API Calls | IMI Life Science

API Calls

Purpose: Inspect API destinations and request/response diagnostics for IMI Life Science.

What You Can Do Here

  • Find calls: Filter by exact destination host, request path, service, direction, outcome, and time range.
  • Inspect a call: Expand a row to see its destination, payload, and response. Developer Tools reveals the retained raw representation.
  • Follow logs: View in logs opens the exact record with surrounding context. Application logs flag API calls with their direction and destination.
  • Refresh: Reload the list to include recently completed calls.

Capture and Retention

  • Scope: Calls recorded after the updated services start include inbound API requests and outbound factory-managed HTTP calls. Direction is relative to the emitting service.
  • Bodies: JSON and form bodies are limited to 256 KiB and credential fields are redacted. Sensitive and diagnostic endpoints, unsupported content, and oversized bodies show omission reasons.
  • Streaming: Outbound records complete when the response is consumed or disposed. These records describe logical HTTP calls, not individual retry attempts.
  • Access: Log-read permission is required. Standard log retention and purge rules apply; expired and historical uncaptured payloads cannot be recovered.

External Integrations

  • Integration: Select ERP, Customer Services, Salesforce, or Stripe. Labels remain stable when hosts are routed through a gateway.
  • Destination: The internal-services gateway records its real upstream HTTP hop separately. Direction is relative to the emitting service; SQL and SMB operations are not HTTP calls.
  • Slow calls: Use duration thresholds or exact millisecond bounds, then sort by Duration. Time to headers separates the wait for a response from body consumption time.

Inspect Long Payloads and Responses

  • Browse: Open fields or array indexes and follow the exact path shown above the body. Collection pages show 50 entries at a time without discarding later items.
  • Search: Search nested field names and scalar values, then open a matching path. Searches are capped at 500 matches.
  • Expand: Open either body in a large independent inspector. Copy and download include only the retained, redacted content.
  • Developer Tools: Use the read-only source view with syntax highlighting, line numbers, search, a minimap, and the current line range.
  • Capture bounds: JSON and form bodies up to 256 KiB are retained. Larger, incomplete, sensitive, and unsupported bodies display omission reasons; omitted data cannot be reconstructed.

Filter and Sort

  • Filters: Combine integration, direction, service, method, exact host, path, HTTP status, duration, and custom dates.
  • Sorting: Sort the complete filtered result by time, duration, status, service, method, or path, not just the current page.
  • Share: Filters and ordering remain in the page URL. Use destination tags for quick filtering, remove individual filters, and select 25, 50, or 100 calls per page.

Connections Versus Requests

  • Flow: Distinguish HTTP requests, SignalR, WebSocket, Server-sent events (SSE), gRPC, and explicit HTTP streams.
  • SignalR transport: Negotiation and ordinary message requests remain separate from WebSocket, SSE, and long-polling connections.
  • Duration: Connection lifetime, stream lifetime, and poll wait include intentional waiting. These records do not receive a Slow label solely because they remain open. Failed handshakes and transport errors remain visible.
  • Handshake: An outbound WebSocket record measures only the HTTP upgrade handshake. Inbound records measure the lifetime observed by the request middleware.
  • Slow calls: A duration threshold excludes explicitly identified long-lived connections by default. Enable Include long-lived connections to inspect those lifetimes as well.

Response Availability

  • Not empty: WebSocket frames, SSE events, and streaming messages are not retained as a single HTTP response body. A compact state explains what is not captured; it does not imply that the connection transmitted no messages.
  • Idle polls: SignalR long polling can legitimately return no messages after waiting.
  • gRPC: Binary messages are not captured. HTTP metadata alone cannot determine whether an RPC is unary or streaming.
  • Historical records: Protocol badges may be inferred from saved status, path, or content type. A long duration alone never identifies a connection. Server flow filters apply to explicitly classified records; unclassified history remains visible.

Other Integrations and SOAP

  • Other: Includes every captured call outside ERP, Customer Services, Salesforce, and Stripe, including missing, empty, and unrecognized historical integration tags. No integration registration is required for factory-managed HTTP calls.
  • Narrow the results: Combine Other with Outbound, a service, destination host, or SOAP flow. Other can include internal calls as well as external vendors.
  • Validation services: Swiss UID SOAP validation and VIES/HMRC HTTP validation use the same captured transport. SOAP faults remain visible with their HTTP status and retained XML response.
  • XML bodies: Browse and search XML elements, attributes, and repeated nodes as structured fields. Developer Tools shows the retained XML source. Copy/download preserve the redacted representation.
  • Safety: XML capture is limited to 256 KiB. SOAP headers and secret-named elements/attributes are redacted; external entities, DTDs, malformed content, and unsupported bodies are not exposed.
  • Coverage: This is HTTP diagnostics, not a packet recorder. Binary/stream messages and sensitive bodies can be omitted while the call remains visible. Database, SMB, and browser-only third-party traffic are outside server HTTP capture.