How to Use Local AI in IntelliJ with Ollama and Continue

In short: install Ollama, ollama pull qwen2.5-coder:7b (and a small autocomplete model), install Continue from the JetBrains Marketplace, point provider: ollama in config, reload, then chat. Honest note: Continue’s JetBrains plugin is community-maintained; Continue also recommends the Continue CLI (cn) beside your IDE — both are valid.

Why local AI in IntelliJ

Cloud assistants are great, but class assignments and client code sometimes must stay offline. Ollama runs the model on your machine. Continue is the bridge many JetBrains + VS Code users share with one config file.

What you need

• Ollama: macOS / Linux / Windows install from ollama.com. Enough RAM for the model (8GB minimum habit; 16GB+ more comfortable).
• Coding models: Example tags used in our VS Code lesson: qwen2.5-coder:7b for chat, qwen2.5-coder:1.5b for autocomplete — pull exact tags; they change.
• Continue plugin: JetBrains Marketplace search “Continue”. Same config family as VS Code’s Continue.continue.
• Optional CLI: Continue docs push active development toward the Continue CLI (cn) if the plugin feels stale.

How it connects

Install steps

1. Install Ollama. Check: ollama --version.
2. Start the service if needed (ollama serve on some setups). Check: curl http://localhost:11434 → “Ollama is running”.
3. Pull models with pull (not interactive run): ollama pull qwen2.5-coder:7b and ollama pull qwen2.5-coder:1.5b. Confirm with ollama list.
4. IntelliJ → Settings → Plugins → Marketplace → install Continue → restart.
5. Open the Continue tool window / sidebar.
6. Open config (gear → config.yaml). Paste an Ollama block (sample below). Model names must match ollama list exactly.
7. Reload config. Pick the Qwen chat model. Ask a tiny question.

Configuration

name: Local Config (JetBrains)
version: 0.0.1
schema: v1
models:
  - name: Qwen2.5-Coder 7B
    provider: ollama
    model: qwen2.5-coder:7b
    apiBase: http://localhost:11434
    roles:
      - chat
      - edit
      - apply
  - name: Qwen2.5-Coder 1.5B
    provider: ollama
    model: qwen2.5-coder:1.5b
    roles:
      - autocomplete

1. apiBase: http://localhost:11434 is the usual local URL.
2. Use a small model for autocomplete so typing stays snappy.
3. If IntelliJ cannot reach Ollama, check firewall / WSL networking; Continue logs under ~/.continue/logs/ often show the failure.
4. Same file can also list DeepSeek or xAI providers — pick per task.

Settings tips

1. Turn off autocomplete if Copilot + Continue both suggest at once.
2. Keep @-file context small — local models choke on huge pastes.
3. If the Marketplace plugin misbehaves, try Continue’s CLI (cn) in the IntelliJ terminal with the same config — Continue’s own JetBrains README steers active work there.
4. JetBrains AI Assistant can also talk to Ollama, but that path needs a JetBrains AI subscription (Ollama’s JetBrains page) — different product from Continue.

First task (5 minutes)

1. Apply to a new file only if the diff is clear.
2. Write 3 asserts yourself (or ask for tests, then edit).
3. Reject any request to read secrets.

Quick quiz

Save tokens and money

Local models are “free” of API fees but cost electricity and RAM. Use a 1.5B-class model for autocomplete; save larger models for hard chat.

Save these cards