v1.6.0 — Local AI Tooling

Local AI Tooling feature

Adds the local-ai devcontainer feature — a single opt-in addition that brings a full local AI coding platform into any devcontainer.

What's included

Tool	Description
Aider	AI pair programming in the terminal
Codex CLI	OpenAI terminal agent via local Ollama
GitHub Copilot CLI	GitHub Copilot in the terminal
Claude Code	Anthropic agentic coding assistant
Continue.dev	VS Code AI code completion and chat

Usage

Add to your project's .devcontainer/devcontainer.json:

{
  "image": "ghcr.io/dever-labs/devcontainers/dotnet-dev:latest",
  "features": {
    "ghcr.io/dever-labs/devcontainers/features/local-ai:1": {}
  }
}

The feature automatically handles:

• ✅ Installing all AI tools at image build time
• ✅ Detecting host Ollama and writing tool configs on container start
• ✅ Pulling default models (qwen2.5-coder:14b, deepseek-coder:16b) in the background
• ✅ Shared credential volumes so sessions persist across rebuilds and projects
• ✅ VS Code extensions (Copilot, Claude Code, Continue.dev)

Prerequisites

Ollama must be running on the host. Add to your devcontainer.json:

"initializeCommand": "bash -c 'curl -sf --max-time 1 http://localhost:11434/api/tags >/dev/null 2>&1 || (command -v ollama >/dev/null 2>&1 && nohup ollama serve >/tmp/ollama-serve.log 2>&1 &) || true'"

See features/src/local-ai/README.md for full documentation.