ContributorMural::Embedder
Fetches avatars in parallel and turns them into base64 data URIs. Results are cached by URL so multiple render targets reuse fetches.
Constants
The work here is almost entirely waiting, so the ceiling is set by politeness to the host rather than by anything local. Connections are kept alive now, so a worker pays for one handshake instead of one per avatar — which is what makes a wider pool worth having rather than just louder.
Constructors
Instance methods
Returns users with embedded avatars (input order preserved) and the people
skipped due to fetch failures. Raises the first failure instead when
fail_on_missing is set.
Who kept the face the previous wall already had, because this run could not fetch theirs. Accumulated here rather than returned per target: the cache is what makes it a run-wide fact, and the caller reports it once.
Fetches everything every target will ask for in a single fan-out. A config with more than one output used to fetch per target and wait out a separate round of latency each time, even where the two targets wanted the same faces at the same size.