Bug: Cannot Parse Response From `deepseek-r1:latest` Correctly in JetBrains

[zzyu17]

Before submitting your bug report

• I've tried using the "Ask AI" feature on the Continue docs site to see if the docs have an answer
• I'm not able to find a related conversation on GitHub discussions that reports the same bug
• 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: Windows 10 22H2
- Server OS: MacOS 15.3 (24D2059)
- Continue version: 1.0.59
- IDE version: PyCharm 2025.3.1.1
- Ollama version: 0.15.5
- Model: deepseek-r1:latest
- config:
  
name: Local Assistant
version: 1.0.0
schema: v1
models:
  - name: deepseek-r1:latest
    provider: ollama
    model: deepseek-r1:latest
    apiBase: http://[server-ip]:11434
    roles:
      - chat
      - edit
      - apply
      - rerank
      - autocomplete
context:
  - provider: code
  - provider: docs
  - provider: diff
  - provider: terminal
  - provider: problems
  - provider: folder
  - provider: codebase

Description

When using chat mode in Pycharm with deepseek-r1:latest, deployed on ollama on a remote MacOS server, Continue fails to parse the response correctly and return only the thinking contents.

How should this be fixed?

To reproduce

No response

Log output

[continue]

Related issues:

• Support configurable thinking output formats #5992 - Support configurable thinking output formats
• think tags #3901 - think tags (DeepSeek R1 via Ollama, tags stripped)
• Not able to disable thinking mode for the self hosted models which use OpenAI as a provider (please read quick description) #9102 - Not able to disable thinking mode for self hosted models using OpenAI provider
• BUG: Edit mode with gpt-oss-120b inserts reasoning text into file #8990 - Edit mode with gpt-oss-120b inserts reasoning text into file

[RomneyDa]

@zzyu17 is this a regression which showed up in a recent version or does it fail on older versions for you as well?

[zzyu17]

It showed up in a recent (actually the lastest) version of Continue, Pycharm and ollama. But I hanven't try it on an older version. Should I give it a try?

[zzyu17]

Well I just tried a more couple times and found a temporary "solution" to this bug. I did the following two things:

1. Close the GUI window in my remote server;
2. Add a snippet configuring completionParser under the model in local config as follows:

name: Local Assistant
version: 1.0.0
schema: v1
models:
  - name: deepseek-r1:latest
    provider: ollama
    model: deepseek-r1:latest
    apiBase: http://100.108.196.57:11434
    roles:
      - chat
      - edit
      - apply
      - rerank
      - autocomplete
    completionParser: |
      (data: string) => {
        const parsed = JSON.parse(data);
        return parsed.message.content;
      }
context:
  - provider: code
  - provider: docs
  - provider: diff
  - provider: terminal
  - provider: problems
  - provider: folder
  - provider: codebase

The result is sometimes it worked but sometimes didn't, and I can't figure out when it will work and when it won't. When it works, it shows the response like:

[zzyu17]

@RomneyDa Also I found when it worked, it might mess up the question and response, and the response might repeat exactly the same as the former one.

Continue_bug.mp4