generated: '2026-10-09'
method: derived
generator: derive-data-model.py
source:
- openapi/tale-dev-openapi.yml
notation: has_one = property $ref; has_many = array of $ref; belongs_to = <entity>_id field
summary:
  entities: 68
  entities_in_spec: 68
  relationships: 34
entities:
- name: Actor
  fields: 2
  description: The verified member a machine caller acts FOR when it relays a person’s own gesture — named by e-mail,
    never by an id the caller made up. The door resolves the address against the organization’s membe
- name: AnthropicError
  fields: 3
  description: A refusal on the Anthropic-compatible model endpoint, in Anthropic’s own error shape — the REST door’s
    refusals included. `error.code` is this API’s stable refusal code, an addition to Anthropic’s sha
- name: ApprovalPart
  fields: 4
- name: AttachmentPart
  fields: 7
- name: Automation
  fields: 10
- name: AutomationSummary
  fields: 8
- name: AutomationVersion
  fields: 7
- name: BrowserSession
  fields: 8
- name: BrowserSessionImport
  fields: 7
- name: BulkCreateResult
  fields: 4
  description: 'A body the schema refuses fails the whole call with 400; a row the directory refuses (a duplicate
    email or external id) fails that row alone and is listed under `errors` with its code. `created` here '
- name: ChatModel
  fields: 10
- name: Contact
  fields: 15
- name: ContactInput
  fields: 10
  description: Unknown keys are refused (`INVALID_BODY`). Strings are trimmed, and a blank optional field — or one
    sent as `null` — reads as left out (a CSV-shaped row or a JSON export imports cleanly); the same bod
- name: ContactPatch
  fields: 11
  description: 'Every field optional. `null` clears any optional field, and a blank string reads as `null`; `metadata`
    merges per RFC 7396 while `address` is replaced whole. A patch may not clear the last of `name`, '
- name: ConversationDelivery
  fields: 12
  description: One native Inbox reply in a source’s delivery queue, as the listing shows it — never the claim token
    or the body
- name: ConversationMirror
  fields: 10
- name: Document
  fields: 16
- name: DocumentInput
  fields: 10
- name: DocumentPatch
  fields: 10
  description: Every field optional; null clears a nullable field. Applies only to Knowledge Hub documents. A patch
    that changes nothing writes nothing and leaves `updatedAt` alone.
- name: Error
  fields: 4
- name: ExternalStatusDecision
  fields: 2
- name: ExternalStatusWorkflow
  fields: 1
- name: KnowledgeEntry
  fields: 10
- name: KnowledgeEntryCreated
  fields: 2
- name: KnowledgeHit
  fields: 13
- name: KnowledgeSearchResult
  fields: 2
- name: LegacyRunQuarantine
  fields: 5
  description: A legacy execution whose external effects or termination cannot be proven. Present on held runs.
    A recorded stop request does not resolve the hold or establish that work stopped.
- name: LegacyRunStopRequest
  fields: 4
- name: Me
  fields: 5
- name: Message
  fields: 13
- name: OpenAiError
  fields: 1
  description: A refusal on the OpenAI-compatible model endpoints, in OpenAI’s own error shape — the REST door’s
    refusals (a missing key, the organization header, the request budget) included. `code` is this API’s s
- name: PendingAsk
  fields: 8
  description: 'The question a run is parked on (`waitingFor: "ask"`): what the agent asked, on which node, and
    until when an answer still resumes the run. Answer at `POST …/runs/{runId}/asks/{askId}`.'
- name: Product
  fields: 16
- name: ProductInput
  fields: 11
  description: '`name` is required on create; every field is optional on update. Strings are trimmed, and a blank
    optional field — or one sent as `null` — reads as left out. Unknown keys are refused (`INVALID_BODY`).'
- name: ProductPatch
  fields: 12
  description: Every field optional. `null` clears any optional field, and a blank string reads as `null` (a blank
    `name` is refused); `metadata` merges per RFC 7396. Send `expectedUpdatedAt` from the last product r
- name: Project
  fields: 9
- name: ProjectAgent
  fields: 17
- name: ProjectAgentInput
  fields: 9
  description: Full project-agent configuration. name, harness, model, skills and connectors are required. Harness
    must support project-agent execution — one the managed lane runs unattended on the platform’s own cr
- name: ProjectAgentUpdate
  fields: 10
  description: The full configuration a `PUT` replaces — `ProjectAgentInput` plus the optional `expectedUpdatedAt`
    precondition.
- name: ProjectFile
  fields: 7
- name: ProjectFolder
  fields: 3
- name: ProjectFolderResult
  fields: 2
- name: ProjectUploadHandoff
  fields: 6
- name: QuestionSet
  fields: 2
  description: 'A structured question set an agent may ask instead of a plain sentence: up to four questions, each
    a choice among two to four labelled options; the person may also answer a question in their own words'
- name: ReasoningPart
  fields: 2
  description: The model’s reasoning ahead of its reply — display-only, never replayed to the model; it may quote
    the assistant’s own instructions verbatim, so it is not safe to render as the answer.
- name: RetryIndexingResult
  fields: 2
  description: What a retry-indexing request did — the one answer the Hub document door and the project-file door
    share.
- name: Run
  fields: 25
  description: 'A run in full — the single read: identity, status, input, output, trace, effects, and per-node checkpoints.
    Listings answer `RunSummary` rows.'
- name: RunProjection
  fields: 25
  description: 'A run read with `?fields=`: exactly the keys of `Run` the query named, no others — none is required,
    since the caller chose them.'
- name: RunSummary
  fields: 23
  description: One run as a listing reports it — identity, scope, status and timing. The input, output, trace, effects
    and checkpoints of the full `Run` are inlined only when `include` names them; the single read an
- name: Skill
  fields: 17
  description: 'The skill bundle: its summary fields plus the markdown body and the bundle’s file list'
- name: SkillSummary
  fields: 15
- name: Task
  fields: 17
- name: TaskReview
  fields: 11
  description: A task’s ordinary pending review, owned by the captured person or agent. Native agent verdicts use
    task_review; the public review POST relays a person’s decision.
- name: TaskStatusSnapshot
  fields: 7
- name: TaskUpsertResult
  fields: 3
- name: Team
  fields: 3
  description: 'One team of the organization: the `id` an audience names, the `name` the app shows for it, and whether
    the key holder is a member.'
- name: TextPart
  fields: 2
- name: Thread
  fields: 11
- name: ToolCallPart
  fields: 4
- name: ToolResultPart
  fields: 5
- name: Trigger
  fields: 16
  description: 'The binding, and its health: `lastFiredAt` and `lastRunId` name the last run it started, `lastSkippedAt`
    and `lastSkipReason` the last time it came due and started nothing. A binding is alive when `la'
- name: Website
  fields: 15
- name: WebsiteInput
  fields: 5
- name: WebsitePage
  fields: 13
- name: WebsitePageList
  fields: 6
- name: WebsitePatch
  fields: 4
  description: 'The domain is immutable after create (the crawl registration is keyed by it): a body carrying a
    `domain` other than the stored one is refused with 400 `WEBSITE_DOMAIN_IMMUTABLE` — delete the website a'
- name: WebsiteSearchHit
  fields: 5
- name: WebsiteSearchResults
  fields: 2
relationships:
- from: Document
  to: Team
  type: belongs_to
  via: teamIds
- from: Document
  to: Team
  type: belongs_to
  via: teamId
- from: DocumentInput
  to: Team
  type: belongs_to
  via: teamIds
- from: DocumentInput
  to: Team
  type: belongs_to
  via: teamId
- from: DocumentPatch
  to: Team
  type: belongs_to
  via: teamIds
- from: DocumentPatch
  to: Team
  type: belongs_to
  via: teamId
- from: WebsitePageList
  to: WebsitePage
  type: has_many
  via: pages
- from: WebsiteSearchResults
  to: WebsiteSearchHit
  type: has_many
  via: results
- from: ConversationMirror
  to: Contact
  type: belongs_to
  via: contactId
- from: ConversationDelivery
  to: Message
  type: belongs_to
  via: messageId
- from: Project
  to: Team
  type: belongs_to
  via: teamIds
- from: ProjectFolderResult
  to: ProjectFolder
  type: has_one
  via: folder
- from: Task
  to: Project
  type: belongs_to
  via: projectId
- from: PendingAsk
  to: Run
  type: belongs_to
  via: runId
- from: PendingAsk
  to: QuestionSet
  type: has_one
  via: questions
- from: PendingAsk
  to: Task
  type: belongs_to
  via: taskId
- from: TaskStatusSnapshot
  to: ExternalStatusWorkflow
  type: has_one
  via: workflow
- from: TaskReview
  to: Task
  type: belongs_to
  via: taskId
- from: TaskReview
  to: Run
  type: belongs_to
  via: runId
- from: TaskUpsertResult
  to: Run
  type: belongs_to
  via: runId
- from: AutomationSummary
  to: Project
  type: belongs_to
  via: projectIds
- from: Automation
  to: Project
  type: belongs_to
  via: projectIds
- from: RunSummary
  to: Run
  type: belongs_to
  via: runId
- from: RunSummary
  to: Project
  type: belongs_to
  via: projectId
- from: RunSummary
  to: LegacyRunQuarantine
  type: has_one
  via: legacyQuarantine
- from: Run
  to: Project
  type: belongs_to
  via: projectId
- from: Run
  to: LegacyRunQuarantine
  type: has_one
  via: legacyQuarantine
- from: RunProjection
  to: Project
  type: belongs_to
  via: projectId
- from: RunProjection
  to: LegacyRunQuarantine
  type: has_one
  via: legacyQuarantine
- from: KnowledgeSearchResult
  to: KnowledgeHit
  type: has_many
  via: hits
- from: Thread
  to: Project
  type: belongs_to
  via: projectId
- from: ProjectAgent
  to: Project
  type: belongs_to
  via: projectId
- from: KnowledgeEntry
  to: Document
  type: belongs_to
  via: documentId
- from: KnowledgeEntryCreated
  to: Document
  type: belongs_to
  via: documentId