Unable to run bolt.diy locally and unsure of problem

Hey. I’m trying to run bolt.diy on my computer, and while the ui loads, it’s spitting out some awfully vague error messages. Currently on a macbook m1 pro from 2021. This is the first time I’ve tried to run an application from a file downloaded on GitHub, so any help would be much appreciated. I’ve attached my terminal output below (I’m only allowed to attach one image, the other one is simply"There was an error processing your request: An error occured"):

From what I can gather, the server is not sending a response, or something like that. I have entered an API key, but, I don’t really know what to do. Also, if anyone could suggest alternatives to bolt that would be much appreciated.

Hi @GiantWizard ,

please verify the API is working outside of bolt first. Use curl to do that. There should be an example in the deepseek backend.

Maybe you get an error and know then whats wrong.

Sorry- I’m very new to most of coding.

is this what you mean by backend?

Copied the request and put my api key in and it came out like that?

no idea though, I tried OpenAI and Anthropic as well. Both don’t work and simply say it’s sending a request to the servers.

Actually, I do see that the API requests aren’t returning the right value. Here’s one for OpenAI, though I don’t know how to fix it:

It says you dont have “Balance”, so no credits available. The API is not free, you have to buy credits for this and if no available you get this error.

You could use Google AI for free, if you dont want to pay.

You can checkout my video at about minute 11:50 I show the usage of Google AI:

Okay. Is Google any good for coding? I’ve heard a lot of things about Gemini’s benchmarks, and they’re not very good…

I did the whole task-list-advanced project on Gemini 2.0 Flash, what I also showcase on my youtube channel.

holy crap it worked. how do i mark this as solved?

1 Like

if you close it, how are the steps again when you used it? should it be the whole process again?

@GiantWizard at the moment there is no feature to solve topics, but its in discussion to enable / install

@ColeMedin see here and also my topic in Staff

Actually Gemini Flash 2.0 (exp-1206) is free and very good. It rivals that of ChatGPT 4o to me and in some cases even Claude Sonnet 3.5. You can get access to many models through HuggingFace for free, including Qwen 72B and 32B Coder (pretty good), LLama3.3 (okay), and you can Get ChatGPT 4o free through GitHub (Azure; prety good).

And if you are willing to pay some money, I would recommend Openrouter which gives some free credits to start, offers several models for Free and is like a one-stop shop. Currently everyone is raving about DeepSeek-V3 which is available for ~20¢/MTok (which is crazy cheap for as good as it is).

Best to worst imo:
Claude 3.5 & o1 > DeepSeek-V3 & Gemini Flash 2.x > 4o > Qwen 72B Instruct & 32B Coder, LLama3.3 70B Instruct…

P.S. @ColeMedin I like the new Summary with AI feature and the Chatbot.

1 Like

Glad you like the new summary with Ai feature!

Also great best to worst list - I agree 100%

1 Like

Thanks for the ping! Saw this and that’s something I’m looking into. Not 100% still it’s available in Discourse so I’m looking at external plugins.