Autocomplete for code is not working

[Magnarks]

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 11
- Continue version: 1.2.22
- IDE version: 1.118.1 VS Code
- Model: Qwen-Coder 2.5
- config:
  
name: Local Config
version: 1.0.0
schema: v1
models:
  - name: Local Chat
    provider: ollama
    model: qwen2.5-coder:7b
    roles:
      - chat
  - name: Local Autocomplete
    provider: ollama
    model: qwen2.5-coder:1.5b
    roles:
      - autocomplete
  - name: Autodetect
    provider: ollama
    model: AUTODETECT
  
  OR link to agent in Continue hub:

Description

Autocomplete for code is not working, Code autocomplete isn't working; it simply does nothing. It doesn't show any code suggestions. I disabled Copilot, but that didn't work either; it still doesn't show any suggestions as I type code. I've tried everything, but nothing makes it work. I've even tried small models, even though my system can run larger ones, and it still doesn't work.

https://youtu.be/lBG_1x52X-E

To reproduce

It simply doesn't work as shown in the video.

https://youtu.be/lBG_1x52X-E

Log output

[Ayushraj6202]

Can you please provide screenshot of log in developer tools?

[RoboMico]

Also having the same issue, attempting to use Deepseek V4 Flash as autocomplete model.

Log from developer tools: vscode-app-1778289229131.log

OS: Linux (CachyOS)
VSCode 1.119.0
Continue extension: 1.2.22
Continue config:

name: Local Config
version: 1.0.0
schema: v1
models:
  - name: DeepSeek V4 Pro
    provider: deepseek
    model: deepseek-v4-pro
    apiKey: *****
    apiBase: https://api.deepseek.com/beta
    roles:
      - chat
      - edit
      - apply
      - summarize
      - autocomplete
    contextLength: 1024000
    defaultCompletionOptions:
      temperature: 0.0
      maxTokens: 256
  - name: DeepSeek V4 Flash
    provider: deepseek
    model: deepseek-v4-flash
    apiKey: *****
    apiBase: https://api.deepseek.com/beta
    roles:
      - chat
      - edit
      - apply
      - summarize
      - autocomplete
    contextLength: 1024000
    defaultCompletionOptions:
      temperature: 0.0
      maxTokens: 256
  - name: Embedding Gemma
    provider: ollama
    model: embeddinggemma:latest
    apiKey: ollama
    roles: [embed]
context:
  - provider: code
  - provider: docs
  - provider: diff
  - provider: terminal
  - provider: problems
  - provider: folder
  - provider: codebase

[userpasta]

I'm having the same issue with local autocomplete. No prompt is being sent to the autocomplete model.

Kogs from "Continue Console":

Type: Complete
Result: Success
Prompt Tokens: 0
Generated Tokens: 0
ThinkingTokens: 0
Total Time: 0.03s

{
  "model": "local-fim-model",
  "maxTokens": 4096,
  "temperature": 0.01,
  "stop": [
    "<fim_prefix>",
    "<fim_suffix>",
    "<fim_middle>",
    "<file_sep>",
    "<|endoftext|>",
    "</fim_middle>",
    "</code>",
    "/src/",
    "#- coding: utf-8",
    "```"
  ],
  "raw": true
}

Also getting a lot of not found in AST tracker console errors (might be red herring):

[Extension Host] Document /Users/<redacted>/Projects/scripts/cond_example.go not found in AST tracker
log.ts:117   ERR [Extension Host] Document /Users/<redacted>/Projects/scripts/cond_example.go not found in AST tracker

Tried the following:

1. All combinations of VS Code 1.119 and 1.120 with Continue 1.2.22 and 1.3.38 with no luck.
2. Using llama.cpp and openai as a provider in the config.yaml.

[DeruiZhao]

I have the same issue as @RoboMico.

I checked DeepSeek's documentation, which says that the "thinking" mode is enabled by default. To use autocomplete, this mode must be disabled. According to the docs, disabling it requires passing a parameter like extra_body={"thinking": {"type": "disabled"}}.

However, it seems that Continue does not support the extra_body parameter. Could this be the root cause of the problem?

[userpasta]

not found in AST tracker was a red herring.

I was able to fix my issue by extending contextLength from 4096 to 8192 in my config.yaml.