Error streaming edit diffs: Token limit reached. File/range likely too large for this edit

[AgainPsychoX]

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 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: Linux x64 
- Continue version: 1.2.1
- IDE version: VS Code 1.103.2
- Model: `deepseek-r1:8b` via ollama
- `config.yaml`:
  
name: Local Agent
version: 1.0.0
schema: v1
models:
  # DeepSeek with separate no thinking version, because other roles need speed
  - name: DeepSeek-R1 8b Thinking (local)
    provider: ollama
    model: deepseek-r1:8b
    roles: [chat]
    defaultCompletionOptions:
      contextLength: 8192 # 7.5 GB
  - name: DeepSeek-R1 8b NoThink (local)
    provider: ollama
    model: deepseek-r1:8b
    roles: [autocomplete, edit, apply]
    defaultCompletionOptions:
      contextLength: 8192 # 7.5 GB
    requestOptions:
        extraBodyProperties:
          think: false

Description

I can't understand why edit mode doesn't work. I work with very small file, 50 lines. I also have open (in other tabs) 2 other files, also very small. I tried to use Deepseek R1 8B with and without thinking. Neither works... It works properly in the chat mode, even if I copy-paste my code there.

BTW, can't find Continue: Enable Console, can't troubleshoot properly:

To reproduce

1. Setup ollama with deepseek-r1 8b in Continue
2. Open my file
3. Select some code
4. Use Ctrl+I to invoke the edit mode
5. Enter my command
6. It fails :(

Log output

Error output:
> Error streaming edit diffs: Token limit reached. File/range likely too large for this edit

Developer Tools error:

 ERR [Extension Host] Error handling webview message: {
  "msg": {
    "messageId": "64bdc51d-2467-4058-a62d-66e79da72427",
    "messageType": "edit/sendPrompt",
    "data": {
      "prompt": "I want to read newest HTML file in the directory",
      "range": {
        "range": {
          "start": {
            "line": 5,
            "character": 0
          },
          "end": {
            "line": 5,
            "character": 83
          }
        },
        "filepath": "file:///home/againpsychox/Public/Tools/CAN/run.ps1",
        "contents": "$pattern = '<tr[^>]*>.*\\n.*\\n.*>(\\d+)<.*\\n.*\\n.*\\n.*\\n.*\\n.*>(\\d+\\.\\d+\\.\\d+\\.\\d+)<'"
      }
    }
  }
}

Error: Error streaming edit diffs: Token limit reached. File/range likely too large for this edit

[8bit-coder]

I have the same issue, I tried increasing the context limit to 65535 and it does nothing to help

[Shourya-Agarwal]

Yes, this issues always comes when editing files.

[tomasz-stefaniak]

@AgainPsychoX you can press CMD + , to open settings, then search for "continue" and you should see a checkbox to enable the console. Then, you should be able to see Continue console in the bottom bar:

[ernesto-elsaesser]

I have the same issue with gpt-oss:20b (Ollama, default config), and I believe the problem is that it thinks a lot, and Continue does not ignore the thinking messages. Actually, in the cases where I don't run into the token limit error, the thinking output is included in the code diff. Maybe properly parsing and discarding the thinking blocks would resolve the "token limit reached" issue for thinking models, as the actual "final" response only contains the edited code (as requested), which normally should not exceed any limits.

[chhu]

Also happens on ollama with new qwen3-coder, no thinking involved. Context size is model max (250k). Big bummer!
Looking at the console, the model answers exactly as it should be but after 1668 tokens just cancels!

Also it seems I cannot find the error message in the code base:

Error streaming edit diffs: Token limit reached. File/range likely too large for this edit

Where does it come from?

[milgner]

Confirming this with qwen3-coder:30b and ollama, running with a context window of 100k. Generating the diff works fine, good performance - but pressing "Apply" will take a long time then report

Token limit reached. File/range likely too large for this edit

Since I use this project for experimentation, it's extremely small with just a single file.

Confirming that context size is set correctly:

NAME               ID              SIZE     PROCESSOR          CONTEXT    UNTIL               
qwen3-coder:30b    06c1097efce0    41 GB    24%/76% CPU/GPU    100000     29 minutes from now