Model Use Track

The track for actually using the Global Model built by federated learning. You attach the model to an LLM Agent as a tool (Tool AI), open it as an HTTP API, and call it from your app.

Prerequisite: you have finished Owner Track chapter 7 and have Global Model v1. (You can also skip federated learning and use the Initial Model from Owner Track chapter 4.) This track assumes the Owner account. Whether a Client account can do the same has not been verified yet.

flowchart LR
  G["Global Model v1"] --> A["1 Build Agent<br/>LLM + Tool AI + rules"] --> B["2 Serving API<br/>port + token"] --> C["3 Plug into app"]
Base LLM (language)  +  Tool AI (my model)  +  Harness (rules)  =  Agent

Numbers come from the model, sentences from the LLM. In the Agent’s answer, the prediction (no disease / disease probabilities) comes from the Tool AI built by federated learning, and the LLM only explains those numbers in plain words. This principle runs through the whole track.

#PageWhere
1Build an AI AgentStudio Agent Builder
2Call It over an APIStudio Agents + terminal
3Demo App · Plug into Your ProjectBrowser + your code

Table of contents


Back to top

© 2026 Cognitive Computing Lab, Gachon University