Malformed JSON error when cursor movement cancels autocomplete streaming with OpenAI provider

[fsytsuta]

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: Windows 11
- Continue version: 1.0.24
- IDE version: VSCode 1.103.1
- Model: Qwen/Qwen3-Coder-30B-A3B-Instruct (Locally running on vllm)
- config:
  
name: Local Assistant
version: 1.0.0
schema: v1
models:
  - name: Qwen3-Coder-30B-A3B-Instruct
    provider: openai
    model: Qwen/Qwen3-Coder-30B-A3B-Instruct 
    roles:
      - autocomplete
    apiBase: http://localhost:8000/v1
    apiKey: EMPTY

Description

When using the OpenAI provider for autocomplete functionality, moving the cursor during active streaming causes incomplete JSON responses to be processed, resulting in Malformed JSON sent from server errors.

To reproduce

1. Configure Continue extension with an OpenAI provider model (e.g., Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8)
2. Enable autocomplete functionality in VSCode
3. Trigger code completion and move cursor while streaming is in progress
4. Observe the following error:

Continue Autocomplete Error: Malformed JSON sent from server: {"id":"chatcmpl-...","object":"text_completion",...,"usage":{"completion_tokens":0,"prompt_token

Log output

[fsytsuta]

I analyze this issue using AI.

Note: This analysis was conducted with limited technical expertise and AI assistance. The root cause analysis and proposed solutions may contain inaccuracies and should be verified by experienced developers.

Root Cause Analysis

The issue occurs in the streamSse function implementation stream.ts:124-149 . When cursor movement triggers request cancellation via AbortSignal, the streaming response is interrupted mid-transmission. The buffer containing incomplete JSON data is then passed to parseSseLine → parseDataLine, causing JSON parsing errors stream.ts:76-108 .

The OpenAI provider's FIM streaming implementation OpenAI.ts:141-169 and the core OpenAI class implementation OpenAI.ts:371-403 both rely on streamSse for processing responses, making them susceptible to this issue.