Skip to content

docs: clarify timeout-vs-retry behavior and retry headers for API Request Tool - #1285

Draft
stephenvapiai wants to merge 1 commit into
VapiAI:mainfrom
stephenvapiai:helpcenter-update/api-request-tool-timeout-retry-headers
Draft

stephenvapiai wants to merge 1 commit into
VapiAI:mainfrom
stephenvapiai:helpcenter-update/api-request-tool-timeout-retry-headers

Conversation

@stephenvapiai

Copy link
Copy Markdown
Contributor

Clarifies two recurring API Request Tool reliability questions on the existing "Handle API Request tool latency and retries" page.

@stephenvapiai

Copy link
Copy Markdown
Contributor Author

Why this change

Recurring customer confusion about API Request Tool retry/timeout behavior, evaluated via the Enterpret Context Graph over the last 7 days vs. the prior 7 days and a trailing 28-day baseline (customer-facing sources: Plain support tickets, Fern Ask AI conversations, Discord, Posthog NPS).

  • 9 distinct customer conversations this week, 22 more in the prior 3 weeks, asked about backoffPlan / timeoutSeconds / excludedStatusCodes for the API Request Tool — a sustained, recurring pattern rather than a one-off spike.
  • Two specific sub-questions came up repeatedly and are not answered by the existing "Handle API Request tool latency and retries" page:
    1. Whether a request that times out with no response (vs. a received non-2xx response) is treated as retryable under backoffPlan.
    2. What request headers accompany a retry attempt vs. the original request.
  • The existing /observability/logs/webhook-logs page corroborates that this is a genuinely open question in the docs today: it states a long request duration "can indicate retries or a timeout, but does not prove either one by itself," rather than defining the relationship.

Classification: Incomplete content — the page covers backoffPlan/timeoutSeconds/excludedStatusCodes well for the non-2xx-response case, but doesn't address the timeout-with-no-response case or retry-attempt headers.

What this PR does

  • Adds a new "Timeouts versus retries" section stating what's confirmed from the schema (retries are defined in terms of HTTP status codes on a received response) and explicitly flagging the two unconfirmed behaviors as open questions, with cross-links to Webhook logs and support.
  • Does not assert a specific timeout-retry or retry-header behavior that isn't already confirmed in the docs/schema — both open items are left as explicit <!-- TODO (docs team) --> comments for the docs team to confirm with engineering and then finalize, per the customer-quote/evidence-integrity rule against fabricating unverified product behavior.

Scope note: Per the current run's candidate filter, security/auth/authorization/privacy/regulatory/compliance topics were explicitly excluded from consideration this run. This gap (HTTP retry/timeout mechanics for the API Request Tool) does not touch any of those categories.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant