Running Gencode NG completely offline with Ollama

For teams working on proprietary code, sending a single line to a cloud API is a non-starter. Gencode NG is built to accommodate that: pair it with Ollama and you can run open models like DeepSeek-R1 or Qwen-2.5-Coder entirely on your workstation.

Install Ollama

Download Ollama for macOS, Linux, or Windows. Once installed, pull a model with a single command: ollama pull qwen2.5-coder. The model runs locally against your CPU or GPU — nothing leaves your machine.

Point Gencode NG at your local runtime

Open your Gencode NG config file and add a provider block pointing at http://localhost:11434 with the model name you pulled. You can even assign a fast local model to Autocomplete and a heavier one to Chat using multi-model splitting.

Verify the air gap

Disconnect from the network and start typing. Autocomplete should keep firing, chat should keep responding, and inline diffs should keep rendering. That is a real offline AI coding loop — no bandwidth, no data exfiltration risk, and no vendor in the middle.