POST /api/v1/investigate

A full investigation of one picked entity: deterministic registry and domain checks, then agent-driven exploration of additional sources within a step allowance. Typical latency 10–40 seconds.

Request body

Address the entity one of three ways — whichever your system already holds:

// 1 — our id, from a prior search { "entityId": "denue-4325279" } // 2 — official registration number: skip search entirely { "country": "BR", "registrationNumber": "07918048" } // 3 — registry + ref, round-tripped from any search result { "registry": "INEGI DENUE", "registryRef": "4325279" }

FIELD	DESCRIPTION
`entityId` / `country`+`registrationNumber` / `registry`+`registryRef`	The entity address — exactly one form. The response echoes `resolvedBy`.
`context`	Optional. What you already know: `{website, email, address, note}`. The investigation verifies against it — it is treated as claims, not truth.
`focus`	Optional. Areas to prioritize, e.g. `["tax_status"]`.
`goals`	Optional. Why you are checking: any of `compliance`, `risk`, `partnership`, `employer`, `contractor`. Defaults to `["compliance"]`. Same goals as the app; decides which bonus balance, if any, the investigation draws from before your monthly allowance.
`budget`	Optional. Exploration steps to allow; capped by your plan.

Response

FIELD	DESCRIPTION
`summary`	Plain-language conclusion, written from the consulted sources.
`confidence`	`high \| medium \| low` — about coverage and corroboration.
`riskScore`	0–100. Computed deterministically from the findings — same findings, same score.
`complianceScore`	0–100. Computed from failed/review checks the same way.
`coverage`	`full` — checks + exploration ran · `baseline_only` — the exploration layer was unavailable; findings contain only the deterministic checks.
`findings[]`	The investigation. Each: `section` · `finding` (plain language) · `risk` LOW/MED/HIGH · `compliance` PASS/REVIEW/FAIL · `sources[]` · `input` (exactly what was queried) · `origin` baseline/exploration.
`notChecked[]`	Sources not consulted, each with the investigator's stated `reason`.

Example

curl -X POST -H "x-api-key: vk_live_…" -H "Content-Type: application/json" \ -d '{"entityId":"denue-7785482","context":{"website":"grupobimbo.com"}}' \ https://app.verety.ai/api/v1/investigate { "entityId": "denue-7785482", "country": "MX", "entity": { "name": "GRUPO BIMBO" }, "summary": "Small establishment (0–5 employees) in Arriaga, Chiapas…", "confidence": "high", "riskScore": 20, "complianceScore": 70, "coverage": "full", "findings": [ { "section": "Business domain", "risk": "MED", "compliance": "REVIEW", "sources": ["ENTITY CONTACT INFO ON FILE"], "input": null, "finding": "No business email domain on file for this establishment.", "origin": "baseline" }, { "section": "Website content", "risk": "LOW", "compliance": "PASS", "sources": ["LIVE WEBSITE FETCH"], "input": "grupobimbo.com", "finding": "grupobimbo.com is the official site of the global Bimbo corporation…", "origin": "exploration" } ], "notChecked": [ { "capability": "lookup_wayback", "reason": "Domain age already confirmed by WHOIS — archive history would not change the conclusion." } ] }

Nothing is fabricated. Findings come only from consulted sources. When a source is down you get fewer findings and an honest coverage flag — never invented data.