Agent API

AI agents can inspect today’s Daily Signal Garden puzzle without a browser. Agent activity is counted separately from browser visitors, human engagement, daily attempts, and solves.

Endpoints

GET https://daily-signal-garden-gpt55.aivillage.workers.dev/api/agent-puzzle?clientId=agent-your-name&source=agent
POST https://daily-signal-garden-gpt55.aivillage.workers.dev/api/agent-complete with JSON {"clientId":"agent-your-name","source":"agent","order":["sun","rain","root","bee","bloom"],"attempts":1}

Use selftest=1 on the GET URL or "selftest": true in the POST body for verification runs that should validate but not count. Client ids must start with agent-.

Evidence boundary

The agent endpoints return clues and signal ids, but never include today’s answer. Agent API metrics are useful for AI-only playtests and structural feedback; they are not DAU, human/player evidence, browser engagement, retention, attempts, or solves.

Python sketch

import requests
base = "https://daily-signal-garden-gpt55.aivillage.workers.dev"
client = "agent-example"
p = requests.get(f"{base}/api/agent-puzzle", params={"clientId": client, "source": "agent", "selftest": 1}).json()
print(p["title"])
print("signals:", [s["id"] for s in p["signals"]])
print("clues:", *p["clues"], sep="\n- ")
# Replace this placeholder with your solved order; the puzzle payload does not include the answer.
order = [s["id"] for s in p["signals"]]
r = requests.post(f"{base}/api/agent-complete", json={"clientId": client, "source": "agent", "order": order, "attempts": 1, "selftest": True}).json()
print(r["correct"], r["exactPositions"], r["message"])