generated: '2026-07-19'
method: searched
source: >-
  https://docs.langdock.com/en/developer/overview/api-introduction plus derivation
  from openapi/langdock-openapi-original.yml — the cross-cutting request/response
  conventions that apply across Langdock endpoints rather than to any single operation.
description: >-
  How the Langdock API behaves across operations: authentication style, the
  provider-compatibility routing model, regional endpoints, pagination, versioning,
  error envelopes, and rate-limit signaling. Langdock's defining convention is
  provider-shape compatibility — its completion and embedding endpoints deliberately
  mirror the OpenAI, Anthropic, Google Vertex AI, and Mistral wire formats so that
  existing vendor SDKs work unchanged against a Langdock base URL.
base_url: https://api.langdock.com
dedicated_deployment_base_url: https://<your-domain>/api/public
api_style: REST over HTTPS, JSON request and response bodies, multipart for file upload
authentication:
  scheme: HTTP Bearer — workspace API key
  header: 'Authorization: Bearer YOUR_API_KEY'
  key_prefix: sk-ld-
  browser_calls: blocked — backend access only
  detail: authentication/langdock-authentication.yml
  docs: https://docs.langdock.com/en/developer/overview/api-introduction
provider_compatibility:
  description: >-
    Model endpoints are namespaced by upstream provider and mirror that provider's
    native request/response shape, so the official vendor SDK can be pointed at
    Langdock by changing only the base URL and API key.
  routes:
  - provider: openai
    path: /openai/{region}/v1/chat/completions, /openai/{region}/v1/embeddings
    compatible_with: OpenAI SDKs
  - provider: anthropic
    path: /anthropic/{region}/v1/messages
    compatible_with: Anthropic SDKs
  - provider: google
    path: /google/{region}/v1beta/models/{model}:generateContent
    compatible_with: Official Vertex AI SDKs (Python / Node)
  - provider: mistral
    path: /mistral/{region}/v1/fim/completions
    compatible_with: Mistral Codestral FIM
  - provider: langdock
    path: /agent/v1/chat/completions
    compatible_with: Vercel AI SDK format
regions:
  supported: true
  mechanism: '{region} path segment on every model endpoint'
  values:
  - eu
  - us
  notes: >-
    /mistral/{region}/v1/fim/completions declares eu only; the OpenAI, Anthropic, and
    Google routes declare both eu and us. Derived from the region path-parameter enums
    in the OpenAPI.
idempotency:
  supported: false
  detail: >-
    Langdock documents no idempotency contract. There is no Idempotency-Key header or
    parameter anywhere in the OpenAPI, and the term does not appear in the published
    documentation index. No Idempotency pointer is emitted for this provider.
pagination:
  style: cursor
  applies_to:
  - GET /skills/v1
  - GET /audit-logs/{workspace_id}
  request_params:
    cursor: opaque cursor marking the start of the next page
    limit: page size
  scope_note: >-
    Pagination is only declared on the two list endpoints above. The knowledge-folder
    list endpoint (GET /knowledge/{folderId}/list) declares no pagination parameters.
  source: openapi/langdock-openapi-original.yml
filtering:
  description: Query-parameter filtering is declared on the paginated list endpoints.
  params:
    skills:
    - query
    - slug
    audit_logs:
    - from
    - to
    - entity_type
    - actor_id
field_expansion:
  supported: false
metadata:
  supported: false
request_tracing:
  documented: false
  detail: No request-id or correlation header is declared in the OpenAPI or documented.
versioning:
  scheme: uri-path, per product family
  detail: >-
    Version is carried in the path segment of first-party endpoints (/agent/v1,
    /assistant/v1, /skills/v1, /attachment/v1, /user-management/v1). Provider-compatible
    routes carry the upstream provider's own version segment (/v1, /v1beta). The
    OpenAPI document itself declares info.version 3.0.0.
  current:
    agents: v1
    skills: v1
    attachments: v1
    user_management: v1
  legacy: /assistant/v1 is deprecated in favour of /agent/v1 — see lifecycle/langdock-lifecycle.yml
error_envelope:
  format: json
  problem_json: false
  detail: >-
    Errors are provider-shaped JSON, not RFC 9457 application/problem+json. The
    OpenAPI defines a family of named error schemas — ErrorResponse, APIError,
    AuthenticationError, InvalidRequestError, NotFoundError, PermissionError,
    RateLimitError, OverloadedError, UsageExportError, UserManagementError,
    InviteUsersError. The Anthropic-compatible route defers to Anthropic's own error
    documentation.
  catalog: errors/langdock-problem-types.yml
rate_limiting:
  default_rpm: 500
  default_tpm: 60000
  scope: per model, per workspace
  status_code: 429
  headers_documented: false
  custom_limits: BYOK workspaces can configure custom per-model limits in Settings.
  detail: rate-limits/langdock-rate-limits.yml
streaming:
  supported: true
  detail: >-
    Completion endpoints support streaming responses; the OpenAPI defines
    CompletionChunk / CompletionEvent / DeltaMessage streaming schemas.
file_handling:
  upload: POST /attachment/v1/upload (multipart) for agent attachments; POST /knowledge/{folderId}
    for knowledge-base files
  delete: DELETE /attachment/v1/delete, DELETE /knowledge/{folderId}/{attachmentId}
  limits: POST /skills/v1/import returns 413 when the uploaded archive is too large.
related:
  authentication: authentication/langdock-authentication.yml
  errors: errors/langdock-problem-types.yml
  lifecycle: lifecycle/langdock-lifecycle.yml
  rate_limits: rate-limits/langdock-rate-limits.yml
  data_model: data-model/langdock-data-model.yml