> ## Documentation Index > Fetch the complete documentation index at: https://docs.veri.studio/llms.txt > Use this file to discover all available pages before exploring further. # Create a monitor > Scores a sampled share of one deployment's live traffic with one evaluator. A single_turn evaluator scores traced requests as they ship; a thread evaluator scores a thread once it has been idle for `idle_minutes` (again after it resumes and goes idle again). Scores carry `source: monitor` and the monitor's id; a failed evaluation emits the `score.failed` webhook. The deployment must store trace bodies (the judge reads them). ## OpenAPI ````yaml /api-reference/openapi.json post /v1/monitors openapi: 3.1.0 info: title: Veri API description: >- REST API for the Veri RL post-training platform. All requests require a Bearer API key (`vk_` prefix). license: name: '' version: 0.1.0 servers: - url: https://api.veri.studio description: Production security: [] tags: - name: Training jobs description: Create, monitor, and manage training jobs. - name: Datasets description: Upload and connect training datasets. - name: Deployments description: Serve trained models and run inference. - name: Volumes description: Persistent file storage mounted into jobs. - name: Models description: Custom model registry deployments serve from. - name: Regions description: Discover available launch regions. - name: GPU description: Live GPU availability by provider and region. - name: Code artifacts description: Upload custom training script bundles. - name: Billing description: Credit balance and transaction history. - name: API keys description: Create and revoke API keys. - name: Account description: The authenticated caller's identity. - name: Settings description: Account-level integrations (Weights & Biases). - name: Metrics description: Prometheus metrics export for your own observability stack. - name: Evaluators description: 'Evaluators: versioned scoring rules (LLM judge, code, human).' - name: Experiments description: >- Experiments: offline runs of a pinned dataset snapshot through a target, scored by pinned evaluators. - name: Annotation queues description: >- Human review: queue traces, threads and experiment items, reserve one at a time, score them and feed corrections back into datasets. - name: Monitors description: >- Monitors: evaluators scoring a sampled share of a deployment's live traffic. - name: Observability description: Agent conversations, agents and their traffic. - name: Public runs description: Unauthenticated reads of runs their owners published to Explore. paths: /v1/monitors: post: tags: - Monitors summary: Create a monitor description: >- Scores a sampled share of one deployment's live traffic with one evaluator. A single_turn evaluator scores traced requests as they ship; a thread evaluator scores a thread once it has been idle for `idle_minutes` (again after it resumes and goes idle again). Scores carry `source: monitor` and the monitor's id; a failed evaluation emits the `score.failed` webhook. The deployment must store trace bodies (the judge reads them). operationId: create requestBody: content: application/json: schema: $ref: '#/components/schemas/MonitorCreateRequest' required: true responses: '201': description: The monitor content: application/json: schema: $ref: '#/components/schemas/MonitorResponse' '400': description: >- Invalid name, rate, cap, idle minutes or filter; an evaluator a monitor can't run (human, local code, requires a reference); or evals or observability disabled '404': description: >- No such deployment, evaluator or evaluator version in the caller's workspace '409': description: >- trace_bodies_required: the deployment does not store trace bodies; or the evaluator is archived security: - bearerAuth: [] components: schemas: MonitorCreateRequest: type: object required: - name - deployment_id - evaluator_id properties: name: type: string description: 1-64 lowercase letters, digits, '.', '_' or '-'. deployment_id: type: string description: A deployment of the workspace with trace_bodies on. evaluator_id: type: string description: |- Evaluator id (evl_...) or name. llm_judge or cloud code; human and local code evaluators, and evaluators that require a reference answer, can't run on live traffic. evaluator_version: type: - integer - 'null' format: int32 description: Pin a version; null (default) tracks the latest. sampling_rate: type: number format: double description: 0..1 (default 1). filter: type: object description: '{agent?} (default {}).' idle_minutes: type: integer format: int32 description: 1..1440 (default 15). daily_cap: type: integer format: int32 description: 1..100000 (default 1000). enabled: type: boolean review_below_confidence: type: - number - 'null' format: double description: >- Confidence routing (set both or neither): scores whose confidence is below this (0..1) send their target to `review_queue_id`, an annotation queue (id or name) reviewing traces (single_turn monitors) or threads (thread monitors). 400 while Jev is off. review_queue_id: type: - string - 'null' MonitorResponse: type: object description: One monitor. required: - object - id - name - deployment_id - evaluator_id - evaluator_name - scope - sampling_rate - filter - idle_minutes - daily_cap - enabled - stats - created_at - updated_at properties: object: type: string description: Always "monitor". id: type: string description: '`mon_...`' name: type: string deployment_id: type: string description: The deployment whose traffic is scored. evaluator_id: type: string evaluator_name: type: string evaluator_version: type: - integer - 'null' format: int32 description: The pinned version; null tracks the latest. scope: type: string description: |- single_turn (scores traced requests) | thread (scores idle threads), from the version the monitor runs. sampling_rate: type: number format: double description: Share of the traffic scored, 0..1. filter: type: object description: '{agent?}: only requests or threads sent with this X-Veri-Agent.' idle_minutes: type: integer format: int32 description: >- Thread monitors: minutes without a request before a thread is scored. daily_cap: type: integer format: int32 description: Most targets queued per UTC day. enabled: type: boolean review_below_confidence: type: - number - 'null' format: double description: |- Confidence routing: an ok score whose confidence is below this (0..1) adds its target to `review_queue_id`; null = off. review_queue_id: type: - string - 'null' description: The annotation queue low-confidence targets go to. stats: $ref: '#/components/schemas/MonitorStatsResponse' created_at: type: string format: date-time updated_at: type: string format: date-time MonitorStatsResponse: type: object required: - scored_today - flagged_today - scored_7d - flagged_7d properties: scored_today: type: integer format: int64 description: ok scores written today (UTC). flagged_today: type: integer format: int64 description: Of those, the ones with passed = false. scored_7d: type: integer format: int64 flagged_7d: type: integer format: int64 last_scored_at: type: - string - 'null' format: date-time human_agreement: oneOf: - type: 'null' - $ref: '#/components/schemas/HumanAgreement' description: Null until a human label shares a target with this monitor. HumanAgreement: type: object description: |- Agreement between a monitor's verdicts and human labels of the same evaluator on the same targets. required: - 'n' - rate properties: 'n': type: integer format: int64 description: Targets with both a monitor verdict and a human verdict. rate: type: number format: double description: Share of them where the verdicts agree, 0..1. securitySchemes: bearerAuth: type: http scheme: bearer bearerFormat: API key description: API key with the `vk_` prefix. Create one from the dashboard. ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.