generated: '2026-07-19'
method: searched
source: openapi/linqalpha-openapi-original.json
docs: https://docs.linqalpha.com/quickstart
authentication:
  style: api-key-header
  header: X-API-KEY
  scoped: false
  multi_tenant_selector: organization_id (platform-tier keys)
  artifact: authentication/linqalpha-authentication.yml
idempotency:
  supported: false
  evidence: >-
    No Idempotency-Key header, parameter, or documented replay contract appears anywhere in the
    OpenAPI or the published documentation. Write operations (POST /v1/briefings,
    POST /v1/source_batches, POST /v2/connectors, POST /v2/vault/confirm) carry no client-supplied
    request key, so a retried create may duplicate. Long-running generation is instead made safe by
    the fire-and-forget + poll pattern below, not by idempotency.
pagination:
  style: page-number
  parameters:
  - page
  - per_page
  applies_to: list endpoints (e.g. GET /v1/briefings, GET /v1/briefings/{id}/deliveries)
  cursor_support: false
streaming:
  protocol: Server-Sent Events
  media_type: text/event-stream
  operations: 7
  endpoints:
  - POST /v2/chat/sse
  - POST /v1/analytics/sse
  - POST /v2/analytics/sse
  - POST /v1/rms_chat
  - POST /v1/briefings/sse
  event_names:
  - conversation
  - message
  - status (start | keep_alive | finish)
  - tool_use_block
  - tool_result_block
  - think
  - answer
  - chat_message_id
  keepalive: status keep_alive heartbeat emitted if there is a gap greater than ~15s
  cancellation: POST /v1/stop_stream pauses an in-flight RMS chat stream
async_pattern:
  style: fire-and-forget then poll
  examples:
  - enqueue: POST /v1/briefings/preview
    poll: GET /v1/briefings/deliveries/{delivery_id}
  - enqueue: POST /v2/vault/confirm
    poll: GET /v2/vault/status until status is Synced
  - enqueue: POST /v1/sources
    poll: GET /v1/sources/{source_id} until status is success
versioning:
  scheme: uri-path
  versions:
  - v1
  - v2
  note: >-
    v1 and v2 run side by side rather than v2 replacing v1; several resources are published in both
    (documents presigned URLs, references, analytics, chat). v2 variants add structured source
    selection, hard/soft filters, and RMS/vault document support. No version header, no date-pinned
    version train, and no announced sunset for v1.
  artifact: lifecycle/linqalpha-lifecycle.yml
error_envelope:
  format: vendor-json
  rfc9457: false
  shape: '{ "error": { "code", "message" }, "payload": null }'
  machine_readable_code: error.code
  artifact: errors/linqalpha-problem-types.yml
rate_limiting:
  documented_for: MCP server
  limit: 60 requests per minute per user
  signal: JSON-RPC error code -32000
  response_headers: none documented
  rest_api: no published REST rate limit
  artifact: rate-limits/linqalpha-rate-limits.yml
request_tracing:
  request_id_header: none documented
  correlation_ids:
  - conversation_id
  - chat_message_id
  - delivery_id
  note: >-
    There is no transport-level request-id header. Tracing is done at the domain level: an SSE
    stream emits conversation_id and chat_message_id events, and those ids are the handles used to
    fetch references, submit feedback, and run the grounding judge.
citations:
  note: >-
    Answers carry inline [N] markers that resolve to evidence references retrieved via
    GET /v2/analytics/conversations/{conversation_id}/references. Source documents are fetched
    through short-lived presigned URLs (10 minute expiry).
field_expansion:
  supported: false
metadata:
  supported: false