Call the World Token Factory API to run geospatial risk assessments on any business. Decomposes businesses into operational steps, identifies risk factors, and runs multi-depth AI analysis returning structured risk tokens with failure rates, uncertainty scores, loss ranges, and cited evidence.
npx @senso-ai/shipables install wingdingspenpal/world-token-factoryUse this skill to assess geospatial and operational risk for any business via the World Token Factory API.
https://king-prawn-app-kbwtp.ondigitalocean.app
Pass an API key via X-API-Key header. Get a key from the World Token Factory dashboard.
X-API-Key: your_key_here
Break any business into operational steps and risk factors.
POST /api/decompose
Content-Type: application/json
{
"description": "Gulf Coast oil pipeline operator focused on Permian Basin extraction",
"max_steps": 5
}
Response: NDJSON stream of steps, each with risk factors and initial metrics (failure_rate, uncertainty, loss_range).
Run AI analysis on a specific risk factor at three depth levels.
POST /api/analyse
Content-Type: application/json
{
"risk_factor_id": "rf_1_1",
"risk_factor_name": "ERCOT Grid Failure Risk",
"business_context": "Gulf Coast Oil Operator",
"step_context": "Permian Field Operations",
"depth": 2,
"data_domains": ["oil"]
}
Depth levels:
1 — Quick Scan: filename matching, ~350 tokens, instant2 — Research Brief: reads source docs + Senso RAG, ~3k tokens, 30-60 seconds3 — Deep Run: parallel agents, multimodal (GeoTIFF, PDF, video), ~200k tokensResponse: NDJSON stream with events: step, signal, file_found, token_update, complete.
Fan out to all data sources in parallel, synthesize with model-routed LLM.
POST /api/orchestrate/analyse
Content-Type: application/json
{
"business_name": "Gulf Coast Oil Operator",
"step_name": "Permian Field Operations",
"risk_factor_name": "ERCOT Grid Failure Risk",
"risk_factor_description": "Power curtailment to field compressors during ERCOT stress events",
"domain": "oil",
"lat": 31.5,
"lng": -102.5,
"depth": 2
}
What it does:
Response: NDJSON stream with reasoning chain, evidence sources, risk metrics.
Conversational risk assessment interface.
POST /api/chat
Content-Type: application/json
{
"prompt": "What are the top 3 risks for a warehouse near the Mississippi River?",
"system_message": "You are a World Token Factory risk analyst.",
"max_tokens": 2048,
"temperature": 0.3
}
See which AI model would be selected for a given prompt.
POST /api/model-route
Content-Type: application/json
{
"prompt": "Assess seismic risk for Permian Basin pipeline infrastructure",
"system_message": "risk analysis"
}
Response: { "task_type": "geospatial", "selected_model": "kimi-k2.5", "reason": "..." }
GET /api/health
GET /api/sponsor-status
GET /api/orchestrate/systems
GET /api/models
Every analysis produces a structured Risk Token:
{
"risk_factor_id": "rf_1_1",
"summary": "Cited analysis with specific numbers...",
"gaps": ["Specific knowledge gap 1", "Gap 2"],
"metrics": {
"failure_rate": 0.22,
"uncertainty": 0.76,
"loss_range_low": 48000000,
"loss_range_high": 360000000,
"loss_range_note": "Wide range reflects ERCOT event duration uncertainty"
},
"artifacts": [
{"filename": "permian_basin_risk_brief.md", "type": "document", "relevance": "..."}
],
"tokens_used": 3200,
"depth": 2
}
Key insight: failure_rate and uncertainty are SEPARATE metrics. A 5% failure rate with 80% uncertainty is very different from 5% with 10% uncertainty. The width of the loss range is itself a signal about information quality.
import httpx
API = "https://king-prawn-app-kbwtp.ondigitalocean.app"
KEY = "your_unkey_api_key"
# Step 1: Decompose
r = httpx.post(f"{API}/api/decompose",
headers={"X-API-Key": KEY, "Content-Type": "application/json"},
json={"description": "Solar panel installation company in Florida"})
# Step 2: Analyse each risk factor
for step in r.json().get("steps", []):
for rf in step["risk_factors"]:
analysis = httpx.post(f"{API}/api/analyse",
headers={"X-API-Key": KEY, "Content-Type": "application/json"},
json={
"risk_factor_id": rf["id"],
"risk_factor_name": rf["name"],
"business_context": "Solar installer, Florida",
"step_context": step["name"],
"depth": 2,
"data_domains": ["general"]
})
print(f"{rf['name']}: {analysis.json()}")
Railtracks (agent orchestration) · Senso (RAG knowledge base) · Nexla (data pipelines) · DigitalOcean (hosting + Gradient AI inference) · Unkey (API key management) · Augment (code context) · Google Gemini (LLM) · assistant-ui (chat interface)