AI Agent
Reason, call a tool, stream the answer.
A research lab exploring authenticated, streamed agent conversations and real-time tool use.
A research system. Public runtime access is provided only when approved.
> system.status
ACTIVE_LAB
> capability.scan
STREAMING RESPONSES
TOOL SELECTION
LOCAL MODEL OPTION
AUTHENTICATED SESSIONSThe idea
This lab combines a Next.js conversational client with a FastAPI agent service. Responses stream using server-sent events while a graph-based agent selects tools and composes an answer.
The current demonstration includes current, historical, and forecast weather tools, a locally hosted language model, Redis-backed session infrastructure, and Markdown output.
Product system
Designed around the real loop.
Features are built to reinforce the core experience—not compete with it.
Streaming responses
Server-sent events make intermediate answer progress visible without waiting for a completed payload.
Tool selection
A ReAct-style graph lets the model choose current, historical, or forecast data tools.
Local model option
The current agent can run against a locally hosted model rather than requiring a public model endpoint.
Authenticated sessions
JWT and Redis support protected conversation sessions and service state.
Answers / AI Agent
Good questions,
clear answers.
Is this a consumer product?+
No. It is a public-facing lab used to demonstrate and evaluate agent architecture.
Is the demo always available?+
Continuous public availability is not promised. A live link will appear only after uptime and abuse controls are approved.