Model Center

Manage which local LLMs are active — TinyLlama, Claude, and more — and switch between models from one place.

Overview

Model Center is the single place to manage all language models available to TotalApp's AI features. It covers both cloud-hosted models (accessed via API key) and locally-running models (downloaded and served on the same machine). The active model set is what appears in the model selector dropdown in Agent Architect and Prompt Lab.

Screen preview

A quick look at the screen. Use the toggle in the corner to switch between the dark and light themes.

app.totalapp.app/orch-model-center

Models

Claude Sonnet 4.6Cloud · Active

Llama 3.2 3BLocal · Active

TinyLlama 1.1BLocal · Standby

GPT-4oCloud · Disabled

Model Center: cloud and local models

Cloud Models

Cloud models are accessed via API keys stored in TotalApp's settings. Currently supported providers:

Anthropic — Claude Sonnet, Claude Haiku, Claude Opus. Requires ANTHROPIC_API_KEY.
Cohere — Command R7B and Command R+. Requires COHERE_API_KEY.

Cloud model cards in the Model Center show their current availability status. A green dot means the API key is valid and the model is reachable. A red dot means the key is missing or invalid — open Settings → Writer Engine to enter the key.

Local Models

Local models run on the same machine as TotalApp's server. They require no API key and no internet connection once downloaded.

TinyLlama 1.1B — Fast, low RAM. Good for classification, routing decisions, and short summarisation tasks.

Use Local Models for Privacy-Sensitive Tasks

If your workflows process personally identifiable information or confidential documents, route those steps through a local model. No data leaves the machine.

Local Model Downloads

First-run downloads can be several gigabytes. Ensure the host machine has sufficient disk space and a stable network connection before initiating a download from the Model Center.

Activating / Deactivating Models

Toggle the switch on any model card to activate or deactivate it. Deactivated models do not appear in the Agent Architect or Prompt Lab model selectors. This helps keep the selector list short when you have many models installed but only use a few regularly.

Existing Agents Are Not Affected

Deactivating a model prevents it from being selected in new agents. Existing agents that already reference the deactivated model continue to work — the model is still available at runtime, just hidden from the selector UI.