Resilience, retries and fallbacks
How a call survives slow or failing AI services, from provider retries and alternate speech providers to caller feedback while waiting and ordered final actions, with defaults that teams and flows can override. Browse the feature map or configuration finder.
Availability: Next. Entries that depend on an optional feature, a role or a deployment choice name it in their own Availability line.
Features in this family
- Resilience controls availability
- Resilience defaults and inheritance
- Provider timeouts, retries and reconnects
- Retry on selected HTTP status codes
- Alternate speech recognition and synthesis providers
- Flow-level resilience settings
- Feedback while the bot is slow to answer
- Final call actions on runtime failure
Resilience controls availability
An optional deployment feature shows the platform, team and flow editors for recovery settings; hiding them leaves saved policies in force at runtime.
Availability: Requires the Resilience controls feature; available from 0.9.18. The editors are hidden unless resilience controls are enabled; with subscription management they appear only for teams on the custom plan. Saved policies keep applying when the editors are hidden.
For: Deployment operators, Team administrators
| Level | Where to change it | Owner section |
|---|---|---|
| Portal and feature profiles | Portal › Feature profile › Resilience controls(resilience) | Resilience controls |
| Services and containers | Web service environment(FEATURE_RESILIENCE) | Resilience controls |
Read more:
- Reference: Resilience controls
Related features: Resilience defaults and inheritance, Flow-level resilience settings.
Resilience defaults and inheritance
Recovery settings are defined once for the platform, can be replaced per team and per flow, and the closest complete block for a provider role wins.
Availability: Requires the Resilience controls feature. The editors follow the resilience controls availability; stored defaults still apply when hidden.
For: Deployment operators, Team administrators, Application builders
| Level | Where to change it | Owner section |
|---|---|---|
| Platform settings | Admin Settings › Platform Settings › Resilience defaults(resilienceDefaults, providerDefaults, failover) | Other Platform Settings controls |
| Team settings | Team › Resilience defaults(resilienceDefaults) | Set team resilience defaults |
| Flow and node settings | Provider editor › Timeouts & retries(resilience) | Resolve the complete retry policy |
Read more:
- How it works: Resolve the complete retry policy
- Set it up: Set team resilience defaults
- Reference: Other Platform Settings controls
Related features: Provider timeouts, retries and reconnects, Final call actions on runtime failure, Resilience controls availability.
Provider timeouts, retries and reconnects
Each speech, realtime or bot provider can get its own request timeout, number of extra attempts, delay with backoff, and a limit for reconnecting a dropped speech connection during the call.
Availability: Requires the Resilience controls feature; available from 0.9.18. The editors follow the resilience controls availability; stored settings still apply when hidden.
For: Application builders, Team administrators
| Level | Where to change it | Owner section |
|---|---|---|
| Flow and node settings | Provider editor › Timeouts & retries(timeoutMs, retryCount, reconnectAttempts, retryDelayMs, backoffFactor, retryOn, keepalive) | Retries, timeouts and fallback providers |
Read more:
- Reference: Retries, timeouts and fallback providers
- Troubleshoot: Resolve the complete retry policy
Related features: Retry on selected HTTP status codes, Resilience defaults and inheritance, Alternate speech recognition and synthesis providers.
Retry on selected HTTP status codes
A provider retry policy can name exactly which HTTP response codes count as worth another attempt, and an empty list turns off status-based retries only.
Availability: Requires the Resilience controls feature. The editors follow the resilience controls availability.
For: Application builders, Team administrators
| Level | Where to change it | Owner section |
|---|---|---|
| Flow and node settings | Provider editor › Timeouts & retries › Retry HTTP status codes(retryStatusCodes, DEFAULT_PROVIDER_RETRY_STATUS_CODES) | Retry HTTP status codes |
Read more:
- Reference: Retry HTTP status codes
- How it works: Resolve the complete retry policy
Related features: Provider timeouts, retries and reconnects.
Alternate speech recognition and synthesis providers
In modular voice flows an ordered list of up to five backup recognition or synthesis services takes over when the primary one cannot connect or fails during the call.
Availability: Requires the Resilience controls feature. The editors follow the resilience controls availability.
For: Application builders
| Level | Where to change it | Owner section |
|---|---|---|
| Flow and node settings | Provider editor › STT fallback providers / TTS fallback providers(sttFallbacks, ttsFallbacks) | STT and TTS fallback providers |
Read more:
- Set it up: STT and TTS fallback providers
Related features: Provider timeouts, retries and reconnects, Final call actions on runtime failure, Azure speech recognition, Speech synthesis providers (OpenAI, Azure, Inworld).
Flow-level resilience settings
A flow's resilience tab groups provider recovery, waiting feedback and final call actions, and can replace the inherited call-level policy for that flow.
Availability: Requires the Resilience controls feature. Without resilience controls the drawer is labelled Variables and shows only the variable catalogue.
For: Application builders
| Level | Where to change it | Owner section |
|---|---|---|
| Flow and node settings | Flow Designer › Variables & Resilience › Resilience(gatewayProfile.failover, resilience) | Set flow-level resilience |
Read more:
- Set it up: Set flow-level resilience
Related features: Final call actions on runtime failure, Feedback while the bot is slow to answer, Resilience defaults and inheritance.
Feedback while the bot is slow to answer
While a bot reply is outstanding the caller can hear a holding phrase or audio, repeated a set number of times, and recognition can resume so the caller is not left in silence.
Availability: Requires the Resilience controls feature. The editors follow the resilience controls availability.
For: Application builders
| Level | Where to change it | Owner section |
|---|---|---|
| Flow and node settings | Variables & Resilience › Resilience › Bot delay and terminal failover(botNoInputTimeoutMs, botNoInputSpeech, botNoInputUrl, botNoInputRetries, abortPlayOnBotMessage, resumeRecognitionTimeoutMs, resumeRecognitionSpeech, resumeRecognitionPlayUrl) | Give callers feedback while the bot is |
Read more:
- Reference: Give callers feedback while the bot is
Related features: Give up when the bot stops responding, Final call actions on runtime failure.
Final call actions on runtime failure
When a bot timeout, bot error, recognition, synthesis or internal failure reaches the call, an ordered list of up to ten steps transfers or ends the call, with a safety hangup if every step fails.
Availability: Requires the Resilience controls feature. The editors follow the resilience controls availability; saved policies keep applying when hidden.
For: Application builders, Team administrators
| Level | Where to change it | Owner section |
|---|---|---|
| Flow and node settings | Variables & Resilience › Resilience › Configure final call actions(failover.enabled, policies, triggers, bot_timeout, bot_error, stt_error, tts_error, system_error, transfer, hangup, botNoInputGiveUpTimeoutMs) | Configure final call actions |
| Team settings | Team › Resilience defaults › failover(failover) | Set team resilience defaults |
Read more:
- Set it up: Configure final call actions
- How it works: Set flow-level resilience
Related features: Give up when the bot stops responding, Resumable call sessions, Flow-level resilience settings, Alternate speech recognition and synthesis providers.