vocabulary:
  name: Evidently AI API Vocabulary
  description: >
    Domain terms and concepts used across the Evidently AI platform for ML/LLM
    observability, evaluation, and monitoring.
  version: '1.0.0'
  url: https://www.evidentlyai.com/

terms:

  - term: Snapshot
    definition: >
      A serialised Evidently report or test suite result captured at a point in time.
      Snapshots are the core unit of storage in the Evidently platform; each contains
      computed metrics, widget data, and optional metadata tags.
    type: resource
    relatedTerms:
      - Report
      - TestSuite

  - term: Report
    definition: >
      An Evidently output object that computes and visualises a set of metrics
      over a reference and current dataset. Stored as a snapshot on the platform.
    type: resource
    relatedTerms:
      - Snapshot
      - Metric

  - term: TestSuite
    definition: >
      An Evidently output object that runs a set of conditional tests over data
      and returns pass/fail results. Stored as a snapshot on the platform.
    type: resource
    relatedTerms:
      - Snapshot
      - Test

  - term: Project
    definition: >
      The top-level organisational container in the Evidently platform. A project
      groups related snapshots, dashboards, datasets, and artifacts belonging to
      a single ML or LLM system under observation.
    type: resource

  - term: Dashboard
    definition: >
      A configurable monitoring view in the Evidently platform that displays
      time-series panel charts derived from snapshot metrics. Each project has
      one dashboard.
    type: resource
    relatedTerms:
      - Project
      - Panel

  - term: Metric
    definition: >
      A computed statistical measure calculated over a dataset within a Report.
      Evidently ships over 100 built-in metrics covering data drift, data quality,
      model performance, and LLM output quality.
    type: concept
    examples:
      - DatasetDriftMetric
      - ColumnDriftMetric
      - DatasetMissingValuesMetric
      - TextEvals

  - term: DataDrift
    definition: >
      A class of Evidently metrics that detect statistical distribution shifts
      between a reference dataset and a current dataset. Supports tabular,
      text, and embedding column types.
    type: concept
    relatedTerms:
      - Metric

  - term: DataQuality
    definition: >
      A class of Evidently metrics that measure data completeness, correctness,
      and consistency — including missing values, duplicates, and column statistics.
    type: concept
    relatedTerms:
      - Metric

  - term: LLMEvaluation
    definition: >
      Automated scoring of LLM-generated text outputs using Evidently's TextEvals
      metric suite, which includes built-in judges for correctness, relevance,
      toxicity, and custom LLM-judge prompts.
    type: concept
    relatedTerms:
      - Metric
      - Trace

  - term: Trace
    definition: >
      A recorded interaction with an LLM application (prompt + response pairs)
      collected via the Tracely SDK or OpenTelemetry-compatible instrumentation.
      Traces are uploaded to the platform for evaluation and monitoring.
    type: resource
    relatedTerms:
      - LLMEvaluation

  - term: Dataset
    definition: >
      A tabular data file (CSV/Parquet) uploaded to the Evidently platform and
      associated with a project. Datasets can be linked to snapshots and used
      for reproducible evaluations.
    type: resource

  - term: Artifact
    definition: >
      A versioned model asset stored in the Evidently platform — for example,
      a trained model, a prompt template, or an evaluation configuration — tracked
      across multiple versions.
    type: resource

  - term: ArtifactVersion
    definition: >
      An immutable, numbered revision of an Evidently Artifact. Each version
      stores the artifact payload and creation metadata.
    type: resource
    relatedTerms:
      - Artifact

  - term: BearerToken
    definition: >
      The authentication credential required to call the Evidently Platform REST API.
      Obtained from the Evidently platform settings page and passed as the
      Authorization: Bearer <token> header or via the EVIDENTLY_API_KEY env var.
    type: securityScheme

  - term: ColumnMapping
    definition: >
      An Evidently configuration object that specifies the roles of columns in a
      dataset — target, prediction, features, text columns, embeddings, etc.
      Required for task-specific metrics.
    type: configuration

  - term: Preset
    definition: >
      A pre-packaged bundle of related Evidently metrics or tests covering a
      common monitoring scenario (e.g. DataDriftPreset, DataQualityPreset,
      TextOverviewPreset). Presets simplify report configuration.
    type: configuration

  - term: Panel
    definition: >
      A single visualisation widget on an Evidently Dashboard that plots a
      metric time-series. Panels are configured with a metric type, optional
      label filters, and display settings.
    type: resource
    relatedTerms:
      - Dashboard

  - term: EVIDENTLY_API_KEY
    definition: >
      The environment variable used to supply the API authentication token to the
      Evidently Python SDK (CloudWorkspace) and CI integrations.
    type: configuration

  - term: CloudWorkspace
    definition: >
      The Evidently Python SDK client class that connects to the hosted Evidently
      platform (app.evidently.cloud) or a self-hosted instance using an API key.
    type: sdkConcept

  - term: Litestar
    definition: >
      The Python ASGI web framework used to build the Evidently Platform REST API.
      Routes are mounted under /api and /api/v2.
    type: implementation