generated: '2026-08-13'
method: derived
source: >-
  npm @bigfootai/noded-sdk@0.1.2 (dist/index.js, dist/index.d.ts),
  https://www.getnoded.ai/developers, https://www.getnoded.ai/developers/AGENTS.md
note: >-
  Derived from Noded's own published SDK and developer page. Noded publishes no API reference or
  convention guide, so anything not observable in the SDK is recorded as not-documented rather
  than guessed.
api_style: GraphQL (Apollo)
endpoint: https://api.getnoded.ai/api/v1/graph
transport: HTTP POST, application/json
authentication:
  style: bearer
  header: Authorization
  see: authentication/noded-ai-authentication.yml
versioning:
  style: path
  current: v1
  evidence: /api/v1/graph
  policy_published: false
  note: >-
    GraphQL schema evolution (additive fields, deprecated directives) is the usual mechanism for
    an Apollo API; Noded documents no version or deprecation policy. See lifecycle/.
idempotency:
  supported: unknown
  documented: false
  header: null
  note: >-
    No idempotency key, header, or retry-safety guidance appears anywhere in the SDK, the
    developer page, or AGENTS.md. The one mutation the SDK exposes is named upsertOneNote, whose
    upsert semantics imply write convergence on a key, but Noded does not document that as an
    idempotency contract — so no Idempotency pointer is emitted.
pagination:
  style: limit
  params:
    - name: limit
      in: SearchOptions / *Input
      default: null
      note: SDK passes limit through to SearchTagsInput / FindBlocksInput
  cursor: not-documented
  offset: not-documented
  response_fields: not-documented
  note: >-
    The SDK exposes only { limit, select }. No cursor, offset, page token, or total-count field
    is documented, so a consumer cannot page past the first limit deterministically.
field_selection:
  supported: true
  mechanism: GraphQL selection set
  sdk_param: select
  note: >-
    Every helper accepts a `select` string that overrides the SDK's default selection set. This
    is the GraphQL-native equivalent of sparse fieldsets — the provider's guidance is to widen,
    not narrow.
  default_selections:
    tag_fields: default TAG_FIELDS selection compiled into the SDK
    block_fields: default BLOCK_FIELDS selection compiled into the SDK
escape_hatch:
  method: noded.graphql(query, variables)
  note: any Noded GraphQL operation; inputs use the schema's *Input types
metadata:
  supported: unknown
  documented: false
request_id_tracing:
  supported: unknown
  documented: false
  note: no request-id, trace, or correlation header is set or read by the SDK
error_envelope:
  style: graphql
  shape: 'GraphQL response { data, errors[] } — errors[].message joined by the SDK'
  rfc9457: false
  client_error_type: NodedError
  client_error_kinds:
    - auth
    - graphql
    - network
  see: errors/noded-ai-problem-types.yml
rate_limiting:
  documented: false
  response_headers: not-documented
  see: rate-limits/noded-ai-rate-limits.yml
retries:
  documented: false
  sdk_behavior: none — the SDK does not retry; callers wrap in try/catch
permissions:
  model: tenant + signed-in-user permissions enforced server-side
  client_side_enforcement: false
  note: >-
    Provider guidance is explicit — "scope everything to the signed-in user; the API already
    enforces their tenant + permissions" — and never ship a secret to a browser.
cross_links:
  authentication: authentication/noded-ai-authentication.yml
  scopes: scopes/noded-ai-scopes.yml
  errors: errors/noded-ai-problem-types.yml
  lifecycle: lifecycle/noded-ai-lifecycle.yml
  rate_limits: rate-limits/noded-ai-rate-limits.yml
  data_model: data-model/noded-ai-data-model.yml
  graphql: graphql/noded-ai-graphql-operations.yml