Engineering blog
Real engineering problems we ran into while building an LLM API gateway.
-
What is an LLM API Gateway? A Practical Introduction
A technical introduction to LLM API gateways — what they do, when you need one, and how they differ from traditional API gateways. Covers architecture, BYOK, billing, and cross-protocol translation.
-
Choosing an LLM Gateway in 2026: SaaS, Library, or Self-Hosted
The LLM gateway category has settled into three architectural shapes — managed SaaS, in-app library, and self-hosted service. A practical comparison of trade-offs, plus a decision matrix and migration paths.
-
BYOK Pricing for LLM Gateways: The Billing Invariants That Matter
BYOK looks like a feature checkbox but is the largest class of correctness bugs in production LLM gateways. Five billing invariants we enforce by test, with concrete examples of what breaks when each is violated.
-
Prompt Caching Across LLM Providers: What Translates and What Doesn't
Anthropic, OpenAI, and Gemini all support prompt caching with different APIs, TTLs, and pricing. A gateway has to translate between them — and some of the translation is lossy. The rules, the edge cases, and the tests that catch the bugs.