> ## Documentation Index
> Fetch the complete documentation index at: https://docs.near.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Recompute usage_hourly for a UTC window (Admin only)

> **Deprecated:** this endpoint is moving to cloud-platform at the same path; the cloud-platform base URL will be announced. Behaviour may differ in edge cases.

Rebuilds `[start, end)` from raw usage one UTC day per transaction, then reports raw vs
aggregate parity for every day the window touches. Parity always covers the whole UTC day,
so a window covering part of a day can report `ok: false` from hours outside it; repair
whole days to clear a day's parity. Use it after a backfill or when the nightly parity check
warns; the scheduler only re-reads the last 3 hours. A window must not start past the next
hour the scheduler has yet to compute (409), or the hours in between would never be computed.

Admin permissions: admin session or read_write token. A read_only token receives HTTP 403 with error.type insufficient_permissions before the handler runs.



## OpenAPI

````yaml /api-reference/openapi.json post /v1/admin/usage-hourly/recompute
openapi: 3.1.0
info:
  title: NEAR AI Cloud API
  description: >-
    NEAR AI Cloud API for private AI model inference and organization
    administration.
  contact:
    name: NEAR AI Team
    email: support@near.ai
  license:
    name: MIT
  version: 1.0.0
servers:
  - url: https://cloud-api.near.ai
    description: NEAR AI Cloud
security:
  - session_token: []
  - api_key: []
tags:
  - name: Chat
    description: Chat completion endpoints for AI model inference
  - name: Images
    description: Image generation endpoints
  - name: Decisions
    description: Typed decisions through the System One protocol
  - name: Audio
    description: Audio transcription endpoints
  - name: Rerank
    description: Document reranking endpoints
  - name: Score
    description: Text similarity scoring endpoints
  - name: Privacy
    description: Privacy classification (PII span detection) endpoints
  - name: Models
    description: Public model catalog and information
  - name: Conversations
    description: >-
      Temporary authenticated, workspace-scoped migration access for export and
      account deletion. Only `POST /v1/conversations/batch`, `GET
      /v1/conversations/{conversation_id}`, `GET
      /v1/conversations/{conversation_id}/items`, and `DELETE
      /v1/conversations/{conversation_id}` are available. Conversation creation
      and every other mutation return `410 Gone`; this surface will be removed
      after data migration. Migration responses use `Cache-Control: no-store`.
  - name: Files
    description: >-
      Temporary authenticated, workspace-scoped migration access for export and
      account deletion. Only `GET /v1/files`, `GET /v1/files/{file_id}`, `GET
      /v1/files/{file_id}/content`, and `DELETE /v1/files/{file_id}` are
      available. Upload and every other mutation return `410 Gone`; this surface
      will be removed after data migration. Migration responses use
      `Cache-Control: no-store`.
  - name: Responses
    description: >-
      Stateless response inference. Both paths normalize omitted or null `store`
      to `false` and reject `store: true`. Raw request/response content,
      response items, and history are not persisted. Clients must include any
      prior context in each request. Canonical `openai/` models use one upstream
      native Responses request after alias resolution. Other models use the
      typed adapter, with bounded request-scoped inference rounds when hosted
      search is requested.


      Both paths accept client-managed `function` and `custom` tools, including
      tools grouped in a `namespace`, and leading developer `additional_tools`
      declarations. Cloud returns calls without executing them; clients replay
      each `function_call` or `custom_tool_call` with its matching output and
      caller-managed history. Preserve `call_id`, names, namespaces, raw
      arguments or custom input, and provider metadata such as
      `thought_signature`. The typed replay path accepts assistant `output_text`
      parts and plaintext `reasoning` items, using reasoning content or its
      summary as prior assistant context. Encrypted-only reasoning requires a
      native Responses backend. The typed adapter supports text custom tools and
      the Codex `exec` Lark source grammar; other grammars require a native
      backend.


      Each generated item has its own `output_index`. Tool calls finish with
      `response.output_item.done`; a successful model turn ends with
      `response.completed`, including when tools await client execution.
      Reasoning uses `response.reasoning_text.delta`; failures expose
      `response.error` in `response.failed` and preserve partial usage.


      The typed adapter executes `web_search` and `web_context_search` using a
      bounded request-scoped loop, emits search lifecycle events and URL
      citations, and accumulates usage across inference rounds without storing
      history. Optional `web_search.count` (0–20) overrides model-selected
      counts; 0 skips the search provider. If a model emits client tool calls
      and searches in the same round, client calls are returned without
      executing that round's searches. On the typed path, completed search
      output markers may accompany replayed assistant text, but raw search
      results are not restored across requests. Native Responses also accepts
      top-level `web_search`, executed by the upstream provider without a Cloud
      search loop. Complete native `web_search_call` replay items, search
      configuration, and `include` are forwarded unchanged; provider-returned
      search citations and sources retain their native shape. `file_search`,
      `code_interpreter`, `computer`, and remote `mcp` are unsupported.
      Image-generation/editing models are rejected on both paths. The separate
      `POST /mcp` endpoint exposes its `web_search` tool independently of
      Responses; use `/v1/images/*` for image generation/editing. Existing
      completed-response gateway attestation is preserved best-effort: when the
      signature write succeeds, `GET /v1/signature/resp_*` retrieves signatures
      over SHA-256 request/response digests, never raw content. Interrupted
      streams create no `resp_*` attestation record or legacy disconnect
      fallback. Responses rejects conversation linkage, response history, and
      file input; the separate Conversation and File endpoints are not part of
      Responses inference.
  - name: Organizations
    description: Organization management
  - name: Organization Members
    description: Organization member and invitation management
  - name: Workspaces
    description: Workspace and API key management
  - name: Users
    description: User profile and token management
  - name: Invitations
    description: Token-based invitation handling
  - name: Usage
    description: Usage tracking and billing information
  - name: Reporting
    description: Read-only customer usage reporting
  - name: Billing
    description: Billing costs endpoint (HuggingFace integration)
  - name: Staking Farm
    description: House of Stake farm credit configuration and synchronization
  - name: Health
    description: Health check endpoints
  - name: Attestation
    description: Attestation and verification endpoints
  - name: Gateway
    description: Model gateway integration endpoints
  - name: Admin
    description: Administrative endpoints (admin access required)
  - name: Services
    description: Public platform services (e.g. web_search pricing)
paths:
  /v1/admin/usage-hourly/recompute:
    post:
      tags:
        - Admin
      summary: Recompute usage_hourly for a UTC window (Admin only)
      description: >-
        **Deprecated:** this endpoint is moving to cloud-platform at the same
        path; the cloud-platform base URL will be announced. Behaviour may
        differ in edge cases.


        Rebuilds `[start, end)` from raw usage one UTC day per transaction, then
        reports raw vs

        aggregate parity for every day the window touches. Parity always covers
        the whole UTC day,

        so a window covering part of a day can report `ok: false` from hours
        outside it; repair

        whole days to clear a day's parity. Use it after a backfill or when the
        nightly parity check

        warns; the scheduler only re-reads the last 3 hours. A window must not
        start past the next

        hour the scheduler has yet to compute (409), or the hours in between
        would never be computed.


        Admin permissions: admin session or read_write token. A read_only token
        receives HTTP 403 with error.type insufficient_permissions before the
        handler runs.
      operationId: recompute_usage_hourly
      requestBody:
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/UsageHourlyRepairRequest'
        required: true
      responses:
        '200':
          description: Window recomputed
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/UsageHourlyRepairResponse'
        '400':
          description: Invalid window
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
        '401':
          description: Unauthorized
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
        '403':
          description: >-
            Missing admin privileges (forbidden) or insufficient token
            permission (insufficient_permissions).
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
        '409':
          description: Window starts past the hours usage_hourly has computed
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
        '500':
          description: Internal server error
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ErrorResponse'
      deprecated: true
      security:
        - session_token: []
        - admin_token: []
components:
  schemas:
    UsageHourlyRepairRequest:
      type: object
      required:
        - start
        - end
      properties:
        end:
          type: string
          format: date-time
          description: Exclusive end; a whole UTC hour, at most 31 days after `start`.
        start:
          type: string
          format: date-time
          description: Inclusive start; a whole UTC hour.
    UsageHourlyRepairResponse:
      type: object
      required:
        - start
        - end
        - rows_written
        - days
      properties:
        days:
          type: array
          items:
            $ref: '#/components/schemas/UsageHourlyDayParity'
        end:
          type: string
          format: date-time
        rows_written:
          type: integer
          format: int64
          minimum: 0
        start:
          type: string
          format: date-time
    ErrorResponse:
      type: object
      required:
        - error
      properties:
        error:
          $ref: '#/components/schemas/ErrorDetail'
    UsageHourlyDayParity:
      type: object
      required:
        - day
        - ok
        - raw_requests
        - aggregate_requests
        - raw_total_cost_nano
        - aggregate_total_cost_nano
      properties:
        aggregate_requests:
          type: integer
          format: int64
        aggregate_total_cost_nano:
          type: integer
          format: int64
        day:
          type: string
          format: date
        ok:
          type: boolean
          description: True when raw and aggregate totals for the whole UTC day match.
        raw_requests:
          type: integer
          format: int64
        raw_total_cost_nano:
          type: integer
          format: int64
    ErrorDetail:
      type: object
      required:
        - message
        - type
      properties:
        code:
          type:
            - string
            - 'null'
        message:
          type: string
        param:
          type:
            - string
            - 'null'
        type:
          type: string
  securitySchemes:
    session_token:
      type: http
      scheme: bearer
      bearerFormat: JWT
      description: >-
        JWT access token for user authentication (Authorization: Bearer
        <jwt_token>). Create via POST /users/me/access_tokens.
    api_key:
      type: http
      scheme: bearer
      bearerFormat: api_key
      description: 'API key for programmatic access (Authorization: Bearer sk-<api_key>)'
    admin_token:
      type: http
      scheme: bearer
      bearerFormat: adm_
      description: >-
        Opaque admin access token. Permission is immutable and applies across
        organizations: read_only allows approved reads; read_write allows admin
        mutations. Both require the creator to remain an admin. Token management
        requires an admin session. Usage bookkeeping and audit logging remain
        enabled.

````

This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.