DeepSeek-R1 + Cline: BEST AI Coding Agent! Develop a Full-stack App Without Writing ANY Code!

[@intheworldofai]

🚨 Subscribe To The FREE Newsletter For Regular AI Updates: intheworldofai.com/

[@AbilHakim-f1g]

I use ollama + deepseek R1 and Cline... too much hallucination, better I use Windsurf

[@bmqww223]

Hei can this ai do gdscript programming too? for game dev related tasks? and does deepseekcoder have reasoning like deepseek r1? which one is better?

[@caseyarmstrong7113]

Pro tip: in lieu of an image, you can upload the image to chat GPT and have it describe it. Then use the text. :-)

[@bimoputra23]

Have you ever done it? I've thought about doing it, but I'm confused about what the prompt is like in GPT😅

[@andersberg756]

​@@bimoputra23just tell it to describe the image for someone's who wants to recreate the webpage. chatGPT and Claude van handle high level instructions like that well in my experience. So another example would be if you want to generate new images in the style of an existing image, you'd give it that task - describe the style, colors...

[@Techpeasy]

@@bimoputra23 Just mention in chatGPT, analyze the image and describe all the features present in the app, in detail. It will give you the outputs.

[@Ewakaa]

It can read images

[@peteranders6893]

@@Ewakaa I know right?

[@intheworldofai]

[Must Watch]: Srcbook: NEW AI Coding Agent That Can Generate Full-Stack Applications! (Opensource): kzbin.info/www/bejne/oHfIc4RsiJeWftUsi=V7mxeGBhJmUAWb9q Microsoft's Phi-4 14B: NEW Opensource LLM is a TINY BEAST! Beats GPT-4o! (Fully Tested): kzbin.info/www/bejne/baq2oqWHi7F8fJosi=QXJ-RmpJTPrXWECT Windsurf IDE UPDATE: AI Editor - Cursor Alternative That's FREE & LOCAL!(Autonomous, Web Commands): kzbin.info/www/bejne/jXWtoJx9Ysajobssi=DKinSFvXNbsrmSrA

[@intheworldofai]

Bolt.DIY + Deepseek-R1: Develop a Full-stack App For FREE Without Writing ANY Code! (OPENSOURCE): kzbin.info/www/bejne/pnSXn3WtfdOJpbs

[@intheworldofai]

Deepseek-R1 Engineer: POWERFUL AI Coding Agent Can do ANYTHING! (Opensource): kzbin.info/www/bejne/anmsgq2efNiHg7M

[@intheworldofai]

Deepseek-R1 + RooCode: BEST AI Coding Agent! Develop a Full-stack App Without Writing ANY Code!: kzbin.info/www/bejne/aHfVkqyllJ2irLM

[@intheworldofai]

Cline v3.2 UPDATE: Fully FREE Autonomous AI Coding Agent! (FREE API - Sonnet 3.5, Planning Agent): kzbin.info/www/bejne/np7XmKJrmbugb8U

[@intheworldofai]

Want to HIRE us to implement AI into your Business or Workflow? Fill out this work form: www.worldzofai.com/ 💗 Thank you so much for watching guys! I would highly appreciate it if you subscribe (turn on notifcation bell), like, and comment what else you want to see! 🚨 Subscribe To The Newsletter For Regular AI Updates: intheworldofai.com/ 🔥 Become a Patron (Private Discord): patreon.com/WorldofAi 🧠 Follow me on Twitter: twitter.com/intheworldofai 👾 Join the World of AI Discord! : discord.gg/NPf8FCn4cD 📆 Book a 1-On-1 Consulting Call WIth Me: calendly.com/worldzofai/ai-consulting-call-1 Love y'all and have an amazing day fellas. Thank you so much guys! Love yall!

[@rakly347]

I just got rickrolled by this LLM 😂 I asked it to make an app (one that I already made) to see how it does. It proceeded making a small script that uses my existing app and then gave me the lyrics to 'Never Gonna Give You Up'

[@wfcsargon]

Funniest coding LLM story I've ever heard! I actually burst out laughing on this one. That was absolutely spot on, DeepSeek!

[@pupcakestv3382]

Thanks for sharing chief. Ive been trying to get a api ket from deepseek but the site has been down for days. I downloaded and ran the model to my local pc. can I get it done running it locally? thanks again

[@BEYONDPREMIERE]

was waiting for this :D

[@ntokozoshaba]

You Sir have saved me a million hours on software building. Thank you a bunch. When I make my first buck I hope to remember you.

[@KishanLal-s4k]

I don't think deepseek-reasoner model support tools calling natively. I observed some inconsistencies even with Cline + DeepSeek-R1. I asked it to read a bunch of project files and update the readme file. DeepSeek just read the readme file and updated it with wrong info without reading the project files. Whereas Claude through AWS Bedrock was able to read my project files, understand the current state of the project and able to add correct info in the readme doc.

[@tuapuikia]

The model is known for hallucinations. Try to reduce the temperature and see.

[@KishanLal-s4k]

@@tuapuikia temperature will have no effect on deepseek-reasoner model. You can check out their official API documentation.

[@m.c.4458]

Check your prompt and refine it. Use a sample MD or supporting docs in your dir. So stop determinism. I note that doesn't like using my agents.

[@rakeshshukla9159]

​@@KishanLal-s4krun it locally on your setup

[@roberthuff3122]

Great! Could you please share your prompts? Thank you again.

[@mattiasw8465]

Next time, show how you evolve an existing application. I have noticed that very often I get duplicate code etc.

[@AlexJohnson-g4n]

DeepSeek-R1 and Cline are revolutionary! AI-driven no-code development could reshape coding. Excited to try these tools and see their impact.

[@NITHZAEL92]

Would you think Cline is better than bolt? 🤔

[@ViperWenom]

Dont say "revolution", socialism will punish you 😄

[@intheworldofai]

Codename Goose: NEW FREE AI Software Engineer Can DO Anything! (Opensource): kzbin.info/www/bejne/rYaygayJrJWJetk

[@envt]

Do you know if it's possible to use a local version of deepseek (ollama) with cline? I can't get it to work. The cloud version works flawlessly.

[@m.c.4458]

I tried too. But you would need the larger model and that needs a big computer

[@joeseabreeze]

@@m.c.4458 But you can use the 7b locally using ollama, so not sure why you can't get it to work with Cline.

[@digital_down]

I’m interested in this to if anyone finds an answer or solution

[@m.c.4458]

@@digital_down try visiting hugging face

[@ZeerakImran]

a local version of deepseek isn't deepseek. It's a modified/butchered deepseek. I know my comment isn't helpful but it may save you a lot of time and energy. Unless we have a budget of £30k perhaps, deepseek locally isn't going to be anything useful. Deepseek v3 sucks in my opinion (full most capable version). By sucks, I mean that it will waste your time. yes there's plenty we could say about it but I'm not here to waste your time. deepseek r1 is good but doesn't support tool calling and can't be run locally. Distilled versions may sound good in benchmarks, but so does 4o. And as good as 4o is, it's very annoying in the same way that deepseek v3 is. And is completely useless at coding. Claude 3.5 sonnet (latest) is the only one that's remotely useful. Its just the unfortunate reality.

[@titusfx]

Could you put the costs of your examples?

[@quertie]

what kind of hardware are you using for this?

[@jacquesdupontd]

128k input is very low tho ? If you want to work on an existing project with lots of lines of code. So to me, you can't create a full-stack app without writing any code because if there's only one error, and you don't know how it works and where is it and the project is having more than 128k tokens of length, then you'll be stuck. edit: Also, I just tried it to implement a Google Drive upload of a webapp backup in a PHP script. We went back and forth with how to get the Oath credentials and such matters and when it finally worked, it costs me 11 cents. It's not nothing. I'm not sure that Windsurf is more expensive with it's 500 prompts for 10 dollars to early adopters.

[@BEYONDPREMIERE]

Roo Cline doesn't read the full code just what you tell him to work my project is 300k token window and doesn't affect

[@jacquesdupontd]

​@@BEYONDPREMIEREBut if you need to analyse your whole code to know what's wrong, it needs to read it yeah ? Thanks

[@BEYONDPREMIERE]

@ yeah I organize it by components works well for me, I copy it to gemini 2 million for any issues and notify roocline

[@jacquesdupontd]

@BEYONDPREMIERE Kind of smart. You copy paste to Google A.I studio or you don't in VS code ?

[@wealthysecrets]

I HIGHLY doubt this model can create a full app, it's actually terrible in my coding tests

[@jermesastudio]

Thank you for this tutorial. It has been very helpful for my commercial projects.

[@korytny]

Hello, bro! Thanx for your research! i subscribed to your newslatter. Can you explain, how to connect deep-seek-r1 installed locally in ollama to cline in vsc ?

[@reserseAI]

Still roo cline + claude sonet via copilot 🔥

[@haidersyed6554]

Please zoom in a little on your prompts so we can see what you typing

[@tommy516]

So you have to communicate with their China servers giving them all your code???

[@uhack2417]

They shall be tested on larger projects instead of creating new ones from scratch each time

[@wolfganggermain7175]

Can it understand context of an existing code?

[@Mars2152]

Best AI bootleggers outfit out there, but don't worry about your great ideas because you give them away for free...

[@jinghuacao8407]

I have ollama running the same model locally in the same PC(127.0.0.1 works), when I try to cline, it says API request failed. 404 page not found.

[@cupideffect]

Keep it coming!

[@GallyTi]

I have a question. So Deepseek cant use files? You cant upload files, like images or code and so on? What about whole project? Can i use Cline on my already made project? Thank you :)

[@阿霖-v7b]

just open cline in current project, and type "read project". Done, no upload needed

[@abj136]

I’m going to try to replicate what you did in this video. I’ve not done any LLM programming to API before now.

[@NyloXD]

What about using the reasoner as an architect and using the R3 to implemnent it? I Heard that using the reasoner to do everything was not working properly with cline.

[@IamKonstantin]

thank you thank was great. subscribed.

[@John-ek5bq]

You created that in react?

[@yumna-y2p]

Which one is best for coding? I'm considering an open-source model, so please suggest what I should choose: 1.Llama 3.2 2.Qwen 2.5 3.DeepSeek-R1

[@hjchen-hq6ku]

Deepseek-R1

[@medivalwizard]

DeepSeek-R1

[@tuapuikia]

Gemini 😊

[@yumna-y2p]

@@tuapuikia gemini is free?

[@jameswang6714]

Use DeepSeek-R1, if you stuck, try others to get side ideas.

[@mr.gk5]

Do I have to create a new API key for this or just reused what I had for the new model? getting a 401 authentication fail here

[@justinberken]

Is this being run locally? What are your specs?

[@iamleifjohansen]

Does anyone else notice the API can be incredibly slow at times? Sometimes it works fine for long stretches, but then suddenly it takes 5 to 15 minutes or longer to respond. So frustrating! I discovered this with v3 over the past 2 weeks many times and now with R1. Any advice?

[@learnings.academy]

could be due to increase in number of users causing this issue

[@Nacktmull88]

yeah its completely useless. I got 1 response out of 5 prompts. No way im waiting 15 minutes for a response. Then i rather pay 10x for a claude API

[@NattaWang]

402 Insufficient Balanced was out as my response after followed your setup

[@LatinFunk-b6f]

same but i have to out some morrrrney

[@Tips-r-us]

tell ou another reason, OpenAI blocks many regionds, Deepseek doesnt,

[@d4rkside84]

Is it possible to link cline to a local deepseek r1 destillation?

[@whatisayoutubehandlebro]

wanted to ask the same, also would it eliminate the tokens and paid model? let me know if you find out

[@Dygit]

you can use LM studio to host your own r1 API and then connect it to cline with a proxy like ngrok

[@djayp34]

Yes, but it's really bad at function calling (just tried with 14b). If someone knows a way...

[@hippiworld]

I'm using VSCode, Cline, and Deepseek R1 Distill Qwen (7B) with LM Studio locally. I've got it all hooked up and Cline replies to my prompts in the Cline console inside of VS Code, but for some reason Cline won't actually make any changes to my files even though I've given it permission to do so. It will write code in the Cline console, but it never asks to implement the code, and once the response is completed, it just keeps looping by attempting to reply to my initial prompt over and over and over again and it never stops looping. I've been trying to troubleshoot this with no luck :(

[@madhavkwatra5888]

Any updates?

[@JasonEala]

Need to set this up asap🤙

[@haha-eg8fj]

I still prefer Claude. DeepSeek R1 is not bad for improving my code but not good enough. Many times it failed to improve anything.

[@llIllIlI]

If it runs locally, how / why are we charged for tokens? I don't understand how that works if it's local on your own machine. Any info on this?

[@jonclement]

via Dygit: "you can use LM studio to host your own r1 API and then connect it to cline with a proxy like ngrok"

[@syvern7197]

To run a model locally, what is the decisive element for it to run smoothly, is it the CPU or the Gpu ram or is it General ram?

[@RenRagDJ]

Mostly its DRAM and VRAM

[@redlobsta1]

is there a way to have the cline sidebar to the right? i guess im so used to cursor.

[@psynchro]

Yes, just drag the icon over.

[@a3igner]

deepseek doesn't work with Cline anymore. What's going on? I see they have some API problems. Can you confirm?

[@gmp782]

yes i have the same issue. It will be back soon I think... Could you please let me know when it will be ok for u again ?

[@DePistolero]

Well awesome!!! Thanks!!!

[@xelerator2398]

Does R1 support images?

[@m.c.4458]

If cyberattacked nothing we can do to use it. Massive issue...

[@burger1886]

Can it make files and use frameworks ?

[@Jamescito_]

This is a game changer

[@Ren_Zekta]

Cline don't work for me, any LLM just ignores the prompt and answers to the directories that cline writes as part of the prompt.

[@theamazings2296]

cline now supports DeepSeek API, no need to select OpenAI Compatible for API key input

[@ArdiMuluk]

if you don't mind helping, my API request never complete and always loading, any suggestion?

[@GoldenVlAdStix]

@@ArdiMuluksame here

[@solomslls]

Can i use cline with Lm-studio and deepseekr1 distil llama model ??? That would be game changer

[@intheworldofai]

Yessir! You can do that! Even with ollama

[@mateocarino]

Yes you can, locally with llama 8b i think

[@solomslls]

@@intheworldofai I never used cline , I'll try it, hopefully I can lm-studio api and use it ,

[@KohliInCanada]

I tried the setup with Lm-studio and deepseekr1 distil qwen model and it gave me the error " Cline uses complex prompts and iterative task execution that may be challenging for less capable models. For best results, it's recommended to use Claude 3.5 Sonnet for its advanced agentic coding capabilities. "

[@solomslls]

@@KohliInCanada that's so weird ! Hopefully there is a way to workaround that .. I'll try and tell you if I manage to

[@Derick99]

How come deepseek doesnt support image's

[@hnchanhk]

It seems related to cline, not deepseek

[@gani2an1]

It is much much cheaper than OpenAi o1 but it is still too expensive to run a large codebase on. Any chance of downloading this and usign it locally??

[@slotsmaster8877]

It is much cheaper than human coder for sure

[@echo-nomade]

Try the distilled versions

[@whatisayoutubehandlebro]

@@echo-nomade that would make it 100% free right? also can cline work with it?

[@echo-nomade]

​@@whatisayoutubehandlebro yes and yes But keep in mind that the distilled versions are not as performant than the original R1 model.

[@whatisayoutubehandlebro]

@@echo-nomade I thought the distilled versions were pretty decent

[@ron2344-f2f]

i wanted to build a android app, can ai agents build it ?? like no code app building for android apps??

[@jackpre3399]

Yep, I've built four or more apps with Windsurf. And I didn't know a thing about coding or Android Studio!

[@MeditationOasis_BaGRoS]

You can use expo + cline

[@Sanyat100]

how are you using R1? R1 does not have a web version yet. You can only run it locally.

[@test-zg4hv]

wrong

[@lazycoder01]

It does

[@KriGeta]

hey, can you please make an blogger XML theme using this? when showing these examples?

[@nithinreddy7556]

it says 402 Insufficient Balance but i just started using it. what went wrong please help

[@Relics]

you have to add balance stupid

[@simplyebeats6847]

.......Did you add money?

[@cheyer8869]

I’m not very impressed by this combination. It’s overly buggy, repeatedly looping through code changes and sometimes duplicating-or even tripling-the lines of code.

[@查尔斯温哥华生活]

Which tool would you recommend?

[@Quincycle]

@@查尔斯温哥华生活your own fingers

[@ipch]

cursor, codeium, windscribe.

[@dimasound8438]

R1 support images???

[@atchutram9894]

No

[@viniciusargolo3146]

Great!

[@gslvqz8812]

You typed 'stinky' instead of 'sticky' note; that is why the model got confused.

[@intheworldofai]

LOL I just realised omg

[@MrAtomUniverse]

The token uploaded is crazy high wtf

[@MrAtomUniverse]

came here for the clickbait, Deepseek-Coder

[@nguyenanhnguyen7658]

English and Chinese only. The rest of linguals are way hallucinated, and badly.

[@mrGapMan1]

If you're on windows Cline, do you use WSL when coding? Because the Deepseek models insist on using Unix commands and that doesn't fly with Windows Powershell.

[@shazzydog1]

No serious developer would ever code on windows

[@Guroviski]

Create a Windows/Linux dual boot on your computer

[@user-ze9tj9yj4t]

Do you connect to WSL with VS code when you open a project?

[@mrGapMan1]

@@Guroviski will projects created in linux or win/wsl work on Windows natively?

[@TheSoloDIYer]

Just use bash instead of PowerShell. GitHub desktop comes with bash, just use that. I wouldn't dream of using the windows command line for development. That's nightmare territory.

[@nick0lasj0nes73]

You lost me at Base URL

[@nunajah]

What about Roo-Cline?

[@razorbackroar]

Hey bro. Roo cline is easier it has deep seek chat aka deepseekv3 then deep seek reasoning aka R1

[@yumna-y2p]

I'm getting 402 Insufficient Balance using deepseek api

[@saber8387]

Its not free you have to buy credits

[@yumna-y2p]

@@saber8387 so how much i have to buy?

[@celebrityaudiobooks]

I think the problem is that you have insufficient balance.

[@fred3r1k]

@@yumna-y2p 5 or 10 $ i guess

[@drhouseshield]

Deepseek r1

[@DaniMartVTen]

Why not develop DeepSeek as its own negotiation-based operating system? If this was put inside of an empty Orange Pi keyboard, why couldn't it measure its own device capability and then solve all other software and hardware tasks? The trio of Windows/Linux/Mac is defeated!

[@KevinOsterkilde]

04:25 "stinky notes"

[@lighteningnewspodcast]

A powerful coding AI model ? Sounds like the best way to inject backdoor in the whole world

[@n-i-n-o]

Facepalm

[@chun_1310]

Only thing stopping innovation is dumbness of few people

[@magicismagic123]

can you explain, what you mean?

[@chun_1310]

​@@magicismagic123 bro thinks fully offline transformer models have something to do with chinese softwares and backdoors, that he was fed in by media because thinking by himself is too hard, or it need brain to begin with

[@lighteningnewspodcast]

@@magicismagic123 I believe when people use ai agent to write code, they don't review code. Then it is easy for the AI agent to insert some backdoor in your code. Maybe, you can add another AI which is trustable to review code for you 😂

[@Nacktmull88]

Im not getting one response from the API in Cline. I guess the chinese LLMs are exactly what you can expect from chinese LLMs...Exactly the same trash as any other chinese product too.

[@TuoiLevan-n4i]

Con này thua chát GPT về mặt hỗ trợ ngôn ngữ nhưng hơn chát GPT về mặt không giới hạn kí tự...

[@StrongerFX]

Its cheaper because you pay with data hahahaha

[@mechtxofficial]

157th :) 🙋

[@peteranders6893]

Tried so many times getting exactly 157th, but it is hard. Very well done. 🎉

[@megakilllll]

Use RooCline instead of Cline

[@simongentry]

better than ChatLLM?

[@janalgos]

ChatLLM just gives you a drop-down to choose any model you want, including deepseek

[@simongentry]

@ i like that it chooses which llm to use based on input…

[@janalgos]

@simongentry not necessarily, you can choose whichever one you want or give it the ability to choose

[@intheworldofai]

Deepseek-R1 Computer Use: FULLY FREE AI Agent With UI CAN DO ANYTHING! (Beats OpenAI Operator): kzbin.info/www/bejne/hoPFdHmdiNuUmdE