AE-01V5S060
Voice-agent LLM plugin retries at two layers (vendor SDK plus framework), so attempt timeouts are exceeded and failover is delayed
RETRY LOOPseverity: MEDIUMcause: LIKELYoutcome: RESOLVED UNVERIFIEDconfidence: LOW
The Anthropic LLM plugin created its SDK client without max_retries, inheriting the SDK default of 2 internal retries on top of the framework's own retry loop. Each SDK retry got a fresh timeout, so one configured 5-second attempt per fallback tier could take about 16.5 seconds before failing over.
- Framework / agent
- LiveKit Agents · real-time voice agent with LLM fallback adapter
- Remediation attempts
- TESTED
- Recurrence
- not documented
- Source languages
- en
- Updated
- 2026-09-30
Sources
- GITHUB ISSUE anthropic plugin: vendor SDK retries not disabled, double-retrying on top of conn_options.max_retry — github.com/livekit/agents, retrieved 2026-09-29
- GITHUB PULL REQUEST fix(anthropic): disable vendor SDK retries by default — github.com/livekit/agents, retrieved 2026-09-29
Symptoms
- Requests are retried at two independent layers and the configured timeout no longer bounds a turn's wall-clock time
- With max_retry_per_llm=0 and attempt_timeout=5.0, each tier can consume ~3 x 5s plus backoff (~16.5s) before failover
The full record — root-cause evidence, every remediation attempt with its status and verification, failed attempts, patch references, verbatim quotes and recurrence — is a paid lookup (0.018 USDC via x402). Agents:
GET /api/v1/cases/AE-01V5S060. PricingSimilarity to your system is not implied. A remediation that worked in the documented context may not work in yours.