Everything You Need to Get Started with bolt.diy

image
yup, the nodejs is in the path under the environment.

PS: Yup, the pc is named user.

Not exactly what I meant for path set in the environment variable (but this should be taken care of by the Node installer).

It’s what tells Windows how to find various binaries. Otherwise, you would need to put the full path to npm when running it.

And I was curious, because I’ve never seen someone name their user account user. Very amusing.


I’ve added the path under user environment instead of system environment, and it seems able to work, but not quite sure whether it runs successfully or not. I assume it will open a browser window if its up and running? How can i access to the bolt.diy homepage?

Wrong folder, you’re not in the Bolt.diy repo folder.

As a note, I would probably remove the node_modules from the root of your users folder that you just created. And either path works.


thank you for your patient in guiding me.
Now it has another error here showing…

This is a known issue on Windows 11. Install the Latest Version of C++ Redistributable.

Thanks Alias! Finally get it to work!

PS: I saw many talking about the error when first starting the prompt where it will occur even the API key has been inserted, this is due to the API itself have to be subscribed first. Say Claude, you may got the key inserted, but it will not work unless you have credit inside for the API subscription, which i think this has to be highlighted on the tutorial page, else many will raise this error over and over.

In short, by using Bolt.diy isn’t entirely free, unless we have a free and powerful LLM. Feel free to correct me if im saying anything wrong here.

1 Like

Not completely true. While OpenAI and Anthropic require payment processing with a balance, there are MANY other options:

  • Openrouter and many other providers give you free credits to start and they have several free models you can also use.
  • HuggingFace has Interface API (serverless) endpoints that are all free, with pretty generous rate limits.
  • Google AI Studio offers their “Experimental” Gemini Flash 2.0+ models for free (which are quite good and also has generous rate limits).
  • Microsoft offers Azure AI Models through GitHub for free, including ChatGPT 4o.
  • You can also run LMStudio or Ollama locally for free with many of the models available on HuggingFace…
  • And arguably you could connect a Kaggle, Google Collab, or various Jupyter Notebook-type platforms with it.
  • There are even ways to get Anthropic’s Claude 3.5 for free.

So, there’s really no limit of options to get started. Many free and good enough for your average user. But the options can also be overwhelming when first starting out with AI, so I’d suggest just get started with HuggingFace and go from there.

And if you are willing to pay money, the new DeepSeek-V3 model is quite good and available for < 20¢/MTok, which is super cheap. They charge a fee per transaction though, so I generally just keep a rotating balance on it once it gets low.

P.S. I should also mention, you can even host Bolt.diy (and customize it if you wish) completely for free. See my article on Deploying Bolt.diy with Cloudflare Pages (the easy way!). Or you can just use the one’s I’m running (need to provide your own key):

So, you can literally run Bolt.diy COMPLETELY Free! No strings attached.

Good luck!

1 Like

Maybe easier to use Pinokio if you have issues with command prompts and installation:

1 Like

@itqandas see my post here: Installing bolt.new Ottomator using Pinokio - #3 by leex279

1 Like

i keep getting this error

Cannot read properties of undefined (reading ‘toolCalls’)

Window 11 I presume. Install the Latest version of C++ Redistributable.

just install the latest version of C++ Redistribution, but still keeps getting this error

I believe this issue happens if you install using npm install instead of using pnpm install, and it could be for just one or many models. Can also be causes by vite having an issue or missing.

Try deleting the node_modules folder and run npm install pnpm & pnpm install & pnpm run dev.

And sorry, mixed up issues.

1 Like

@DigitalVisions as postet above, Pinocia does not use the stable branch, what can cause some problems.

I would recommend installing it according to the docs, without Pinocio.

Here a install video for Windows from me:

There are also other videos from @dustinwloring1988 :

apparently it does now:

I dont see a reference to a one click installer :smiley:

Thanks to how Gihtub works, the old unofficial github URL redirects to the official repo. Which means, the old pinokio launcher using the “unofficial” BOLT was ALREADY working with the official BOLT repository automatically when the transition happened thanks to the redirect.

But there were some details such as custom environment variables which needed to be added to the launcher in order to support all the new LLM providers mentioned in OP. This new version does that, and also built for Pinokio 3, with optimizations such as using UV. As well as setting the defaults for OLLAMA API BASE URL and LMSTUDIO BASE URL so it just works out of the box for local use cases.

Just reinstall from scratch. It should just work: program.pinokio

1 Like

Ah ok, thanks for the info. Then just missing that they switch to the stable branch instead of the main branch.

Just got the info it is done: