AI Grimoire
09 — siblings
serving/
tuning/orchestration/

Tooling / serving

Serving and runtimes

Everything between a checkpoint on disk and a token on a wire. The split that matters is throughput against convenience: a datacentre server batching hundreds of concurrent requests is solving a different problem from a binary that runs a quantised model on a laptop, and the two families have almost no code in common.

4 entries.

Entries

09.01.1
llama.cppstandard
A dependency-free C++ inference engine, and the GGUF format everything local uses.
09.01.2
Ollamastandard
A local model runner: pull a name, get a server, forget the flags.
09.01.3
SGLangcommon
A serving runtime built around prefix reuse and constrained decoding.
09.01.4
vLLMstandard
The throughput server: paged KV cache, continuous batching, one endpoint.