sno-ai/llmix Ships Production LLM Call Layer — Hot-Swap Models With Cache, Circuit Breakers, and Key Rotation

GitHub·low signal

LLMix is a production-grade LLM call layer that wraps OpenAI/Anthropic/AI SDK/LiteLLM, letting you hot-swap models via MDA presets while adding caching, retries, circuit breakers, key rotation, and singleflight deduplication. At 129 stars, it targets teams running multiple LLM providers in production who need reliability primitives without building them from scratch. The library separates model selection from application code.