Dynamic reasoning/model switch per user request in Agno AgentOS via AG-UI — is this supported?

[paulsonkevgit]

---

🔴 Core Questions

I’m trying to confirm the following:

1. **Can users dynamically control reasoning and model selection per request/run? : when using AGUI interface

   AgentOS(interfaces=[AGUI(agent=...)])

2. If yes:

   • Is there an official or recommended pattern to dynamically configure the AGNO- agents(whether it should reason and which model should it use) based on AG-UI input passed from user ?

3. If not:

   • Can u tell me WHY its not possible and not expected to work like that ?
   • Is the Agent instance effectively fixed once initialized in AgentOS?
   • Does dynamic behavior require rebuilding/recreating the agent per request? (Which is Bad right)

4.Or only dynamic way is to let the agent decide based on the reasoining tools (https://docs.agno.com/reasoning/reasoning-tools)

5. Also saw the static way of reasoning flag set to true (specified in research section).But in that case always the reasoning would be done and the user doesn't decide it ryt?
   https://docs.agno.com/reasoning/overview
   https://docs.agno.com/reasoning/reasoning-agents

---

🎯MY Goals / (CLICK TO EXPAND )

I’m trying to build a UX similar to ChatGPT / Copilot: (HEY But MY BACKEND AGENT FRAMEWORK IS AGNO - i know its AGUI compatible and is hitting the AGUI endpoint from my front end)

• User can toggle Reasoning: ON / OFF
• User can switch Model: X / Y
• These changes should take effect per run, not just at application startup ryt ?

Expected behavior:

• Each /agui request reflects the user’s selection in how the agent behaves

---

## 🧩 My Setup (CLICK TO EXPAND )

• Using Agno as the agent framework
• Custom agents + team orchestrator (multi-agent setup)
• Integrating AG-UI via:

AgentOS(interfaces=[AGUI(agent=...)])

• Using only the /agui endpoint

AG-UI payload supports:

threadId
runId
parentRunId
state
messages
tools
context
forwardedProps

---

❓ Additional Architecture Question

If dynamic reconfiguration is not supported directly:

• What is the recommended approach?

• Is it only possible to:

  • Keep agent configuration fixed (decided at startup to my understandings)
  • Only reflect reasoning toggle at the UI level (e.g., show/hide reasoning output) ?

Or is there a better pattern to handle:

• Model switching
• Reasoning ON/OFF per user request

---

📚 My Research & Observations (from docs OF AGNO)

Reasoning Support in Agent

From docs:
https://docs.agno.com/reasoning/overview
https://docs.agno.com/reasoning/reasoning-agents

Reasoning can be enabled at the Agent constructor level:

from agno.agent import Agent
from agno.models.openai import OpenAIResponses
from agno.tools.hackernews import HackerNewsTools

reasoning_agent = Agent(
    model=OpenAIResponses(id="gpt-5.2"),
    tools=[HackerNewsTools()],
    instructions=["Use tables to display data"],
    reasoning=True,
    markdown=True,
)

reasoning_agent.print_response(
    "Compare the market performance of NVDA, AMD, and INTC over the past quarter. What are the key drivers?",
    stream=True,
    show_full_reasoning=True,
)

For non-native reasoning models:

What: Transform any regular model into a reasoning system through structured chain-of-thought processing via prompt engineering.

How it works: Set reasoning=True on any agent. Agno creates a separate reasoning agent that uses your same model (not a different one) but with specialized prompting to force step-by-step thinking, tool use, and self-validation.

Works best with non-reasoning models like gpt-4o or Claude Sonnet. With reasoning models like gpt-5-mini, you’re usually better off using them directly.

---

---

🙏 Final Ask

• If this is supported → a minimal working example would help a lot
• If not supported → clarity on why is it because the agent is initiated at startup only with reasoning flag and thereafter we can't reconfigure so that user can decide whether the agen should use reasoning or not.

Also:

Is it not a recommended pattern to allow:

• reasoning ON/OFF per user request
• model switching between interactions

---

Thanks in advance.