Show HN: RelayFreeLLM – Free AI Gateway with Auto-Failover (Update) (github.com)

Author here. Since the last post, RelayFreeLLM now supports NVIDIA's free catalog, adds 4-mode context management (with extractive summarization that doesn't need an LLM), session affinity, output normalization, and a global provider lock to eliminate 429 errors entirely. 8 providers, one endpoint, all free: https://github.com/msmarkgu/RelayFreeLLM