The style output of markdown is incorrect

[thomasbergersen]

Before submitting your bug report

• I believe this is a bug. I'll try to join the Continue Discord for questions
• I'm not able to find an open issue that reports the same bug
• I've seen the troubleshooting guide on the Continue Docs

Relevant environment info

- OS:
- Continue version:
- IDE version:
- Model:
- config:
  

  
  OR link to assistant in Continue hub:

Description

Hello. The markdown style output using Prompt: 'vue vs react output tab' was not rendered correctly.

To reproduce

No response

Log output

[JerryZhongJ]

I have the same issue:

[chezsmithy]

@thomasbergersen and @JerryZhongJ there is a markdown patch coming in a release shortly. I've tested locally and if I ask for the prompt "vue vs react output table", I get a table which is more like a CSV output. However, if I ask for it in markdown format after adding my patch it renders like this:

# Vue vs React Comparison

Here's a comparison table highlighting the key differences between Vue.js and React:

| Feature | Vue | React |
|---------|-----|-------|
| **Learning Curve** | Generally easier to learn with more intuitive syntax | Steeper learning curve, especially with JSX |
| **Syntax** | HTML-based templates with clear separation between template, script, and styles | JSX (JavaScript XML) mixing HTML with JavaScript |
| **Size** | Smaller bundle size (~33KB) | Slightly larger (~42KB) |
| **Performance** | Very good performance, optimized out of the box | Excellent performance with virtual DOM |
| **Data Binding** | Two-way binding available (v-model) | One-way data flow by default |
| **Component Architecture** | Single-file components (.vue files) | Components typically in JSX files |
| **State Management** | Vuex (official), Pinia | Redux, Context API, Zustand, Recoil |
| **Tooling** | Vue CLI, Vite | Create React App, Next.js |
| **Mobile Development** | Vue Native (less mature) | React Native (mature ecosystem) |
| **Community & Ecosystem** | Growing, but smaller than React | Very large community and ecosystem |
| **Corporate Backing** | Independent project led by Evan You | Backed by Facebook (Meta) |
| **Flexibility** | More opinionated, easier to follow conventions | More flexible, multiple approaches to problems |
| **Popular Use Cases** | Small to medium applications, prototyping | Large scale applications, complex UIs |

I think. this is what you were expecting?

Note the PR you are watching for in a preview or public release is: #5624

[chezsmithy]

I belive this is resolved with: #5624
@Patrick-Erichsen could you please validate.

[Patrick-Erichsen]

+1, believe it should be resolved.

[thomasbergersen]

Hi @chezsmithy @Patrick-Erichsen , I think it should be a rendering issue of markdown syntax by 'continuedev'.

Next, I use 'lmstudio' to test the output:

or Output original format

[chezsmithy]

@thomasbergersen is this more in line what you are expecting? #5560

[thomasbergersen]

@thomasbergersen is this more in line what you are expecting? #5560

It's like this. After the merge, an update will be released and tested again. thank you

[thomasbergersen]

Hi @chezsmithy , I have another question. When I deploy Qwen3 with vLLM using the parameters --enable-reasoning --reasoning-parser Qwen3, the content inside the tag in Continue.dev is kept in the background and only the final result is output—this works for both chat and edit functionalities, which is exactly what I expect.

However, when I deploy using the --reasoning-format deepseek parameter with llama.cpp, the output includes the content inside the tags. Currently, this behavior only applies to the chat functionality and not the edit functionality.

Is this an issue with the llama.cpp backend service?

Is there a way to make Continue.dev discard or hide the content inside tags, so it can be used in the edit functionality without outputting the unnecessary think content?

[chezsmithy]

@thomasbergersen I'm not familiar with that model. I'd recommend opening a new issue for tracking.