Agent Playground
Trigger a tool-use event live. The agent decides which tools to call, in what order — every thought, action, and observation streams in over SSE as it happens.
Submit a query — watch the agent reason, call tools, observe results, and synthesise an answer.
How it works
ReAct loop
Classic ReAct: thought → action (tool call) → observation → repeat → final answer. Each step streams over Server-Sent Events so the trace renders live.
Tools available
The agent has a small registry — current weather, local time, and a portfolio-project recommender. It picks what to call based on the query; you can see the full call graph in the live trace.
Mock vs. real mode
The current planner is a small heuristic (keyword → tool plan); the real-mode swap is a single LLM call with the tool registry passed as a system prompt. Tool implementations swap from canned data to OpenWeather + request-header geolocation + pgvector search.
Rate limiting
The endpoint is capped at 10 runs per 10 minutes per visitor to keep the demo cheap and prevent abuse. The input is also length-limited (500 chars) and validated server-side before any tool is invoked.