Configuration & Nicknames

NERVI stores settings inside a JSON configuration file. You can modify settings using CLI commands or by launching the interactive configuration wizard.

📄 Configuration File Schema (config.json)

The config file is located in your configuration directory:

  • Windows: %AppData%\Roaming\nervi\config.json
  • Unix: ~/.config/nervi/config.json

File Format

{
  "model_port": 22828,
  "model_name": "durgasai299792458/Qwen3-4B-OpenVINO-INT4-npu-i",
  "nicknames": {
    "qwen3-4b": "durgasai299792458/Qwen3-4B-OpenVINO-INT4-npu-i",
    "gemma": "durgasai299792458/gemma-4-E2B-OpenVINO-INT4"
  }
}

🛠️ CLI Configuration Command

Get Settings

./nervi config get <key>

Keys: model.port, model.name, nicknames, nickname.<alias>

Set Settings

./nervi config set <key> <value>

Examples:

./nervi config set model.port 11434
./nervi config set nickname.my-model meta-llama/Llama-3-8B-Instruct-OpenVINO-INT4

🏷️ Resolving Model Nicknames

Nicknames are resolved in the following priority order:

  1. User Overrides: Checks custom nicknames defined in the config.json file.
  2. Server Defaults: Fallback to hardcoded nicknames defined in the server code (e.g. qwen3-0.6b, qwen3-4b, gemma-4).
  3. Exact ID Match: Treats input as a fully-qualified repository path if no nickname matches.

🧙 Interactive Config Menu

You can launch a guided menu loop to configure NERVI settings interactively:

./nervi config