BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//programme.europython.eu//europython-2026//speaker//PSCB8
 W
BEGIN:VTIMEZONE
TZID:CET
BEGIN:STANDARD
DTSTART:20001029T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000326T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-europython-2026-BYWVNK@programme.europython.eu
DTSTART;TZID=CET:20260717T153000
DTEND;TZID=CET:20260717T160000
DESCRIPTION:Most LLM tutorials end where production begins. When OpenAI ret
 urns a 429\, when Claude’s latency spikes 10x\, or when your streaming r
 esponse dies mid-generation—what happens to your users?\n\nThis talk cov
 ers battle-tested architecture patterns for production LLM streaming\, mov
 ing beyond simple API calls to resilient systems. We will explore multi-pr
 ovider failover chains (OpenAI → Anthropic → local)\, circuit breakers
  specifically configured for AI workloads\, and token-aware rate limiting 
 that protects both latency and cost.\n\nYou will learn framework-agnostic 
 Python patterns using asyncio and LiteLLM for provider abstraction. We wil
 l examine real incident patterns—including the December 2025 Anthropic o
 utage—and the architectural decisions that separate 99.5% availability f
 rom 99.9%.
DTSTAMP:20260524T121632Z
LOCATION:Chamber Hall B (S3B)
SUMMARY:Beyond the Demo: Production Patterns for Streaming LLM Systems - Ni
 tish
URL:https://programme.europython.eu/europython-2026/talk/BYWVNK/
END:VEVENT
END:VCALENDAR