A few things. I think you are getting a combination of CORS and mixed-content (mixing http and https content) errors as the images are likely from the web, and not in the project itself (neither Bolt.diy/Bolt.new generally bring in their own images). As for Qwen2.5 Coder 32B (open source/free model) being good, it is for simple things as long as you guide it, but it’s not going to one-shot things like Sonnet (commercial/pricy). Heck o1 doesn’t even seem to be as good!
For one, you are running locally, not behind HTTPS… this will cause all sorts of issues when trying to access web resources. Maybe try my tutorial on Deploying Bolt.diy with Cloudflare Pages (the easy way!). This would fix the HTTPS and mixed-content issues, but I’m not sure about the CORS problem. You could also always modify the code to load the images from the project (but you’d have to put them in it).
If you need better debugging, there is an option in the Bolt.diy settings, but I’m not sure it would have helped much in this case.
And while saying “there’s too much needing fixing right now” is fine, but part of it is understanding the thing you are using (differences and limitations), and not assuming it’s a bug. It’s not. But it is what it is.
Good luck!
P.S. Also @shawn, if you don’t mind a little experimentation, maybe try Qwen2.5 72B Instruct and Llama3.3 through HuggingFace for free. Or Google Gemini exp-1206 through Google Studio for free. You can also get 4o for free through GitHub, but Bolt.diy support is kind of pending.