cd /news/developer-tools/local-ai-code-completion-in-vs-code-… · home topics developer-tools article
[ARTICLE · art-88016] src=sourcefeed.dev ↗ pub= topic=developer-tools verified=true sentiment=· neutral

Local AI Code Completion in VS Code with Qwen2.5-Coder and Ollama

A tutorial by Rachel Goldstein explains how to set up a fully local AI coding environment in VS Code using Qwen2.5-Coder models (1.5B and 7B) run through Ollama and the Continue extension, enabling private, offline tab autocomplete and chat. The setup requires at least 8 GB of RAM, about 6 GB of disk space, and was verified against Ollama 0.32.5 and Continue extension 2.1.0. The guide includes steps to install Ollama, pull the models, configure Continue, and verify the setup, emphasizing that no code leaves the machine.

read5 min views1 publishedAug 5, 2026
Local AI Code Completion in VS Code with Qwen2.5-Coder and Ollama
Image: Sourcefeed (auto-discovered)

Wire Qwen2.5-Coder through Ollama and Continue for private, offline autocomplete and chat in VS Code.

Rachel Goldstein

What you'll build #

A fully local AI coding setup in VS Code: Qwen2.5-Coder running under Ollama, wired into the Continue extension for tab autocomplete and inline chat. Nothing leaves your machine — it works on a plane, and proprietary code never touches a third-party API.

Prerequisites #

  • macOS, Linux, or Windows with at least 8 GB of RAM (16 GB is comfortable if you want the 7B chat model alongside your editor)
  • About 6 GB of free disk for the two models
  • A current VS Codeinstall with thecode

CLI on your PATH - Verified against Ollama 0.32.5, Continue extension 2.1.0, and the qwen2.5-coder

tags on the Ollama library, August 2026

1. Install Ollama #

Ollama is the local model server everything else talks to. On macOS or Windows, download the installer from ollama.com/download and run it — the app starts a server on localhost:11434

and keeps it running in the background. On Linux:

curl -fsSL https://ollama.com/install.sh | sh

The script installs Ollama and registers a systemd service, so the server is already running. Confirm:

curl http://localhost:11434

You should get back Ollama is running

.

2. Pull the Qwen2.5-Coder models #

You want two models for two jobs: a small one for autocomplete, where latency matters more than brains, and a bigger one for chat and edits.

ollama pull qwen2.5-coder:1.5b
ollama pull qwen2.5-coder:7b

The 1.5B is a 986 MB download; the 7B is 4.7 GB. Always pin the size tag — a bare qwen2.5-coder

resolves to :latest

, which is the 7B, and that's too slow for autocomplete on most laptops.

3. Install the Continue extension #

Install Continue from the VS Code Marketplace (search "Continue", publisher Continue), or from the terminal:

code --install-extension Continue.continue

On first launch Continue creates its config file and adds a chat icon to the sidebar — you can skip any sign-in it offers, since local models need no account.

4. Point Continue at your local models #

Continue reads ~/.continue/config.yaml

(%USERPROFILE%\.continue\config.yaml

on Windows). Open it directly in your editor, or from Continue's chat sidebar (Cmd/Ctrl+L

): click the agent selector above the input and hit the gear icon next to your local config. Replace the models

section so the file looks like this:

name: Local Assistant
version: 1.0.0
schema: v1

models:
  - name: Qwen2.5-Coder 7B
    provider: ollama
    model: qwen2.5-coder:7b
    roles:
      - chat
      - edit
      - apply
  - name: Qwen2.5-Coder 1.5B
    provider: ollama
    model: qwen2.5-coder:1.5b
    roles:
      - autocomplete

Save the file — Continue reloads automatically, no restart needed. It assumes Ollama's default address; you'd only add an apiBase

if Ollama runs on another machine or port. Finally, make sure VS Code's editor.inlineSuggest.enabled

setting is on (it is by default), and disable GitHub Copilot if you have it installed, so two extensions aren't fighting over the same ghost text.

Verify it works #

First check both models are in place:

ollama list
NAME                  ID              SIZE      MODIFIED
qwen2.5-coder:7b      2b0496514337    4.7 GB    2 minutes ago
qwen2.5-coder:1.5b    d7372fb82b10    986 MB    4 minutes ago

(Your ID

values will differ.) Then in VS Code, create test.py

and type def fibonacci(

— , and gray ghost text should appear with a suggested completion. Press Tab

to accept it. The very first completion takes a few seconds while Ollama loads the model into memory; after that it's near-instant. Running ollama ps

in a terminal should now show qwen2.5-coder:1.5b

loaded.

For chat, press Cmd/Ctrl+L

, pick Qwen2.5-Coder 7B in the model dropdown, and ask it to explain the file. Final proof: turn off Wi-Fi and do it all again — everything still works.

Troubleshooting #

** Error: listen tcp 127.0.0.1:11434: bind: address already in use** — you ran

ollama serve

while the server was already running (the desktop app and the Linux systemd service start it for you). Don't run serve

manually; just use ollama pull

and friends.A Continue error ending in not found, try pulling it first — your config names a tag that isn't downloaded. Continue never pulls models itself. Run

ollama list

and make the model:

values match the listed tags exactly, size suffix included.Ghost text never appears — autocomplete may be d: click the Continue item in VS Code's status bar and re-enable it. Then confirm editor.inlineSuggest.enabled

is true

and that Copilot's inline suggestions are off.

Completions show up but take seconds — the autocomplete role is on a model that's too big, or you're low on RAM. Keep autocomplete on the 1.5B; on older hardware drop to qwen2.5-coder:0.5b

, and use ollama ps

to see what's actually loaded.

Next steps #

If you've got the hardware (24 GB+ RAM or a decent GPU), swap the chat role to qwen2.5-coder:14b

or :32b

— same config shape, better answers. Add codebase-aware chat by pulling nomic-embed-text

and giving it the embed

role, which powers Continue's @Codebase

context. And when completions feel too eager or too sluggish, tune debounceDelay

and maxPromptTokens

— the autocomplete deep dive covers both.

Sources & further reading #

Continue Autocomplete Setup and Configuration Guide— docs.continue.dev - How to Configure Continue— docs.continue.dev - config.yaml Reference— docs.continue.dev - qwen2.5-coder model library page— ollama.com - Download Ollama— ollama.com - Ollama Releases— github.com

Rachel Goldstein· Dev Tools Editor

Rachel has been embedded in the developer tooling ecosystem for nearly eight years, covering everything from IDE wars and package-manager drama to the quiet rise of AI-assisted coding. She has a soft spot for open-source maintainers and an unhealthy number of terminal emulators installed on a single laptop.

Discussion 0 #

No comments yet

Be the first to weigh in.

── more in #developer-tools 4 stories · sorted by recency
── more on @rachel goldstein 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/local-ai-code-comple…] indexed:0 read:5min 2026-08-05 ·