<!--
This file was generated by Spark. Do not edit it by hand.
-->
# AshHooks.OutboundAttempt

<a id="AshHooks-OutboundAttempt"></a>

Turns the consumer's resource into the outbound attempt history: one
durable row per delivery attempt, written by the delivery runtime beside
its fenced result transitions.


<!-- not-run: This resource fragment requires host PostgreSQL configuration and a database migration. -->

    use Ash.Resource,
      data_layer: AshPostgres.DataLayer,
      extensions: [AshHooks.OutboundAttempt]

The extension injects the attempt fields: the complete encoded delivery
key (`delivery_pk`), `endpoint_id`, `event_uuid`, `event_type`, the
attempt number (`attempt`), the outcome (`:succeeded | :failed_retryable |
:dead_letter`), the classified `reason`, `response_status`,
`response_snippet`, `response_headers` (allowlisted capture only),
`latency_ms`, `signing_key_ref` (a store reference, never secret material),
`test`, and `attempted_at`. Every attempted request, including one answered
with HTTP 410, records its duration and signing-key reference. A 410 disable
stored before 3.1.0 carries neither, so its finalization records both as nil.
The extension also injects
the `unique_attempt` identity on `[:delivery_pk, :attempt]`, which
makes the runtime's history writes idempotent per attempt. Your
migration carries the matching unique index.

`outbound_attempt do record_signing_provenance :on end` additionally injects
nullable `signing_key_refs`, the ordered native and host references that
actually signed the request. Migrate before enabling it. Pre-send failures
have no signing references or latency; recovered 410s retain the original
attempt's provenance rather than reading current endpoint configuration.

Response bodies are never stored here unless the ENDPOINT opted in per
row (`response_capture_attributes :on` on the endpoint resource and
`response_capture: true` on the row): an opted-in capture bounds the
bytes by the endpoint's `response_capture_max_bytes`, passes the
package's redaction floor, and keeps only response header names in the
endpoint's `response_header_allowlist`. Without capture, the snippet is
the fixed no-body summary grammar and `response_headers` is nil.

The runtime writes through the injected `:record` machine action
(`authorize?: false`); a history write failure never fails the send;
it surfaces as `[:ash_hooks, :attempt, :record_failed]` telemetry. The
delivery row remains the delivery contract; this ledger is the audit
trail behind the delivery log, health computation, and resend views.

Read access is yours to govern; the package injects no read policies.
Rows carry response summaries and (only under per-endpoint opt-in)
redacted response material.

Policy obligation: protect the injected `:record` create and `:prune`
destroy actions wherever your domain exposes the resource; the
runtime's internal calls bypass authorization by design.


## outbound_attempt







### Options

<table><thead><tr><th style="text-align: left;">Name</th><th style="text-align: left;">Type</th><th style="text-align: left;">Default</th><th style="text-align: left;">Docs</th></tr></thead><tbody><tr><td style="text-align: left;"><a href="#outbound_attempt-record_signing_provenance" id="outbound_attempt-record_signing_provenance"><code class="inline">record_signing_provenance</code></a></td><td style="text-align: left;"><code class="inline">:off | :on</code></td><td style="text-align: left;"><code class="inline">:off</code></td><td style="text-align: left;">Injects nullable signing_key_refs, every native and host key reference actually used by an attempt. Migrate before enabling.</td></tr></tbody></table>







<style type="text/css">.spark-required::after { content: "*"; color: red !important; }</style>
