Add an official Agno AgentOS integration #1061

Is your feature request related to a problem? Please describe.

OpenUI users who run agents on Agno AgentOS currently need to write their own AG-UI normalization and AgentOS session storage glue. That makes it easy to create two sources of conversation truth or lose the identity relationship between the frontend thread and the AgentOS session.

Describe the solution you'd like

Add an official Agno integration as two focused, stacked changes:

1. A new @openuidev/agno package with:

   • createAgnoLLM() for AgentOS AG-UI endpoints
   • agnoAGUIAdapter() for Agno lifecycle and empty tool-parent normalization
   • agnoStorage() for AgentOS-backed thread CRUD and history
   • a deliberate 1:1 identity mapping where OpenUI threadId equals AgentOS session_id

2. A runnable examples/agno-chat application showing the complementary boundary:

   • AgentOS owns agents, teams, tools, memory, knowledge, sessions, authorization, execution, and deployment
   • OpenUI owns generation instructions, rendering, forms, charts, interactions, theming, and the chat shell
   • AG-UI is the streaming boundary
   • a generic thread-menu action seam exposes an "Open in AgentOS" session link after the first run materializes

The example should support both a deterministic no-key mock and a normal Python AgentOS using a real model credential supplied outside the repository.

Describe alternatives you've considered

A copy-paste-only example would demonstrate the happy path but leave every consumer to maintain the same adapter and session code. A package without an example would be harder to validate and would not demonstrate charts, follow-ups, form submissions, backend tools, session persistence, or operational handoff.

Additional context

Agno already merged an OpenUI AgentOS AG-UI cookbook example: agno-agi/agno#9605

The proposed implementation has been exercised locally against a live AgentOS and model as well as the deterministic harness.