> ## Documentation Index > Fetch the complete documentation index at: https://docs.primeintellect.ai/llms.txt > Use this file to discover all available pages before exploring further. # Get Factory Utilization > The team's Model Factory utilization curve: Slurm node allocated/idle counts per bucket, derived from the slurmdbd accounting record with the live jobs read as the truth for the current bucket. Per-source freshness travels in `sources`; gaps are gaps, never zero-filled. ## OpenAPI ````yaml https://api.primeintellect.ai/openapi.json get /api/v1/factory/utilization openapi: 3.1.0 info: title: PI API version: 0.1.0 servers: - url: https://api.primeintellect.ai security: [] paths: /api/v1/factory/utilization: get: tags: - factory summary: Get Factory Utilization description: |- The team's Model Factory utilization curve: Slurm node allocated/idle counts per bucket, derived from the slurmdbd accounting record with the live jobs read as the truth for the current bucket. Per-source freshness travels in `sources`; gaps are gaps, never zero-filled. operationId: get_factory_utilization_api_v1_factory_utilization_get parameters: - name: team_id in: query required: true schema: type: string description: Team whose Model Factory fleet to read title: Team Id description: Team whose Model Factory fleet to read - name: window in: query required: false schema: enum: - 1h - 24h - 7d - 30d - 90d type: string description: >- Curve window; bucket sizes scale with the window (1h->1min, 24h->5min, 7d->hourly, 30d->3h, 90d->daily) default: 24h title: Window description: >- Curve window; bucket sizes scale with the window (1h->1min, 24h->5min, 7d->hourly, 30d->3h, 90d->daily) responses: '200': description: Successful Response content: application/json: schema: $ref: '#/components/schemas/FactoryUtilizationResponse' '401': description: Authorization failed '422': description: Invalid request data content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' security: - HTTPBearer: [] components: schemas: FactoryUtilizationResponse: properties: schema_version: type: integer enum: - 1 const: 1 title: Schema Version default: 1 as_of: type: string format: date-time title: As Of window: type: string enum: - 1h - 24h - 7d - 30d - 90d title: Window bucket_seconds: type: integer title: Bucket Seconds clusters: items: $ref: '#/components/schemas/FactoryClusterUtilization' type: array title: Clusters additionalProperties: false type: object required: - as_of - window - bucket_seconds - clusters title: FactoryUtilizationResponse description: |- Envelope for the cluster utilization curve (public + internal routes). ErrorResponse: properties: errors: items: $ref: '#/components/schemas/ErrorDetail' type: array title: Errors type: object required: - errors title: ErrorResponse FactoryClusterUtilization: properties: display_name: type: string title: Display Name status: type: string enum: - online - offline title: Status total_nodes: anyOf: - type: integer - type: 'null' title: Total Nodes derived_from: type: string enum: - accounting const: accounting title: Derived From default: accounting series: items: $ref: '#/components/schemas/FactoryNodeUtilizationPoint' type: array title: Series default: [] source: anyOf: - $ref: '#/components/schemas/FactorySourceStatus' - type: 'null' additionalProperties: false type: object required: - display_name - status title: FactoryClusterUtilization description: |- One cluster's node-allocation curve (the "Cluster utilization"). DERIVATION (honesty contract): the series is derived from the slurmdbd accounting job records — a node counts as allocated for the whole bucket in which a job's [start, end) interval overlaps it. Approximations, documented here and rendered as such: - a node held by ANY job counts as allocated (Slurm's own allocated/mixed distinction is not recoverable from this source); - the denominator (`total_nodes`) is the deployment's CURRENT claimed-node count from the capacity snapshot — node-count history is not persisted, so `idle` assumes the fleet was its current size across the window; - `draining` (and drain/down nodes inside `idle`) are not derivable from accounting; draining stays null and idle may include nodes that were down or draining; - a failed accounting read surfaces through `source` (ok/stale/ error) with the partial curve still served, exactly like the workloads feed's degraded-source rule — never a silent gap. ErrorDetail: properties: param: type: string title: Param details: type: string title: Details type: object required: - param - details title: ErrorDetail FactoryNodeUtilizationPoint: properties: timestamp: type: string format: date-time title: Timestamp allocated: anyOf: - type: integer - type: 'null' title: Allocated idle: anyOf: - type: integer - type: 'null' title: Idle draining: anyOf: - type: integer - type: 'null' title: Draining additionalProperties: false type: object required: - timestamp title: FactoryNodeUtilizationPoint description: |- One bucket of the node-allocation curve (bucket-start timestamp). `allocated` = distinct nodes held by any job during the bucket (union across overlapping jobs; None = unreadable node evidence for the bucket — never zero-filled). `idle` = the cluster's current claimed node count minus allocated; None when the denominator is unknown or allocated is null (never zero-filled, never negative). `draining` is always None in v1: node states over time need a sampler the platform does not have — the field exists so the contract never lies about what is known. FactorySourceStatus: properties: kind: type: string enum: - capacity - training - inference - slurm title: Kind status: type: string enum: - ok - stale - error title: Status observed_at: anyOf: - type: string format: date-time - type: 'null' title: Observed At additionalProperties: false type: object required: - kind - status title: FactorySourceStatus description: |- Freshness/coverage entry for one evidence source. EXACTLY these three keys. No source id, adapter detail, endpoint or raw error may ever be added here. securitySchemes: HTTPBearer: type: http scheme: bearer ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.