sno-ai/llmix: Production LLM Call Layer Adds Cache, Retries, Circuit Breakers, and Key Rotation Across Providers

GitHub·low signal

sno-ai/llmix (129 stars, early-stage) is a production LLM call layer that keeps existing OpenAI/Anthropic/AI SDK/LiteLLM code while adding hot-swappable models via MDA presets, caching, retries, circuit breakers, key rotation, and singleflight deduplication. It targets reliability concerns that bite agents at scale. For builders, it's an early but focused resilience wrapper worth tracking as it matures.