware before diving into this project lol cause it will not be as easy as any of these "TUTORIALS" show it to be xD.... if I didn't have experience I wouldn't of been able to even get the ENV running lmao
Maybe post a comment with helpful information that could be included instead of just bitchin in a comment?
HE is right through. I got it working but this video was clearly rushed. He did not even show his own example working. He provided one that came with the program..@@Levicandoit
How did you get the env working it wasnt working for me.
There's video now! 😊😂
Hey, I keep gettting stuck in the API Key part. When I paste the Key it and hit enter in anaconda it says the following: The filename, directory name, or volume label syntax is incorrect. Would really appreciate it if I can get help for this!
Thank you!
Me too
I partnered with the main developer for Maker DAO. He has a 5 year built API system that will scrape any data with his API system.
Hi Matt , good video , which software you use for Video recording on Mac ? i use Obs studio
Yep that’s what I use!
Hey there, Matthew!
Hope you doing fine,
Thank you so much for your content and for the help! I'm a real beginner, never touched codes, and with your tips and step by step, i could config ChatDev nicelly, Only one thing, I didn't have git installed, so I needed to pause your video and search on google why I couldn't git clone.
Anyway, Now I just need to understand how to put the AI workers to do their jobs 😅, I'll search on your channel and other sources.
Thanks again,
Wish you the very best!
Hi Matt. II can only get it to do something basic. I cant get to a certain point and then iterate until i get something that us usable. My prompt was 'create a website to upload epub,docx or pdf and allow it to be summarised or analysed by openai'
As a windows user, env:OPEN_AI_KEY didnt work. per the instructions url, set OPEN_AI_KEY=THE_API_KEY worked. Thanks for the tutorial!
I watched your video about LM Studio, could this be used with LM Studio instead of chatgpt? Point to the localhost server started with LM Studio?
Is there a way of using a different llm model, one that is running locally?
I don't believe so.
Awesome video as always Matt! Quick question. What Mac software do you use for the videos. I am an Engineer tasked with creating E-learning video for our devs and I love the way your videos are put together. Thanks in advance, Gregory
OBS for recording, Davinci Resolve for editing.
Thanks @@matthew_berman
This is cool, but I get more excited when these AI projects work with local LLMs... Want my data to be MY Data.
Agreed
open ai api calls aren't used for their data afaik, at least according to their website
Add llama models locally
Is there a way to get the ChatDev team to prompt the user for files to reference in the code? Or can I put files in a folder that the agents will pick up and use?
Which version of gtp did it use and what are the costs? Is there a way to set limits so they don't get stuck in a loop and accumulate a ton of tokens? I used that gpt assistant extention for chrome and it got stuck looping and used up all my tokens.
Seems like it used GPT3.5 and it cost about 13 cents...I also ran it 2-3 times. So it was very inexpensive.
@@matthew_berman Wow, awesome, thanks!
I made space invaders with it using the GPT3.5 version and was the same around 13 cents. although I got what a paid for if you get what I mean, tomorrow I'll run it with GPT4
@@kiwiboysl What was the GPT4 cost?
I would like to add an HR robot, review and fire the low performed code boy on weekly basis, and see how do other code boys react.
Can you give a tutorial on AutoGen?
Yep I’ve had multiple requests for this!
I can't seem to get my api working,
It says that the filename, directory name, or volume label Syntex is incorrect
Can someone help. I'm very new to any type of coding but I wanted to see if I can do this as it seems interesting.
Mine says the same.
I found a solution someone posted that seemed to work for me "
@benjiq732
1: conda env config vars set OPENAI_API_KEY="PUT KEY HERE"
2: conda deactivate
3: conda activate ChatDev_conda_env"
use this command setx OPENAI_API_KEY “”
mine also same problem
When trying to enter my Open AI API key i get this response from the console "The filename, directory name, or volume label syntax is incorrect." Anyone able to help me out with this?
I was waiting for the next snake game :)
Lol I can't tell if you guys like the snake game example or not...
Can someone explain the difference between this and Microsoft Autogen? They both seem to be fairly similar..
Autogen gives you a framework to build your own team, with definitions. It's much more flexible is my understanding. I'll be making a video about it soon.
Command for window to set api key is - set OPENAI_API_KEY=
Hi Matt, great video. I am getting a rate limit error. Any idea how to fix this? I have checked and I have balance in my open AI account but it wont go past the first few steps when i provide the description of my app.
Hi ) You fixed this ?
i have this problem
@@hurry-dentist any fix?
Muchas gracias por el video, muy ilustrativo e intuitivo. Pero sería interesante avisar al comienzo del video que se necesita una cuenta de pago con OpenAi para poder usar estas funciones. Al menos, para los recién llegados a este mundo del IA como yo, que aún no entendemos todo lo relacionado con los tokens y cosa así. Saludos desde Chile
ty for this
Matthew, your next video has to be on AutoGen. Please!
News video coming tomorrow, then Autogen!
I ran into a couple issues on windows. First, I had to change the python command for the API key to start with SET, not $env:.
Later I had issues with your prompt running the visualizer. you had the thing set for python online_log/app.py, but online_log should be visualizer instead. Other than that, it worked a charm.
Oh, my dream just almost came true🎉
Could this be used to create and manage an e-commerce website?
Create, yes. Manage, probably not.
Hey Matthew, love your content. I aspire to teach myself coding and work on my own opensource LLM integrations. Do you think you could post the basic hardware and software you like to use in your bio or something. I'm using VsCode and Conda and figured out a weird powershell workaround to use conda with it, but I'm uncertain if there is a better way in general.
Hey that's a great idea. I'll do that.
I use pip as an installer, and pycharm as an ide. But Mathew also made a video on using LM studio, where you can download any ai you want, "that your computer can handle" and work with it like chat gpt.
Hey Mathew, this isn't working for me :( Iike your video but I wish you went more in depth and showed what it looks like when it works and explained how to setup everything. Mine won't get past the first message from the CEO, please explain how to troubleshoot this
I needed to add some funds to the Open AI API account before it would start working.
Thanks a lot for this tutorial. Since I'm not expert I can't work the Env command for call the API. On windows it says "the syntax of the file name directory name or volume label is incorrect" I know that probably this is a dumb question, but I can't find anything on web.
same
Windows:
set OPENAI_API_KEY="yourkey"
There should be an openai login with aí projects just like a Google login that gets me to automatically put in my openai API key for limited use...
Not sure I understand, can you elaborate?
@@matthew_berman there are so many production AI sites that need you to pay money because they use their openai API under the hood. I would like to be able to use my own, so those projects can keep it practically free since they don't have the API cost. (Same as filling in your key as an environment variable with chatdev)
Also, why not get a way to route API calls directly to the chatgpt+ that I am already paying for anyway. Much of the API functionality chatgpt will be able to handle directly. 🤑
None of these public AI sites will go viral if they have to charge separately.
They need this for creating an Ai Rock Band
Can this use an opensource LLM instead of proprietary like OpenAI? And can you make the video using a Windows computer instead of Mac?
works great thanks
Great videos as usual. Thanks for sharing. It’s a bit clunky…. Type a description of your app in the terminal? Modified it to read a yaml file with all the info…. Ahhh! Much better.
Do I have to pay any royalties to the company for any software I make? Any type of hidden fees or costs involved to the creators of chat Dev?
No, it's open source, you use your own API keys and you pay for the usage so why would you pay royalties to a 3rd party
You use your own API keys so you just obey OpenAI’s permissions
So what are total expenses here? $99/mo for ChatGPT4 access? Then $0.14 per power usage for each app?
Hi,where is the app you generated with chatDev? Why didnt you showcased it?
We are using the openAi API key- any cost involve for using it ?
Yeah man just Google “open ai api pricing” you’ll find it. It’s not that bad but GPT-4 is way more expensive than 3.5
Yes, OpenAI API costs money.
What I would like to know is how expensive was it to develop that flappy bird example. In general, how much AI time do you use up here?
I ran this tutorial and had the team build a blockchain based web app. I got it wrong a couple times before I got it right. total cost $0.14
How much is each token and what type of coins are used. They are all AI bots not people? Facinating
Hey For Windows 10 command setx OPENAI_API_KEY "your OpenAI API key"
Mine didn't work somehow. It didnt move past the CEO's first "speech"
It would be nice if you mentioned how much it cost you, as using an openai api key is not free. :(
Good point. I haven't checked yet lol. I'll add it to the description.
can you use llama for this?
Anyone know any LLM app like this in order to create a mobile application (flutter or react native)?
bro I'm so confused how did u get the blue arrow on the (base) to AI projects
when I get to the step to install the requirements.txt it does not find the file
when I look at what was downloaded it is missing? is there another step I can use to get that file or a workaround
Is there a technical reason why these dev agents seem to only be able to code in python ?
Keep getting this error in terminal: incremental_develop=check_bool(self.config["incremental_develop"]))
KeyError: 'incremental_develop' any ideas on how to solve?
Can this be installed on RunPod, and would it be costly
Is this possible with a local LLM? with something like oobabooga or LM Studio?
Looks fun, what about existing large project? Aider can do it, but it's very limited, with gpt3 it's pretty useless, too small window size. I think we need automatic fine tune and to put the codebase into vector store? Any news on this?
gpt3 can't code, so not worth trying.
before i say goodbye to my software engineering job alert me plz
Why do you use conda not virtualenv?
For some reason it won't get past the first message the CEO. I'm not sure why. I also noticed that it says my API Key was not found, but I entered it correctly. Not sure why this is happening.
great video! i am going to try it out this weekend!
also, I am going to jump into your discord and send you a message. I have something pretty amazing my kiddo and I have been working on to show you!
awesome. and sounds good re: discord!
can it use gpt4all?
Does this work with bard api key?
I do not believe so.
existe forma de hacerlo funcionar con otro LLM?
Cloning into 'ChatDev'...
remote: Enumerating objects: 2738, done.
remote: Counting objects: 100% (314/314), done.
remote: Compressing objects: 100% (240/240), done.
error: RPC failed; curl 56 Recv failure: Connection was reset
error: 1537 bytes of body are still expected
fetch-pack: unexpected disconnect while reading sideband packet
fatal: early EOF
fatal: fetch-pack: invalid index-pack output [ this is the error i get every single time help pls
Nice one , but you need open you wallet wide 😂😂😂
It Still can't do research online
So what happens when a Chatdev game does not work? No way to interact with them
And the todo list app is still in the todo list
can it create apps in another language ? like a php web app?
Why do you use anaconda instead of pip?
Hello. Good morning, can anyone help me please?
It may not be the correct site, and if so I apologize.
But I'm starting to learn how to program.
I'm using MacBook and I only have both times " ".
Does anyone know how to add single quotes?
After my ESC key there are no single quotes.
Can someone help me please?
Thank you very much
Na maioria das vezes tanto aspas duplas:"Olá" quanto aspas simples:'Olá' tem o mesmo valor, tipo, faz a mesma coisa e na programação pelo menos, da na mesma na maioria dos casos '-'
why does everyone keep saying this is free when its not? you have to pay for the api.
yes ..that just stupid ... it should use offline llms like code llama or wizardcoder or somethig similar.
FYI the Conda commands given do not work for windows users
Try conda env config vars set OPENAI_API_KEY="replace with your api key"
After installing conda, I still get this... bash: conda: command not found
Restart your terminal.
Yeah but does the code work? How much did it cost in USD to do this?
The code was 90% there, and it cost me 5 cents.
"cd ChatDev" Command does not work :(
I get the following response: "-bash: cd: ChatDev: No such file or directory"
Don't forget to let us know what masterpieces your AI team has developed for you.
What?
@clray123 Hey! Let's not be selfish here! Don't forget to share your own masterpiece ideas too, I'm sure they would be extremely valuable!
@@Aaronius_Maximus It's not me who is hyping up this crap with each video.
my work is not in warehouse what do i do
Does someone know how better or different is from MetaGPT or others?
is it still working?
can i run this with chatgpt free subscription?
why isnt my work inside warehouse
Can they help me make a simple website?
I do not get anywhere near a working project. All 3 attempts are just terrible.
So its not free/
Anaconda? - no thanks - its a monster. And it uses its own version of python, sets PATH and messes my system up. Also, I do not have a GPU on my PC so I bet this doesn't work as usual. And Visual Studio code? - nope not on my mini PC. But ChatDev does look cool from the video - good job devs.
Concept is nice but really dont get over hiped this wont do anything more complex than tic tac to.