Skip to main content

​
CLI Commands

​
install

The install command downloads and installs MCP model servers from the Cogni+ registry.

​
Usage

cogni install [OPTIONS] MODEL_ID

​
Description

The install command downloads an MCP model server from the Cogni+ registry and installs it locally. This allows you to run the model on your own hardware, testing it before deployment or using it in development environments. The command handles:
  • Authentication with the Cogni+ registry
  • Downloading the appropriate container images
  • Setting up the local environment for the model
  • Configuring default parameters

​
Options

OptionDescription
--version TEXTSpecific version of the model to install (default: latest)
--forceForce reinstallation even if the model is already installed
--no-cacheSkip using cached downloads
--offlineUse only locally available resources (no downloads)
--helpShow help message for the install command

​
Arguments

ArgumentDescription
MODEL_IDThe ID of the model to install (e.g., “gpt4-mini” or “llama3-70b”)

​
Examples

# Install the latest version of a model
cogni install gpt4-mini

# Install a specific version
cogni install --version 1.2.3 gpt4-mini

# Force reinstallation of a model
cogni install --force gpt4-mini

# Install without using cache
cogni install --no-cache llama3-70b

​
Output

Installing model: gpt4-mini (latest)
Authenticating with Cogni+ registry...
Downloading model container (2.1GB)... 
[========================================] 100%
Setting up local configuration...
Installation complete!

To start this model:
  cogni server start
  cogni mcp start gpt4-mini

Model API will be available at:
  http://localhost:8080/v1/gpt4-mini/completions

​
Model Registry

The Cogni+ registry hosts a variety of MCP model servers, including:
  • Large language models (GPT-4, LLaMA, etc.)
  • Multimodal models (text-to-image, image-to-text)
  • Specialized domain models (code generation, scientific, etc.)
You can browse available models in the Cogni+ Platform UI or use the cogni list-models command.

​
Model Storage

Installed models are stored in:
  • Model configuration: ~/.cogni/models/
  • Container images: In your local Docker image repository

​
Tips

  • Ensure you have sufficient disk space before installing large models
  • Some models require specific hardware (e.g., NVIDIA GPUs)
  • Use --version to install specific model versions for reproducibility
  • Models are installed per-user, not system-wide