Summary
A complete setup guide for running Qwen 2.5 Coder locally via Ollama and connecting it to OpenCode (an open-source AI coding CLI) to create a private, offline-capable AI pair programmer in your terminal.
What I Did
- Install Ollama using provided commands for macOS/Linux or Homebrew.
- Pull the Qwen 2.5 Coder model from Ollama with
ollama pull qwen2.5-coder. - Install OpenCode via npm or Homebrew.
- Configure OpenCode's global config file to use Qwen locally by setting up a JSON configuration.
- Increase
num_ctxin the config for better performance and larger working memory. - Start Ollama with
ollama servebefore running OpenCode. - Use OpenCode commands directly in your project directory.
Key Technical Findings
- Proper configuration of the context window (
num_ctx) is crucial for optimal performance. - Ollama and OpenCode provide a powerful, private AI coding solution without cloud dependencies.
Commands Used
Install Ollama: `curl -fsSL https://ollama.com/install.sh | sh` or `brew install ollama`
Pull Qwen model: `ollama pull qwen2.5-coder`
Install OpenCode: `npm install -g opencode-ai` or `brew install sst/tap/opencode`
Verify Ollama: `ollama --version`
List models: `ollama list`
Verify OpenCode: `opencode --version`
Start Ollama server: `ollama serve`
Run OpenCode in project directory: `cd ~/projects/my-project && opencode`
Next Steps
Install and configure Ollama. Pull the Qwen 2.5 Coder model. Install and configure OpenCode. Test the setup with various coding tasks.