Auto-GPT Tutorial - Create Your Personal AI Assistant 🦾
Вставка
- Опубліковано 12 лип 2024
- In this tutorial, I will show you how to set up Auto-GPT and get started with your own AI assistant! Auto-GPT is a pioneering open-source software that demonstrates the potential of the GPT-4 language model. Powered by GPT-4, this application autonomously connects LLM "thoughts" to accomplish any objective you assign. As an initial instance of GPT-4 operating completely independently, Auto-GPT expands the horizons of AI capabilities.
Update from Auto-GPT's GitHub: We've improved our workflow. master will often be in a broken state. Download the latest stable release here: github.com/Torantulino/Auto-G...
Build your own Auto-GPT: • Build Your Own Auto-GP...
👉🏻 Links
platform.openai.com/
github.com/Torantulino/Auto-GPT
beta.elevenlabs.io/
app.pinecone.io/
⏱️ Timestamps
00:00 Introduction
00:24 What is Auto-GPT
02:28 Getting started
04:50 Cloning the repository
07:20 Setting environment variables
10:03 Running the script
11:24 Setting up the assistant
17:49 Speech mode
👋🏻 About Me
Hey there, my name is @daveebbelaar and I work as a freelance data scientist and coach. You've stumbled upon my UA-cam channel, where I give away all my secrets when it comes to working with data. I'm not here to sell you any data course - everything you need is right here on UA-cam. Making videos is my passion, and I've been doing it for 18 years. I just genuinely enjoy helping people.
While I don't sell any data courses, I do offer a coaching program for data professionals looking to start their own freelance business. If that sounds like you, head over to www.datalumina.io/ to learn more about working with me and kick-starting your freelance career. - Наука та технологія
👉🏻UPDATE: Auto-GPT has changed a lot since I've published this video. Check out the latest information here: docs.agpt.co/setup/
that makes sense. This took me 5 hours just to set up properly. Not a quick learn at all and I'm exhausted haha
I would pin your msg :)
im having issue with code. not found error can anyone help me ?
@@ApoAbi Please write the whole message/warning :)
Right now, no one knows when, how and why 😅
@@beastfr0meast93 Thanks for the reminder. I thought I did that! 😅
You are so right about the future. Great work on this AI video.
There's really nothing better than someone teaching you something by flying through it incomprehensibly. Thanks man.
This comment made me laugh my ass off for real. Even though I disagree, I’m a middle aged man with practically zero computer experience and I was able to do this.
What do you need help with? 😃
@@daveebbelaar nothing right now I figured it out. I still can’t get the Google API and location right and I need to double check the voice, it’s cause I’m using Pycharm and I don’t think I’m putting the values either in correctly, or I screwed up the format, maybe I need quotations for everything? And then saving it saying I can’t save a .env over an existing one, which I wasn’t sure if I should rename it entirely
@@Arisbe_obasmoqweef how can I open the pycharm from Mac terminal, the code . doesn't seem to work
very comprehensible and helpful imo.
I appreciate the walk-through, very clear and information.
AutoGPT is crazy!!! I imagine that soon we will all have a cool robot at home, based on AGI
This is nuts. I can not wait to get this up and running. Trying to figure out running it on a VM.
loving your passion to create videos. tqvm. subscribed.
Thanks Sufian 🙏🏻
Thank you. Very informative. wow. Your videos get me right to the front of the AI queue.
Glad you like them!
Deze video is geweldig Dave 👍🏿
Thanks for the ElevenLabs tutorial much appreciated.
I think that is the coolest part haha
🤯 great tutorial
Full autonomous data science projects soon haha!
THIS IS INSANE!
Awesome man, you got me started, thank you so much!
Glad I could help!
you also should add a google api or you'll see a lot of errors and your research may go slower. worked much cleaner for me once I did that.
I asked it to create the ultimate burger and it provided 3 separate outputs for the toppings, flavors and the recipe. very cool.
Thanks for the tip, I will try that as well.
❤
Ll
Where do I find the google API?
How to add that brother, would you be kind enough to explain
Awesome. I am imagining where this tech will be in a few years. It seems it would be possible to send an AI avatar to attend zoom meeting on your behalf and to take notes from the meeting for your to read later.
gotta have your research AI read it later and prepare a presentation for your presentation viewing AI.
Out of curiosity can this be utilized to set up reminders? I'm trying to find an AI option that will let me speak to set reminders and have it remind me via audio on desktop. I'm hoping to utilize GPT to add on Google/Bing searching via voice query as well as 'research' tasks similar to what you're doing in this video. I'm beginning to suspect trying to cobble it all together is going to make me learn Python, however... -_-
Great video Dave! I’m about to go to sleep but am going to follow along with your video and setup autogpt tomorrie. Just fyi, at time 11:30 I think you meant to say “at your service” instead of surface.
Thanks Annette! Yea, that's just my Dutch accent kicking in haha
An alternative to pinecone is Redis
Super useful 👌
Damn this is crazy. Updating my existential risk predictions 🙃
I can't keep up 😅
@@daveebbelaar This is crazy! I'm glad I'm not the only one who can't keep up but it doesn't eliminate how I feel about it all. It's like a rat race and we are being pushed to try a tool for a few days and then a new one comes and we switch to it. Looks like there is no end...
@@asieharati Yep and it looks like that's going to continue for a while 😅
@@asieharati lol I feel like I am in a foreign country, all new terms and words….programmed back in the late 70s….a whole
New world! Exciting
Victoria
This guy deserves millions of subscribers mate waar was je al die tijd lol
he does!
Haha thanks man! 🙏🏻
Thank you
Thank you so much for this very interesting video. Learned a lot! I am so curious about what people will end up doing with this. I am going to use this for my career in software, hopefully, people will use this for good.
Thanks Yan! I am curious as well 😃
@@daveebbelaar You are refering to GPT4. But I am still on the waiting list. Is it possible to do the same with gpt3 or 3.5? .. usig the API key for 3 and followingyou instruction to use it a similar way?
@@michaspringphul Yes, that is possible. You should change SMART_LLM_MODEL in the .env file to gpt-3.5.
@@daveebbelaar thank you :-)
hey, I do all you steps but when I clone the repository I dont have all those scripts that you have in the scripts folder.. what I am doing wrong? I tried master and stable and it seems all broken :(
Awesome tutorial!! I'd love to see a "Jarvis" assistant with voice
dude, thats awesome, all it needs now is image / video generation and you could have AI assistants, AI security with it setup with tensorflow, all kinds of crazy things
just think of entertainment content generation. Example, I want the AI to check my steam account and wow account data to see what types of games and content I enjoy and create a game that employs all of those attributes then also support the game with expansions as my gaming tastes change over time
Yea there are so many possibilities! And this is just the beginning 🚀
when you create a request, is the token used only for that request, or for each one when gpt interacts with each other to solve this task?
3 seconds in and it’s already crazy lol
Yep, this is weird haha
This sounds like a navi from Megaman Battle Network lol. Yeah we will each have a personalized intelligent agent in the not too distant future. :D
That is genius!.
So many opportunities here 😄
@@daveebbelaar Yes. There really are.
@daveebbelaar interesting video,I have a question, how to make the outputs run in a Open API URL or different front end instead of running on terminal?
Dude, did we use the same robot chick in our thumbnails?? 😂
Awesome video by the way.
They are definitely related haha 😂
I'm keeping an eye on Auto-gpt .. and waiting until a few more bugs are sorted .. I read an article saying they paid $14 for a complex inquiry, and a bit later part of the process was needing this inquiry aging but Auto-gpt redid the inquiry when this should have been stored in some way .. but it's exciting where this is going
They made the calculations based on purely GPT-4 token prices afaik (also saw that article today), but AutoGPT doesn't only use GPT-4. It uses way cheaper models for tasks where GPT4 isn't needed. While the redoing is true, that doesn't mean there are no solutions. It all depends on what you are cappable of doing, or how fast somebody else does it
@@Patrick_Haffel yes thx .. I'm sure they're going to sort out these bugs .. if they haven't already
@@Patrick_Haffel With the latest Version you're able to run it with python -m autogpt --gpt3only
Well done sir? I was thinking about trying to clone my with tortoise and use balabolka for my read back agent.
Thank you, Dave, now I have chronic insomnia 🤦🏻 no, but seriously that's frightening in some way. How to trust the output and amount of things people will create, especially for those who are learning? So, we basically become "the supervisor" for AI?!
Yea this is getting weird. I guess "Managing Autonomous Agents" will be a skill we all have to learn.
@@daveebbelaarFantastic content! I didn't know about your channel, but now I'm your fan! Speaking about autonomous agents and agents in general, do you have some introductory materials you could share about these topics? Many thanks!
@@jeffsanaraujo Thanks! I would suggest checking out my latest video: ua-cam.com/video/NYSWn1ipbgg/v-deo.html
@@daveebbelaar Many thanks Dave! Very helpful!
For those who want to enable voice command. as mentioned in vedio generate you elevenkey and than execute it. from data find voice_id field and paste it in env file.
After this run python -m autogpt --speak
Dave Ebbelaar, I was following your projects regarding machine learning tutorial espacially the fitness tracker, very informativ. I would a request a tutorial for a project based on NLP. If you like I can join the project as a NLP Engineer.
Thanks, will consider looking into an NLP project!
Hi nice video, I'd like to use it to find influencers/affiliates but have not succeeded yet. Would be great if you can make a tutorial.
Is it possible to connect auto gpt to a local database like alpaca 7b-native enhanced?
@daveebbelaar I'd like to edit the Pinecode API, and the the Google API. Is there a way to update those through the Powershell app? I'm sorry, I'm new at this and don't know the correct way to articulate with the correct vernacular. Thanks
Thanks for the very insightful and helpful tutorial. I managed to get it work but I'm just having one error when I try to have it save the file and the error message goes: SYSTEM: Command write_to_file returned: Error: 'PosixPath' object has no attribute 'is_relative_to' - I'm running anaconda like you, and its python ver 3.9.6
cool video bro
Ty! 🙏🏻
Is it possible to install this on a smartphone android or apple device?
I tried to write today's date into answers.json file but it has failed even though once it said it did and verified it. Not sure where it writes them to but answrs.json is still empty. Any thoughts? Complained about writing to outside working directory but I gave it full path too for the working directory
I'm all in on several trends that I know of re: AGI, but being sight impaired, and focused on high level transcendent thought (that is the idea..) I need a focused AGI assistant to do this install and basically pull in the universe to my shoe size. So - any advice to that will be well attended, I'm sure..
love the vision mate!!! im hoping to create something like this too... been working on apps for the last couple weeks and the ever evolving AI and tech are amazzzzzzing!!! but also can be a bit too much to quick for me to process all of it.. i feel like using it to help evolve us spiritually will be an amazing use ... there is so much possibility i see now that I feel like i just gotta let it go... "seek first the King/Queendom and all else will be added unto you" peace
@@jeffdgilbert 2023-05-13-Saturday -
Here come the judges.
UA-cam.com/@REVV19
i cant beleve this simple script is making so many waves. And yes it's a simple script..... i've read the whole thing.
I agree, under the hood it is still relatively simple. I think what's causing the hype right now is the idea of "autonomous agents" as the next natural progression from the GPT models.
I think it's because it's just the beginning and we can see the potential. I'm having GPT4 give me new scripts to input and I can create assistants for writing a novel with zero plagiarism and sentiment analysis, webscraping and even some better speech recognition stuff thrown in there lol. The possibilities are a lot of fun and I've never done anything like this before in my life, so I can't imagine what someone who actually knows what they are doing is able to accomplish.
With a lot of prompting I've got it to produce some good stock data
Awesome tutorial as always :). I managed to get it running but questions which come into my mind are. 1. Is it possible to get a nice looking gui for chatting with the AI? Working via the terminal is not so nice. Thanks
Thanks! Not right now, but the project is fully open-source and new updates are released every day. I am pretty sure someone is working on this as we speak!
@@bandui4021 I get the "Warning: Failed to parse AI output, attempting to fix." error also all the time, but it seems to continue just fine. It's just a parsing error of the text in the terminal. I am not sure about the JSON file. That was acting weird for me as well. You try to set it up via Pinecone. You can create a free account and get the API key. Just update the .env file with your key and change memory from "local" to "pinecone".
@@daveebbelaar this invalid json error is also driving me mad and there's a lot of reports on it as well. I was surprised you got it to save an output unto the directory, thanks for this. If you figure out a fix for the invalid json error, please share that as well!
@@IdrisImi I am using the pincone memory right now. Also make sure to update to the latest stable version (see the GitHub).
@@bandui4021 he had the same problem and fixed it in the video
Since Microsoft invested so much mula in chat gpt. I'm saving up money for Black Friday/ Cyber Monday sales to invest in tech like a huge screen and bad ass windows computer. I can't wait for the future.
Does anyone know another way out, since pincone has problems with new indices?
Hey, i have a problem:
Im a beginner, but i want to try AutoGPT for sure. So i watched your video, installed all the stuff, got my environment running but when It comes to start AutoGPT with the “python scripts/main.py” command, it wont work. The terminal is telling me that there is no “main.py” file in the scripts directory. I mean thats correct, “main.py” is a directory locatedat the same directory as scripts, isn’t it? So what am i supposed to do now ? May the problem simply be that i haven’t gave openAI an payment method yet ? I would appreciate some help :D thanks !
With the latest stable version, you now have to use `python -m autogpt` to start running the script. Are you using that?
Today I did successfully start autogpt with the command you told me. Thanks so much, i wouldn’t have done it without your tutorial and the tip You gave me personally! Greetings from Germany 😊
@@maxg.2689 Glad I could help Max, Thanks!
Hey can autogpt create multiple agents working collaboratively, I mean 2-3 agents do the same goal but different ways and feedback each other to complete the goal in unbiased way?
Thats a worthy goal, I can see more reasons why this is possible than not possible, maybe set this as the goal and pair up with a few folk
I followed each step but terminal seems to not go beyond "thinking" after continue step (y/n). Any thoughts?
Try changing the goals maybe.
How do I go about hiring someone to set my auto GPT on my computer for Lead scrap, daily email, Facebook add, funnel on CRM. What softwares needed? How much!
What tasks have you had it succed with? I've set this up but it has this habit of going off the rails and not doing what was asked or ending up in these weird loops.
I am getting mixed results as well. Sometimes it fails to write to files and gets stuck in a loop, like you mentioned. I have had some successful markdown exports after looking up information, but it still has a long way to go.
is it using GPT4? how much in dollars did it cost more less to create these examples?
EDIT: solved. had to change a setting in the .env file as I am still on the waitlist for the gpt-4 api
Nice vid! Even as a no-coder I am having lots of fun with (chat)GPT. I do have a question though, I keep running into the following error:
Command evaluete_code returned: Error: The model: 'gpt-4' does not exist
I assume it has to do with the installation (as said by ChatGPT) but I am not sure how to solve/counter this. Do you have any ideas?
Thanks Rodney!
Yeah - you could in early stage run API's without a credit plan
With the latest version one can run 'python -m autogpt --gpt3only' and it will work without haven a gpt-4 API key.
Haha great so see the code I added appear in a 100k+ video
how does voice_ID works from elevanlab. Is that for authentication only? Or can I use it to input my commands into auto-gpt?
With the VOICE_ID you can select which voice you want to use.
so it is possible to add the voice from Gideon, barry allen AI assistant. i always dreamed of having an AI like her. she also appears on legends of tomorrow.
It's not actually "your own" assistant until you run all the components locally but using them from internet via API :) it's all possible now with open source LLMs
Whenever I try to save it to a file I get the error "Command write_to_file returned: Error: 'WindowsPath' object has no attribute 'is_relative_to'" from autoGPT
are you using openai api 4 keys?
Great vid man. For the life of me, i cant figure out why --speak command wont work. just keep getting Errno 2: no such file or directory. but the file is definitely there lol
Thanks! Are you using ElevenLabs on Mac? Check out this: github.com/Significant-Gravitas/Auto-GPT/issues/856#issuecomment-1506442263
i couldn't use a python virtual environment because the scripts Auto-GPT created wouldn't run. I had to enable virtualization on my machine (windows) and download with docker.
Hey! I'm trying to follow your tutorial and I got an error after typing the commend here 7:01
It's a long error but it starts like error: subprocess exited with error , could you maybe help?😊
My God, so many things we need to install or get before using this auto Gpt thing. Excuse me if I’m overwhelmed. I’ll finish the video, but it makes me feel comfortable just using chat GPT for the next while. Hopefully in a few years, they’ll simplify the process of accessing this other GPT thing.
Hey Dave, is it necessary to have knowledge of Python to create this? I don't have a coding background, so is that possible?
Hi Anurag, it is possible to set it up without, but it helps if you understand a little bit of python. More specifically, how to properly set up an environment and execute code.
@@daveebbelaar thanks Dave. I get your point and it makes sense.
You completely lost me at pip and Conda environment at 5:47min into the video. I used Chat GPT 4 to help me configure my phyton environment, but I'm not quite sure where to resume the install process watching this video.
You know you can preview markdown files directly in VS Code ?
Yep, but I use Craft for my videos, so I wanted to see what it looked like there ;)
After selecting "y" all transactions are completed and indicates what to type to activate auto-gpt. I get this error:CommandNotFoundError: Your shell has not been properly configured to use 'conda activate'. Please help, I have limited experience with this.
Let the real matrix begin, Can we create a team of agents to work together and retain their memories..this is so sick~
I tried AutoGPT and asked it to develop a Shiny app. After half an hour "thinking" and "Googling", AutoGPT said that that the R codes has been witten to an R file. There was an R file, but only one line "" in the file, no actual R code at all! 😭 Any tips to make AutoGPT work?
I tried same question with ChaGPT before using AutoGPT. ChaGPT gave me full R code (immediately) with right Shiny app structure, well defined input variables and calculation. Although there were key errors in the R codes, but was a good draft version.
It is still a bit buggy when writing to files. You never really know what you are going to get.
i followed your instructions several times, I couldn't get ai to work, I then brought it to a local computer store, they could not make docker work as you explain it. Does this have to be done with apple machine. I thought you might have the way it's been several days now. can i call you ?
Twice in nice! in subbed too! best set up video yet.
Thanks! 🙏🏻
when I try
conda create -n auto-gpt python=3.10
I just get an invalid synthx error
Dont we need to pay to use the gpt3.5 api?
How do i save search result in auto gpt workspace ? Nothing saved what i searched
It's possible to change language?
solution for not pay api ?
7:14 I get stuck here since mine says "ERROR: Could not open requirements file: [Errno 2] No such file or directory: 'requirements.txt'" Despite the txt file being in the folder.
Are you in the correct folder in your terminal? Type `ls` and make sure you see requirements.txt there.
I get "Python: can't open file 'scripts/main.py': [Errno 2] No such file or directory" when i try to run. Please help.
python -m autogpt
It is actually in very early stages, it often gets stuck on research loops and doesn't complete the task
Yea, I have experienced this as well.
I've run this a number of times; but, I have yet to see it actually complete anything. It always runs into a loop for me.
why no env file in the new stable version of autogpt, I had to create my own .env file. Is that normal?
It seems that that one is missing from the stable version. You can download/use it from the master version.
i dont know if anyone else has this problem when the files are different, so i can´t find the main.py
can I get this on my Android device?
Is this something anyone can do or do you need to learn code?
I can see "Detroit Become Human" becoming a reality. Well I won't mind it if we get the good ending haha
Can you make a portable version of it?
You could set this up on your personal computer or on a virtual machine in the cloud with services like Azure or AWS, then establish a remote desktop connection with your phone and use it whenever you want from there. You can use this from any machine really with a RDC
So can I hook this up to a robot and have it keep working till it organizes my house
What you think it'll do:
- make you money
- take over the world
- work for you while you chill
What it'll actually do:
- hit usage limits and/or drain your bank account faster than you can say "I wished GPT-4 wouldn't cost 8 cents per 1000 tokens"
One day haha
Hi! I had a issue that when i was running pip install -r requirements.txt in terminal, then it showed zsh: command not found: pip. Why? and how can i fix this. Thankss
Is your python installation set up correctly? What and do you use venv or conda?
Have a look at this also: stackoverflow.com/questions/42870537/zsh-command-cannot-found-pip
@@daveebbelaar I just checked that I use venv, so what can I do about it or adjust it into conda?
@@longtrinh7857 You can first try to run pip3 install -r requirements.txt
@@daveebbelaar thanks a lot! hope you'll reach your sucess!
How I would use this, and I think it might be possible, it’s to help me with projects. For example, if I have a commented PDF, I will want it to work on the comments that I’ve put on the PDF, as prompts. I would like to be able to focus the AI to accept, or restrict to certain parts of the web. So this would basically be a research assistant for me. Taking in my notes from a PDF, and then researching those notes and ideas with relevant quotes and citations to support those ideas.
That's a really good idea! Should eventually be possible 🙌🏻
@@daveebbelaar Yes, I saw a PDF reader based on ChatGPT. Thanks for your tutorial. I will try on the my mac too. I do not know those programming languages as well as you.
@@daveebbelaar LangChain can read multiple pdfs,
I followed your video but the script to activate it doesn’t work for me. It tells me I have to run Python -m autogpt for some reason? If I run that script then it ends up activating. But the one you ran doesn’t work for me? Thoughts?
Do you use Anaconda or a regular Python installation?
Apparently, in the latest version of Auto-GPT, you run the script with `python -m autogpt`. Please see the pinned comment on this video for updates.
@@daveebbelaar Thanks for this. I redid everything using Python as I couldn't figure it out with Anaconda. I did what you showed about creating the Json file, but it's still telling me this: Attempting to fix JSON by finding outermost brackets? It will say it's fixed, but then every few prompts pop back up with this
How does it work being limited 25 queries for every 3 hrs
Hi, can i run this script on windows vps ?
That should be possible.
I keep getting a "cannot import main" error. Even though, everything seems to be in place.
With the latest stable version, you now have to use `python -m autogpt` to start running the script. Are you using that?
code does not seem to run ?
Stuck at 10:26 - get the error /AutoGPT Experiments/Auto-GPT/scripts/main.py': [Errno 2] No such file or directory. Which is true because that file doesn't exist in the folder and not on the git repo either. I'm stumped
Did you solve it ? I got the same error.
@@HearAIStory I was up to 2:30am (formatting the Mac) last night but I think solution is to download the same release as in the video. Looking on GIT it has the correct files
python -m autogpt