FluxyChat Docs

The agent platform is operated through the dashboard (/agents) and Worker admin APIs. LLM keys live in Worker secrets — not in client code.

Create an agent (admin JWT)

curl -X POST "$FLUXY_BASE_URL/agents" \
  -H "Authorization: Bearer $ADMIN_JWT" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Support Bot",
    "handle": "support",
    "provider": "openai",
    "model": "gpt-4o-mini",
    "capabilities": ["chat"]
  }'

Profile fields

Field	Purpose
`handle`	Mention trigger (`@support`)
`provider` / `model`	LLM routing on Worker
Instructions	System prompt templates
`toolExecuteUrl`	HTTP endpoint for tool calls
Fallback model	Secondary provider on failure

Chat in room

Dashboard Agents → Chat in room opens a live WebSocket session (member JWT required). Run banner shows runId, latency, and token usage.

SDK invoke

await client.invokeAgentRest(agentId, roomId, "Help with billing");
const handoff = await client.getRoomHandoff(roomId);

Versioning workflow

Use dashboard drafts for prompt changes. Test in assistant:general or a staging room before pointing production traffic at a new profile.

Platform client (advanced):

import { createWorkerAgentPlatformClient } from "@fluxy-chat/sdk";

See Agents and Assistant room.

Agent platform