Connections and failure notifications
Use integrations only for services and data you are authorized to access. Saving configuration does not test it or send a message; verify behavior with a small controlled ordinary Run.
Browser and HTML network settings
Section titled “Browser and HTML network settings”Supported headers apply before target navigation. Use managed-secret references for supported sensitive header values. User agent overrides have target limits: Browser supports its available user-agent, language and platform controls, while HTML cannot provide a Browser navigator platform.
Request-filter entries match the complete URL with * as the wildcard. Matching target requests are blocked. These settings do not automatically filter unrelated connector calls.
The current proxy workflow supports a single unauthenticated HTTP(S) endpoint, or explicit direct access. An unavailable configured endpoint fails rather than silently switching to direct. Multiple endpoints, rotation, authenticated proxies, SOCKS and VPN remain unsupported.
Console/Desktop do not become Browser-network targets by inheriting these settings. Review effective configuration after switching Robot kind.
Named project connectors
Section titled “Named project connectors”Define a named HTTP connector in project settings, with its method, endpoint, supported headers and optional body. GET/HEAD do not use a request body. Supported Set actions can invoke a selected connector after successful extraction/formatting.
Connector definitions are project resources, not another inherited settings list for every Job/Task. Preserve literal names and review which credential references the selected project may use.
External result destinations
Section titled “External result destinations”Data and file destinations are independent. When an external destination is selected, it is the sole destination for that result family: do not expect a second internal business-data/file copy under Data or Files. Execution history remains separate from result storage.
An external file reference is informational in Eyerobot; it is not an internal downloadable file ID or a general arbitrary-URL preview. Consult the destination using its intended access method.
An uncertain external write may already have happened. Inspect the destination before retrying work that could duplicate it.
Final failure notifications
Section titled “Final failure notifications”The configured error event concerns a final ordinary Session failure after automatic retries. Toolbox Play does not trigger it. Configure its method/destination and optional throttle in Events.
The throttle controls minimum spacing between notification attempts, not a delivery queue. Suppressed or unavailable attempts can be dropped; no automatic notification retry is promised. A failed or uncertain send can consume the throttle interval. Zero disables the limiter, not the rest of the notification policy.
Review Attempt diagnostics for sent/suppressed/failed results. A saved Events form does not prove endpoint readiness or message delivery.
External page resources
Section titled “External page resources”Browser can load supported configured JavaScript resources for page use. A custom function relying on page-owned libraries must explicitly select Run in page context. HTML/native targets do not acquire page JavaScript through this field.
Treat libraries and custom code as part of the automation you author. They can change the target page or send data; do not supply credentials or customer content beyond what the task requires.