A signed go / no-go guard before your AI agent acts.

Humanoids are the highest-stakes acting class - bipedal, working among people - so they're the case we model deepest. Before a humanoid (or any agent) acts, one keyless call returns a signed go / caution / no-go for that class at that location, fused from wind (topple), precipitation (slip), temperature (ice + actuator limits), air quality (people nearby) and hazards. Gate the action on it; keep the signed verdict as proof of what the robot knew the moment it acted.

The call

The humanoid pre-operation preset - one keyless call returns a signed verdict (real example: a cold site where a humanoid gets 'caution' that a wheeled robot does not):

curl -s -X POST https://dynamicfeed.ai/v1/humanoid \
  -H 'Content-Type: application/json' \
  -d '{"location":{"lat":-54.8,"lon":-68.3},"radius_km":50}'
# → { "verdict": { "status": "caution",
#       "advisory": "-0.5C - near/below freezing; ice underfoot + cold-stressed actuators" },
#     "facts": {...}, "signature": {...} }   ·   same spot, class=ground → "go"

Sample response

# pip install requests  —  guard any humanoid (or agent) ACTION on a signed verdict
import functools, requests

def df_check(lat, lon, klass="humanoid", radius_km=25):
    """Signed go / caution / no-go + the grounded facts. Never hangs; a degraded source
    floors the verdict to 'caution' - never a false 'go'."""
    return requests.post("https://dynamicfeed.ai/v1/awareness", timeout=10, json={
        "robot": {"class": klass},                 # humanoid|ground|aerial|marine|orbital
        "location": {"lat": lat, "lon": lon}, "radius_km": radius_km}).json()

def require_clear(klass="humanoid"):
    def deco(act):
        @functools.wraps(act)
        def wrap(lat, lon, *a, **k):
            v = df_check(lat, lon, klass)
            if v["verdict"]["status"] == "no-go":
                raise PermissionError(v["verdict"]["advisory"])
            return act(lat, lon, *a, _awareness=v, **k)
        return wrap
    return deco

@require_clear("humanoid")
def step_into_workspace(lat, lon, _awareness=None):
    ...   # runs only if not no-go. Anchor _awareness via POST /v1/anchor for a permanent record.

# ROS 2? a ready node publishes the verdict to /dynamicfeed/preop_verdict:
#   clients/ros/humanoid_preop_guard.py  (repo: github.com/dynamicfeed/Dynamic-Feed)
# Evidence, not a safety guarantee — your robot's stack owns every action.

Why live data

Humanoid robots are the highest-stakes acting systems coming: bipedal, working in homes, hospitals and factories ALONGSIDE people - where a wrong action has a physical-human consequence. So DF models the humanoid class deepest. POST /v1/humanoid (a preset over /v1/awareness) returns ONE signed go / caution / no-go fused from wind tuned for topple (a tall biped falls where a wheeled robot doesn't), precipitation (wet/icy footing → slip), temperature (ice underfoot + actuator/battery envelope), air quality at a tighter limit (people nearby) and nearby earthquakes - under a hard deadline so it never blocks the robot, and fail-safe: a degraded input floors the verdict to 'caution', never a false 'go'. Live proof of the depth: at a near-freezing site a humanoid returns 'caution' (ice underfoot) while a wheeled ground robot at the same spot returns 'go'. Every verdict is Ed25519-signed (optionally Bitcoin-anchored), so you keep a tamper-evident record of exactly what the robot knew the moment it acted. To be precise: the signature is proof a verdict existed unchanged at a time - tamper-evidence, NOT a claim it is safe and NOT a safety certification. It is evidence; your robot's own stack owns every action. Full guide: dynamicfeed.ai/iot · verify a signature yourself at /playground.

Use it for

• Gate a humanoid action (step into a workspace, lift, hand off to a person) on a signed go/no-go
• Hold or slow when wind (topple), wet/icy footing (slip) or heat/cold cross the humanoid limits
• Floor to 'caution' automatically when a data source is degraded — never act on a false 'go'
• Drop in anywhere: require_clear('humanoid') in Python, or the ROS 2 pre-op node on-robot

FAQ

Related live feeds