generated: '2026-09-12'
method: derived
source: discovery/google-dialogflow-v2.json, discovery/google-dialogflow-v3.json,
  openapi/google-dialogflow-es-v2-openapi.yml, openapi/google-dialogflow-cx-v3-openapi.yml
provider: Google Dialogflow
providerId: google-dialogflow
description: >-
  Entity-relationship graph for the two Dialogflow editions, derived from the resource hierarchy and
  $ref graph in Google's Discovery Documents. Dialogflow encodes its relationships in the resource
  NAME rather than in foreign-key fields — projects/{p}/locations/{l}/agents/{a}/flows/{f}/pages/{g}
  is the containment graph, written out in the URL. So the parent/child edges below are read from
  the path hierarchy, and the cross-cutting references from schema fields that hold another
  resource's name.
identifier_convention:
  style: hierarchical resource name
  format: 'projects/{project}[/locations/{location}]/{collection}/{id}[/{collection}/{id}...]'
  note: >-
    The id segment is a server-assigned opaque string, not a prefixed key. There is no Stripe-style
    `cus_`/`ch_` prefix family, so an id on its own does not tell you what it is — only the full
    name does. Contract patterns are declared per operation, e.g.
    ^projects/[^/]+/agent/intents/[^/]+$.
schema_counts:
  es_v2:
    total_schemas: 692
    core_entities: 24
  cx_v3:
    total_schemas: 667
    core_entities: 26
editions:
- name: Dialogflow ES
  api: v2
  root: projects/{project}
  entities:
  - name: Agent
    schema: GoogleCloudDialogflowV2Agent
    note: Singleton. One agent per region per project — projects/{project}/agent, not a collection.
    relationships:
    - has_many: Intent
      via: projects/{project}/agent/intents
    - has_many: EntityType
      via: projects/{project}/agent/entityTypes
    - has_many: Environment
      via: projects/{project}/agent/environments
    - has_many: Version
      via: projects/{project}/agent/versions
    - has_one: Fulfillment
      via: projects/{project}/agent/fulfillment
  - name: Intent
    schema: GoogleCloudDialogflowV2Intent
    relationships:
    - has_many: TrainingPhrase
      via: trainingPhrases[] (embedded, max 2000 per language)
    - has_many: Parameter
      via: parameters[] (embedded, max 20)
    - belongs_to: Intent
      via: parentFollowupIntentName
    - has_many: Intent
      via: followupIntentInfo[]
    - references: Context
      via: inputContextNames[] (max 5), outputContexts[] (max 30)
  - name: EntityType
    schema: GoogleCloudDialogflowV2EntityType
    relationships:
    - has_many: Entity
      via: entities[] (embedded, max 30000)
  - name: Context
    schema: GoogleCloudDialogflowV2Context
    relationships:
    - belongs_to: Session
      via: projects/{project}/agent/sessions/{session}/contexts/{context}
  - name: SessionEntityType
    schema: GoogleCloudDialogflowV2SessionEntityType
    relationships:
    - belongs_to: Session
    - references: EntityType
      via: name suffix matches the developer entity type being overridden
  - name: Environment
    schema: GoogleCloudDialogflowV2Environment
    relationships:
    - references: Version
      via: agentVersion
    - has_many: EnvironmentHistory
      via: projects/{project}/agent/environments/{env}/history
  - name: Version
    schema: GoogleCloudDialogflowV2Version
  - name: Fulfillment
    schema: GoogleCloudDialogflowV2Fulfillment
    note: Singleton per agent. Holds the webhook configuration.
  - name: KnowledgeBase
    schema: GoogleCloudDialogflowV2KnowledgeBase
    relationships:
    - has_many: Document
      via: projects/{project}/knowledgeBases/{kb}/documents
  - name: Document
    schema: GoogleCloudDialogflowV2Document
  - name: ConversationProfile
    schema: GoogleCloudDialogflowV2ConversationProfile
    note: The Agent Assist configuration object — binds suggestion features, models and the Pub/Sub
      notification topic.
    relationships:
    - references: ConversationModel
    - references: KnowledgeBase
  - name: Conversation
    schema: GoogleCloudDialogflowV2Conversation
    relationships:
    - belongs_to: ConversationProfile
      via: conversationProfile
    - has_many: Participant
      via: projects/{project}/conversations/{conv}/participants
    - has_many: Message
      via: projects/{project}/conversations/{conv}/messages
  - name: Participant
    schema: GoogleCloudDialogflowV2Participant
  - name: AnswerRecord
    schema: GoogleCloudDialogflowV2AnswerRecord
    note: The feedback join between a suggestion and what the human agent did with it.
  - name: ConversationDataset
    schema: GoogleCloudDialogflowV2ConversationDataset
  - name: ConversationModel
    schema: GoogleCloudDialogflowV2ConversationModel
    relationships:
    - belongs_to: ConversationDataset
      via: datasets[]
    - has_many: ConversationModelEvaluation
  - name: Generator
    schema: GoogleCloudDialogflowV2Generator
    relationships:
    - has_many: GeneratorEvaluation
  - name: Tool
    schema: GoogleCloudDialogflowV2Tool
  - name: SipTrunk
    schema: GoogleCloudDialogflowV2SipTrunk
  - name: EncryptionSpec
    schema: GoogleCloudDialogflowV2EncryptionSpec
    note: Singleton per location; holds the CMEK key name.
- name: Dialogflow CX (Conversational Agents)
  api: v3
  root: projects/{project}/locations/{location}
  entities:
  - name: Agent
    schema: GoogleCloudDialogflowCxV3Agent
    note: A collection here, unlike ES — up to 1000 agents per project.
    relationships:
    - has_many: Flow
      via: agents/{agent}/flows (max 50)
    - has_many: Intent
      via: agents/{agent}/intents (max 10000)
    - has_many: EntityType
      via: agents/{agent}/entityTypes (max 250)
    - has_many: Webhook
      via: agents/{agent}/webhooks (max 100)
    - has_many: Tool
      via: agents/{agent}/tools (max 100)
    - has_many: Playbook
      via: agents/{agent}/playbooks (max 50)
    - has_many: Environment
      via: agents/{agent}/environments (max 20)
    - has_many: TestCase
      via: agents/{agent}/testCases (max 2000)
    - has_many: Changelog
      via: agents/{agent}/changelogs (max 20000)
    - has_one: GenerativeSettings
      via: agents/{agent}/generativeSettings
    - references: SecuritySettings
      via: securitySettings
    - references: Flow
      via: startFlow
  - name: Flow
    schema: GoogleCloudDialogflowCxV3Flow
    relationships:
    - has_many: Page
      via: flows/{flow}/pages (max 250)
    - has_many: TransitionRouteGroup
      via: flows/{flow}/transitionRouteGroups (max 100)
    - has_many: Version
      via: flows/{flow}/versions (max 20)
    - references: Intent
      via: transitionRoutes[].intent
  - name: Page
    schema: GoogleCloudDialogflowCxV3Page
    relationships:
    - references: Page
      via: transitionRoutes[].targetPage
    - references: Flow
      via: transitionRoutes[].targetFlow
    - references: Webhook
      via: entryFulfillment.webhook, form.parameters[].fillBehavior
    - references: TransitionRouteGroup
      via: transitionRouteGroups[]
  - name: Intent
    schema: GoogleCloudDialogflowCxV3Intent
    relationships:
    - has_many: TrainingPhrase
      via: trainingPhrases[] (embedded, max 2000 per language)
    - references: EntityType
      via: parameters[].entityType
  - name: EntityType
    schema: GoogleCloudDialogflowCxV3EntityType
  - name: TransitionRouteGroup
    schema: GoogleCloudDialogflowCxV3TransitionRouteGroup
  - name: Webhook
    schema: GoogleCloudDialogflowCxV3Webhook
    note: See asyncapi/google-dialogflow-webhooks.yml — the event surface is a managed resource.
  - name: Tool
    schema: GoogleCloudDialogflowCxV3Tool
    relationships:
    - has_many: ToolVersion
      via: tools/{tool}/versions
  - name: Playbook
    schema: GoogleCloudDialogflowCxV3Playbook
    relationships:
    - has_many: Example
      via: playbooks/{playbook}/examples
    - has_many: PlaybookVersion
      via: playbooks/{playbook}/versions (max 100)
    - references: Tool
      via: referencedTools[]
    - references: Playbook
      via: referencedPlaybooks[]
    - references: Flow
      via: referencedFlows[]
  - name: Example
    schema: GoogleCloudDialogflowCxV3Example
    note: A worked conversation used to steer a playbook; the generative counterpart of a training
      phrase.
  - name: Generator
    schema: GoogleCloudDialogflowCxV3Generator
  - name: Environment
    schema: GoogleCloudDialogflowCxV3Environment
    relationships:
    - references: Version
      via: versionConfigs[].version
    - has_many: Deployment
      via: environments/{env}/deployments
    - has_many: ContinuousTestResult
      via: environments/{env}/continuousTestResults
    - has_many: Experiment
      via: environments/{env}/experiments
  - name: Version
    schema: GoogleCloudDialogflowCxV3Version
    note: Scoped to a flow, not to the agent — flows/{flow}/versions.
  - name: Deployment
    schema: GoogleCloudDialogflowCxV3Deployment
  - name: Experiment
    schema: GoogleCloudDialogflowCxV3Experiment
  - name: TestCase
    schema: GoogleCloudDialogflowCxV3TestCase
    relationships:
    - has_many: TestCaseResult
      via: testCases/{tc}/results
  - name: SessionEntityType
    schema: GoogleCloudDialogflowCxV3SessionEntityType
  - name: SecuritySettings
    schema: GoogleCloudDialogflowCxV3SecuritySettings
    note: Location-scoped, not agent-scoped — retention, redaction and inspect templates live here.
  - name: Changelog
    schema: GoogleCloudDialogflowCxV3Changelog
    note: Read-only audit record of design-time edits.
  - name: AnswerFeedback
    schema: GoogleCloudDialogflowCxV3AnswerFeedback
cross_edition:
  note: >-
    ES and CX share a host and an auth model but not a data model. There is no foreign key between
    them. Google documents a one-way migration path (ES to CX) rather than interoperation.
runtime_vs_design_time:
  note: >-
    The graph above is almost entirely DESIGN-TIME. The runtime surface is thin by comparison —
    sessions:detectIntent, sessions:streamingDetectIntent, sessions:matchIntent,
    sessions:fulfillIntent and the session entity types — and a Session is not a retrievable
    resource: it has no get operation and its data is aged out 30 minutes after the last request.
    This split matters for quota too: design-time reads and writes are quota'd separately and far
    more tightly (60/300 rpm) than runtime turns (600/1200 rpm).
maintainers:
- FN: Kin Lane
  email: kin@apievangelist.com