🌊 Namazu API

POST /building

Register a building to enable floor-by-floor tensor modeling and digital twin generation.

{
  "name": "Osaka City Hall Annex",
  "lat": 34.6937,
  "lon": 135.5023,
  "floors": 3,
  "construction_year": 1978,
  "structure_type": "wooden",  // wooden | rc | steel | src
  "sensors_per_floor": 3
}

GET /building/:id/tensor

Returns the current best structural tensor for a building — the inter-floor stiffness/damping coefficients computed by the leading TGL algorithm.

{
  "building_id": "bld_01HXYZ...",
  "computed_at": "2026-05-14T12:00:00Z",
  "tgl_id": "tgl_lagrangian_v4",  // which TGL produced this
  "floors": 3,
  "tensor": {
    "K": [[2.41, -1.20, 0.0],   // stiffness matrix
           [-1.20, 2.41, -1.20],
           [0.0, -1.20, 1.20]],
    "C": [[0.05, -0.02, 0.0],   // damping matrix
           [-0.02, 0.05, -0.02],
           [0.0, -0.02, 0.03]],
    "confidence": 0.87
  },
  "risk_class": "moderate"   // excellent|good|moderate|poor|critical
}

POST /tgl/sandbox

Check a GitHub repository for malicious code before submitting as a TGL. Clones the repo and runs static security analysis. HIGH severity issues block submission. MEDIUM issues are reported but do not block.

# Request
{ "repo_url": "https://github.com/you/your-tgl" }

# Clean response (200)
{ "status": "clean", "high": 0, "medium": 2, "message": "No high-severity issues found. TGL approved to run." }

# Flagged response (400)
{ "status": "flagged", "reason": "2 HIGH severity issue(s) found.", "high": [...] }

TGL Competition

TGL (Tensor Generating Logic) is Namazu's open competition for structural system identification algorithms. Submit your algorithm, have it tested against real multi-floor sensor data, and earn Namazu tokens if your TGL outperforms others at predicting inter-floor shake propagation.

Writing a TGL — Developer Guide

A TGL is a Python function that models the building as a coupled oscillator system. The equation of motion is:

# M·ẍ(t) + C·ẋ(t) + K·x(t) = f(t)
# Your job: identify K (stiffness) and C (damping) from sensor data.

Your file must expose a predict(shake, floors) function:

# my_tgl.py
import numpy as np

def predict(shake: list[float], floors: int) -> dict:
    """
    shake : flat float array — accelerometer samples (x,y,z at 100Hz, all floors merged)
    floors: number of floors in the building model
    returns: dict with keys 'tensor' (K matrix as nested list) and 'optimized_k' (float)
    """
    n = floors
    K = np.eye(n).tolist()           # replace with your fitted stiffness matrix
    C = (np.eye(n) * 0.05).tolist()  # replace with your fitted damping matrix
    k_opt = float(np.mean(np.diag(np.array(K))))
    return {"tensor": K, "optimized_k": k_opt, "damping": C}

POST /tgl/sandbox

Scan a GitHub repository for security issues before submitting. Flagged TGLs are still accepted — the scan is informational only.

{ "repo_url": "https://github.com/you/your-tgl" }

{ "status": "clean", "high": 0, "medium": 0, "message": "No high-severity issues found." }

POST /tgl/upload

Upload a single .py file directly — no GitHub account needed. Gets a tgl_id you can use in the Simulation UI or API.

curl -X POST https://api.namazu.zweichain.net/v3/tgl/upload/ \
  -F "file=@my_tgl.py" \
  -F "name=My TGL v1"

{
  "tgl_id": "tgl_01HXYZ...",
  "name": "My TGL v1",
  "sandbox": "clean",
  "usable": true,
  "message": "✓ Clean — ready to use."
}

POST /tgl/submit

{
  "name": "My Lagrangian TGL v1",
  "description": "Uses Lagrangian mechanics to identify modal frequencies...",
  "language": "python",
  "entrypoint": "tgl.predict",
  "code_url": "https://github.com/you/your-tgl",
  "docker_image": "ghcr.io/you/tgl:latest"  // optional
}

GET /tgl/leaderboard

{
  "updated_at": "2026-05-14T12:00:00Z",
  "entries": [
    {
      "rank": 1,
      "tgl_id": "tgl_lagrangian_v4",
      "name": "Lagrangian Modal Decomposition v4",
      "author": "namazu_core",
      "rmse": 0.0023,
      "events_tested": 1482,
      "tokens_earned": 12400
    }
  ]
}

Earthquake Alerts

Subscribe to real-time alerts via WebSocket. Alerts fire when the Namazu network detects a cluster of shake events consistent with an earthquake.

# Connect
wss://api.namazu.zweichain.net/v3/alerts?token=nmz_live_xxx&lat=34.69&lon=135.50&radius_km=100

# Alert message received:
{
  "type": "earthquake_alert",
  "alert_id": "alrt_01HXYZ",
  "issued_at": "2026-05-14T13:00:00.000Z",
  "epicenter": { "lat": 34.71, "lon": 135.49 },
  "depth_km": 12,
  "magnitude_estimate": 4.2,
  "intensity_at_location": 2.8,
  "seconds_to_arrival": 18,
  "confidence": 0.91
}

Error Codes