Continue Plugin not properly working

Before submitting your bug report

• I've tried finding an answer on the Continue docs site
• 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

name: Main Config
version: 1.0.0
schema: v1
models:
  - name: Qwen2.5 Coder 32B
    provider: ollama
    model: qwen2.5-coder:32B
    apiBase: http://127.0.0.1:11434
    roles:
      - chat
      - edit
      - apply
  - name: Qwen2.5 Coder 1.5B (Autocomplete)
    provider: ollama
    model: qwen2.5-coder:7B
    apiBase: http://127.0.0.1:11434
    roles:
      - autocomplete
  - name: Qwen2.5 Coder 7B (Embeddings)
    provider: ollama
    model: qwen2.5-coder:7B
    apiBase: http://127.0.0.1:11434
    roles:
      - embed

Description

I am trying to write code for C language using Qwen 2.5 Coder. The LLM suggest ok code, the Continue plugin will modify the code accordingly
but then when it's time to compile the code instead of doing it it will display the steps it needs to do with some HTML tags.

And other times those HTML tags appear after suggesting some code modifications. If I tell the plugin to retry, sometimes it works but sometimes those HTML tags will appear again.

See next example:

"Now that we have verified the contents of the files, let's compile the program:

<function=run_terminal_command>
<parameter=command>
gcc -o main main.c bubble_sort.c


If there are no errors, we can then run the compiled program:

Apply
<function=run_terminal_command>
<parameter=command>
./main

"

Also part of the suggestion shown in Chat aboutwhat next step will be inserted in one of the C files of my code:

"<function=run_terminal_command>
<parameter=command>
./main

"

This previous example should execute ./main binary from compile source code, and not inserted into one of my C files.

To reproduce

To reproduce I am just trying to use the plugin, it's very easy, it appears in Chat or Agent after 4-5 conversations.

Log output sees what I see in Chat/Agent zone where instead of those HTML tags I should see only the command the agents wants to execute and I should be able to "Apply" (allow) agent to run the command. Which I am able to do, to press "Apply" but instead of running that command, that part with HTML tags is added to the file I have opened on the right window (source code file of my C program).

Log output

<function=run_terminal_command>
<parameter=command>
gcc -o main main.c bubble_sort.c
</parameter>
</function>
If there are no errors, we can then run the compiled program:


Apply
<function=run_terminal_command>
<parameter=command>
./main
</parameter>
</function>"

Start by reproducing the issue with the provided Ollama configuration and the logged run_terminal_command output after several Chat or Agent conversations. Trace why the tool-call markup is inserted into the open C file instead of being executed or shown for approval. Done means commands such as gcc and ./main are handled as tool actions rather than file content.