Why don't you use postgres with pgvector for free?

tried, it works! thank you

Amazing work obviously, but sorry, IMO that's not a release. To me a release would be a compiled (optimized) "Build" version that can run without all of the Node dependencies, backend server, etc. For your average user to just launch with the click of a button, and it just works. But as a starting point, I think it's great!

greate video , one question how i can use this in my web or access to it in the web in new page because is just loading without send the message to n8n what ever use localhost or even use ngrok

New to docker. I kept haveing errors trying to install the github files. After a lot of searching the solution was to login through the cmd line before executing the 3rd command. so "docker login -u my-user-name" fixed the failed to authorize: error 401 unauthorized. Eveerything seems to be working now!

i dont see the link to this workflow in the comments or video description as promised in the video

cole I would be very to help in this project bolt.diy

Too many marketing words for me. I need numbers, I need stats, prices, comparison... How am I 5mins in and I've heard bunch of nothing :(

Awesome buddy

Hey amigo! It looks really cool, but sorry-I don’t mean to be mean-I don’t fully get it. Some of the examples you show, like the blog builder, for example, can’t this be done just by using ChatGPT? Or is there something that I am missing? Is using the agent somehow better?

Awesome work!!!!

I have been working on curating a library like this and working towards implementation with Bolt.DIY. The backend and the front end. The LLM will use vision analysis for the user's initial prompt and decide which model best suits the specific use case based on the variable parameters set. I like the way the roadmap goes, the majority of the big tech companies are already ahead of the game on their features and LLMs. That's why it's imperative to get some contributors going.

Wasting time and energy on half decent ide, focus on integration with what developers are using, 99% of users already have existing projects and setups. All of that ui is not needed imo, anyway great job 👏

make it fully automonmous. for testing / debugging and have it write itself.

@ColeMedin if you go through the comment section you’ll see a lot of people especially non-devs are struggling to use the service - this has to be the highest area of focus (user experience) above shipping new features if this project is going to stand the test of time. Please keep a product manager hat on and pay attention to the user feedback. Cheering you on 🙏🏾✊🏾👌🏾

00:00:00 Introduction 00:00:20 Tip 1: Use Superbase for Scalable AI Agents 00:01:15 Tip 2: Choosing the Right Large Language Models 00:02:15 Tip 3: Extracting Text from Different File Types 00:02:45 Tip 4: Referencing Previous Node Outputs 00:03:30 Tip 5: Building AI Agents as API Endpoints 00:04:30 Tip 6: Handling Multiple Items in a Single Node Output 00:05:30 Tip 7: Using Data Pinning to Save Test Event Outputs 00:06:00 Tip 8: Creating Error Workflows 00:06:45 Tip 9: Using the Schedule Trigger to Automate Workflows 00:07:30 Tip 10: Exploring the n8n Workflow Library 00:08:30 Conclusion

I would love if my car had a smart enough LOCAL LLM like Llama 3.3 … but the memory needed for that would prohibitive. BTW, why is memory not getting cheaper ??

apparently the simplest and most elegant way to interact via chatbot with your n8n workflows without having to go through the internet. via VPN I can use it safely from my cell phone without having to expose anything in the public network. I will happily follow your tutorial, incredible how almost everyone else resorts to telegram for no reason, nullifying the effort of self-hosting everything else

amazing stuff ! Thanks Cole

I think PydanticAI would be a game changer. I am already excited to build cool applications with it. Have been using langchain so for. Thanks for the great tutorial!

how to make the sidebar of Windsurf bigger as your? My sidebar is quite small (compare to Cursor or VSCode sidebars) with the same default zoom level

Perhaps the quickstart guide isnt as quick as desired. I setup a docker based instance. Where exactly are the files it generates? I browsed through the container and couldnt find a hello.txt I generated. I used a terminal to get it to tell me the path... alas I cant find it. Seasoned developer, docker noob.

so it is not an open source version, it is an open source fork where the closed source fork doesn't contribute?

Images, yes, thanks. WRT the reasoning step, it could be good to have a pause there, in which the user can read the plan & either go ahead or modify either the prompt or the plan.

It doesn’t work , it stops creating applications , it’s very hard to work with bolt.diy

You are nailing it bro

nice work coleMedin i hope you create best ide and best extension for vs code too so all developers will convert to your tools instead of windsurf ide and dont forget to add button donate

Hi Cole! Already thanked you on the forum but want to say thanks again for working on this amazing project. It surely seems to have paid off with the new collaboration with Stackblitz! I have an issue I haven't been able to solve, I can use the Deepseek V2.5 context 65k (which seems to be using deepseek-chat) with my openrouter api key, but cannot find a way to use the Deepseek Coder V2 236B version you use in the vidéo. I put my OpenRouter API key in both the .env.local and the prompt windows. Do you have an advice for me? I would be so thankful!

Liked and subscribed - Great job Cole!

I have a problem with it I buy api from deepseek the main website but when i link it to my code sometime its not writing write code and sometime the website became very slow and alowes when my code became big its stop working

Hi Cole Have been trying to follow this but the function repository on the OpenWebUI website no long works. is there any chance you can make the script available somewhere else? Many Thanks! :)

It’s a complete waste of time. Spent 1 hr trying to build an app and so many hours debugging it. Then I started something on my own and I can build it in just 3 hours without any issues. 😂😂😂

When I import a folder, unfortunately it does not import more than 1000 files, how to solve this?

the prompt library seems like something that has to be continuously developed and is not a goal with a finish line. dont dilute your focus straight out of the gate. you need to set achievable goals. pick the single best LLM or the most popular, and work around that. then as niche users spend their credits you can analyze the outputs and work on the library.

For the love of God, i cant run the damn thing.. always "There was an error processing your request:" on the bottom right, even the console just shows "ERROR", no network logs, no console logs, sigh.. tried both pnpm and docker instructions to no avail. .env.local saved... nothing works My Ollama is WORKING 1000000% using pinokio/open-webui, so its really on bolt diy side... i wanna try bolt so baaaadd.. :(

Many Features are ready and waiting to be merged.

I have a suggestion. When writing code with bolt.diy it always starts from pakage.json file, but I only want to edit a single file

Your enthusiasm is infectious.

Please fix the lm studio integration.

Hi , what's is the difference with lindy ai please ?

We want a simple install set up, I can't make it work

Congrats🎉 YOU ARE IT !! Build together, Grow together! Pure examples that we all have a chance to build a living from community again! We just got to want to do it together instead of compete. Share the power of who you know and celebrate others success!

I am using qwen2.5-coder:3b and cannot access the preview because my system has low specs. 🖥 The only model I can run is qwen2.5-coder:3b. Could you please help me create a bolt.diy setup on Google Colab? 🙏

Can you add an 'All Auto' option so it automatically takes a screenshot of the web app after each step and send it to the llm as the result. Also, could you include a mechanism to test the web app with 'claude ai computer use'?"

I am getting error: There was an error processing your request: An error occurred.

Hey cole , dont forget the fact that the main thing is the coding power of the assistant at the end.also the fact that it needs to be easy for non programmers Make it so non programmers dont have to debug anything On windsurf you can make context settings for the ultimate goal of your coding , please include this but the easy way where you can give a chunk of of context to make the AI assistant focus on the bigger picture and never lose track of Please know the way , the modifications , the best free LLM that suits bolt.diy and can make you do a full complex app or program because at the end of the, i dont think anyone right now can use to make a complete thing . So this is the biggest obstacle i guess Be strong and finish it, thank you for your job ❤

Can you add flowise as a provider. Then we can use that to call the n8n RAG. Why do you keep avoiding adding a RAG agent to bolt. Then all these things can be done in n8n. I don't see why you won't. It makes sense to have your own RAG as model. Flowise has the API calls that Bolt needs for the base url. You do alot with n8n and RAG agents. But why don't you allow us to use our n8n RAG to bolt.

Nice work! Token calculation is great addition, would love to see that also translated into currency amount (i.e. Total tokens 6520 (~$0.27)). Also to have it as VS Code extension would be amazing.

Can we make MOBILE APPS???

I tried it with chatgpt4o mini and it's TERRIBLE!!!!!!!!!!!!!!!!