# LLMs ## Description The LLMs panel lets you choose which large‑language model powers each agent. Agentic OS is model‑agnostic, so every tutor can run on the LLM that best fits its purpose—OpenAI GPT‑4 for nuanced writing help, Gemini for advanced reasoning, or even a custom model you integrate yourself. ![](/images/llms.png) --- ## Target Audience **Administrator** --- ## Features #### Per‑Agent Flexibility Assign different LLMs to different agents, tailoring performance, cost, and capabilities to each use case. #### Two Quick Access Paths Open the provider list either by clicking the **model name** on the agent card or by selecting the **LLM tab** from the agent dropdown. #### One‑Click Switching Pick a provider, choose a model, and see an immediate **Success** confirmation. #### Provider‑Agnostic Platform Supports **OpenAI, Google**, and other vendors—plus your own **custom integrations**. #### Extensible Model Library Add **new or proprietary LLMs** at any time; they appear alongside built‑in options for seamless selection. --- ## How to Use (step by step) #### Open LLM Settings (Method 1) - On the **agent card**, click the text showing the current LLM name - The provider list opens #### Open LLM Settings (Method 2) - Click the **agent’s name** to open its dropdown - Select the **LLM tab**—arrives at the same provider list #### Pick a Provider & Model - Click a provider (e.g., **OpenAI** or **Google**) - Select the desired model from the list - A **Success** message confirms the switch #### Repeat as Needed - You can **switch providers or models anytime** - Each change shows a **success confirmation** #### Add a Custom LLM (Optional) - If your preferred model isn’t listed, integrate it via the platform’s **custom LLM interface** - Once added, it appears with the built‑in providers and can be selected the same way --- ## Pedagogical Use Cases #### Domain‑Specific Tutors Connect a **healthcare agent** to a medically fine‑tuned model while keeping a **literature agent** on a more creative LLM. #### Cost Management Run high‑traffic, low‑stakes agents on a **budget‑friendly model** and reserve premium models for **advanced courses**. #### Experimental Research Quickly swap models to **compare answer quality, reasoning depth, or speed**—useful for instructional design studies. #### Language‑Focused Agents Choose a **multilingual model** for language courses, ensuring better translation and pronunciation guidance. #### Compliance & Privacy Integrate an **on‑premise or proprietary LLM** for sensitive data scenarios, keeping information within institutional boundaries. --- With simple, **per‑agent switching** and support for **custom models**, the **LLMs feature** ensures each tutor runs on the engine that best meets its educational goals.