Headroom OpenAI-Endpoint Compression Analysis

Why LLM compression proxies achieve zero savings through OpenAI-compatible endpoints, which proxy flags provide value regardless, and when to keep vs remove a passthrough proxy.

• OpenAI-Compatible Endpoints Block LLM Proxy Compression - Architectural mismatch between OpenAI chat format and Anthropic-native compression routing
• Headroom Proxy Flag Reference - Which flags provide value and which are blocked by the content router
• LLM Proxy Architecture Decision - Decision framework for keeping vs removing a non-compressing proxy