Introducing DEVIKA - OpenSource AI Software Engineer | Local Install
HTML-код
- Опубликовано: 28 май 2024
- In this video, we will look at Devika, the open-source alternative to "Devin", the first AI engineer from Cognition Labs. You will learn how to set it up locally and how to get started with both local and proprietery LLMs.
Enjoy :)
🦾 Discord: / discord
☕ Buy me a Coffee: ko-fi.com/promptengineering
|🔴 Patreon: / promptengineering
💼Consulting: calendly.com/engineerprompt/c...
📧 Business Contact: engineerprompt@gmail.com
Become Member: tinyurl.com/y5h28s6h
💻 Pre-configured localGPT VM: bit.ly/localGPT (use Code: PromptEngineering for 50% off).
Signup for Advanced RAG:
tally.so/r/3y9bb0
LINKS:
Devika: github.com/stitionai/devika
PR mentioned in the video: github.com/nalaso/devika/tree...
TIMESTAMPS:
[00:00] Project Devika
[01:57] Installation and Configuration
[05:20] Running Devika on VS Code
[11:55] Experimenting with Different Models and Observations
All Interesting Videos:
Everything LangChain: • LangChain
Everything LLM: • Large Language Models
Everything Midjourney: • MidJourney Tutorials
AI Image Generation: • AI Image Generation Tu... 
Наука
Devika developer releasing the first version in 20 hours of coding is mind blowing. Still cared enough to make it fully open source. But pls give it more complex task.
Without Devin et al, Claude 3 Opus wrote a THREE D version of 2D Life for ,me ... up and running in under 15 minutes.
Would you care to explain furthermore?
@@sisyphs he probably means that he got a stupid code out of claude3 only without generating it using Devika
Yeah it's getting frustrating keeping up with these companies one upping each other. Gpt 4 is now outdated again, Claude opus has the most up to date info and best responses. Before that it was Gemini pro 1.5, which is still great with the 1 million context
@@aloktripathi27 Stupid code? Let's see YOU write a 3D Life program in 15 minutes.
He's likely shitting his pants watching his job loss instead of accepting and excelling himself @@coldlyanalytical1351
test it by giving it a complicated project instead of something that can be written in a single file with like 200 lines of code
Exactly. Local LLM operations suck. Only the expensive GPT-4 seems to "work."
Another great and early choice of what to cover. Thank you. We need an Open Source Software Engineer to fulfill the promises of Devin. Nice work. Of course I would prefer that it work with LM Studio. We shall see if they make this addition.
Thank you, these are really early days and hopefully in near future we will have these agents which are easier to use and are more capable
Exploring the art of storytelling and video creation. With VideoGPT in my toolbox, my content now shines with a subtle yet impactful professional touch.
what content ?
Need to see an actual use case not a overly simplified one anyone can do themselves without an LLM chatbot. We need to see the power of this thing. Have it make an actual game. Lets call it a DOOM clone.
This is something you do in computer sciences courses as a test/for fun. If its half as good as its been hyped, it should do this easy.
Thanks for the video. As of now Devika / Devin is not going to replace software engineers. I can sleep well now :)
Very cool, thanks!
Devin loves devika ❤
Does this only use a single model? Isn't the idea of multiple agents to use specific models for each task and predefined functions for things it already knows? Liked and subscribed.
Thanks for sharing
How do i point this to my ollama docker container?
Nice one, and seems very useful. Can you train Devika to learn how to code for smart contracts using solidity? How to solve bugs in solidity code?
Sir in Devika, The select button of search engine and Select model is not working ,Initially it was working but from last 4-5 days it is not working Please help me
Hello
I need help 🥹
I have done devika's setup and i provided all api keys. But when i run devika and i prompt it to do something after that if just showing echo 'waiting..." In terminal and after restart server it's not disappearing + in backend it's continuously sending get and post requests... (I don't have much knowledge in this and I don't even know how flask works im just a beginner and done setup from instruction) pls help me to solve this 🥹
It seems to be running the back end but nothing is getting output in the front end. Could someone help?
I couldn’t get it to run at all so I hope this video helps.
Can do very basic coding for simple programs but using Llama 3 it can't take on to much without getting confused. Ama zing how only a couple of years in how intuitive it all is in its aim to be autonomous. programmers jobs are safe for now, but can see where this is heading with time and iteration..
looks promising
what is conda ? doesn't work for me, also Ive made cd desktop instead of cd devika and now I can't do cd devika I don't know what to do
Can I use it with LM Studio server?
Yes yu should be able to set that in the Config.toml file as normally.
Could you pretty please do a comparison video comparing the abilities and features of Aider vs Open Devin vs Devika : )
that's a good idea, will look into it.
python3 can't install flask ... 3.10.7 . Conda env not help because of UV maybe ... ohh not working for me...
Thx
Mine doesn't work with ollama it API/sends []
Princeton NLP SWE seems to be near on par to Devin. Can you make a video on that too. Pleeeeease.
Check the latest video on the channel
oh a good initial step huh? Did you bother looking at why it was always locked up? Like was it just spending hundreds of dollars in an infinite loop?
um.... is it supposed to work or just repeatedly suck?
Whoever came up with the name had a brilliant idea. If this project succeed Software Engineers will need to start dating "Devika" 🙂 . She will have a lot of Software Engineer boy friends 🙂 🙂
My kingdom for a batch file...
I am Devika too
Bun for windows is not working for me... 😢
Try with wsl
Does DEVIKA comply with the #VedicThreeFoldAlgorithm?
Devika if ethical should ethically integrate with the ThreeFoldAlgorithm in True
Wow
I wish LM studio could be used as docker container
Have u checked Open Devin ?
What should I learn to become an expert prompts engineer
Prompt engineering course/ tutorial
Search on RUclips
Using Anthropic Meta Prompt Generator
Try stuff, a lot
just keep talking to the robots
Prompt engineering is not really a thing anymore. Most major models can work with simple prompts (Hell, even SDXL is worse with complex prompts).
Well actually it did neither solve the 1 to 100 correctly (identation error) nor did it solve the game of life. Wonder how many tokens got exchanged in total.
Did you use an AI generated voice for this video, It sounds a bit like you did.
Having watched a good few of his videos, this sounds like he always does. I doubt it's AI gen'd voice.
Bro it's a Indian accent
Is she going to give lazy and sub par work and say “I can’t find that on google” when asking it to program something? Lol
At this point, why anyone would invest 4 years getting a software engineering degree is beyond me. AI is so much better at writing code and AGI will develop it's own language to write more efficient applications.
Brev 😅
Devika 😂
Devika seem copy of Devin :)
As an engineer with 40yrs exp in hardware and software, I'm getting pretty sick and tired of the hype being pushed by these AI tool fanbois. AI is incredibly good at mimicking an idiot, we see that in text 2 speech with funny mispronounciation, we see chatGPT YT videos being full of useless generalisms but utterly boring as hell and so monotone. Once we see a lot more of the code, we are going to see that the problem being solved is far simpler than real world software. The issue with software is mostly in the architecture, the implementation of the low level stuff is just a matter or refining the specification to ever lower levels.
Programming was more fun 40 years ago when tools started to get dramatically better like say LightSpeed C, today I can not get my head around XCode or Visual Studio as both are hugely bigger than a decent OS like say HaikuOS and they don't help me one little bit. I would prefer to use a 32b version of Metrowerks CodeWarrior anyday over VS, but it was taken off the market >20 years ago, and it still works too within limits.
imagine the dumb Ptoject managers thinking to replace software engineers with devika and asking it to build a end to end project
we are doomed
seems much more cluncky, slow and inefficient that simply working to openai or claude web interface
useless
devika??? yeah no thanks.
Most of your videos are just useless clickbaits
Yep in 5 years AI is going to be powerful so i think i am going to farm for work
Ai will do that better too😂
@@AFK_Quay no it cant poke its hand inside cow's bum
@@views-re2om It probably will be able to. As for the future of people...
...demand and supply...
I start producing electricity with my body, and work as a generator. I think I will have a safe future :)
@@coderentity2079 good luck brother :) btw we have Nuclear power plants for electricity
there are a lot of these programming agents, so annoying that Devin got all the hype when DevGPT and AutoGPT have been doing this since gpt3.5 came out
hey bro, can you create a new repo with updated code. Thanks
imagine asking this
He does it every video though? Its a normal question..@@brtbstn
@@HistoryIsAbsurd i think he is the new 😂😂😂😂😂 already got it and running well…bus sometimes having issues with conda environment…that’s why i asked
OR, you can make your own