---
title: "List inference calls"
method: POST
path: "/usage/inferences"
tags: ["usage"]
---

# List inference calls

`POST /usage/inferences`

Paginated, filterable list of inference calls (detect + locate) the caller's user was billed for. Filters: date range, endpoint, free-text search. Ordered by call time DESC.

## Request body

- UsageInferencesRequest — Query parameters for listing inference calls.
  - `$schema` string, uri — A URL to the JSON Schema for this object.
  - `end_date` string, date-time, required — End of the inferences query window.
  - `endpoint_filter` string[], nullable — Restricts results to the given vision endpoints ('detect'/'locate').
  - `limit` integer — Number of inferences per page.
  - `model` string — Model restricts results to a single model name.
  - `offset` integer — Pagination offset.
  - `search` string — Filters by inference (event) ID substring.
  - `session_id` string — Restricts results to inferences that ran under one phone session.
  - `sort_by` UsageInferenceSortSpec[], nullable — Ordered list of sort specs; first entry is primary.
    - `field` string, required — Column to sort by.
    - `order` string, required — Sort direction.
  - `start_date` string, date-time, required — Beginning of the inferences query window.

## Response `200`

OK

- UsageInferencesResponse — Paginated list of inference calls.
  - `$schema` string, uri — A URL to the JSON Schema for this object.
  - `inferences` UsageInference[], nullable, required — Page of inference-call records.
    - `api_key_id` string — API key that made the call.
    - `cost_microdollars` integer, required — Final post-markup cost charged for this call.
    - `created_at` string, date-time, required — Time the call completed, in UTC.
    - `endpoint` 'detect' | 'locate', required — Vision endpoint hit ('detect' or 'locate').
    - `inference_id` string, required — Unique inference-call identifier (inference_id).
    - `latency_ms` integer, required — End-to-end call latency in milliseconds.
    - `model` string, required — Canonical provider/model identifier that served the call.
    - `ocr_engine` 'free' | 'premium' — OCR engine used ('free' or 'premium'); empty for /locate.
    - `ocr_pages` integer, required — Number of Textract pages billed (premium OCR only).
    - `session_id` string — Device session this inference ran under, when driven by an allocation-scoped service token; empty for direct API-key calls.
    - `vlm_model` string — Semantic-find model used, when the call was a locate.
  - `limit` integer, required — Page size used for this response.
  - `offset` integer, required — Pagination offset used for this response.
  - `total` integer, required — Total number of inference calls matching the query.

## Other responses

- `default` — Error

---

[API](https://skmtc.net/axilioai/apis/axilio-api.md) · [All operations](https://skmtc.net/axilioai/apis/axilio-api/llms.txt) · [OpenAPI document](https://skmtc-service-staging.skmtc.workers.dev/v1/apis/axilioai/axilio-api/revisions/108ab4b41051/schema)