3 Commits

Author SHA1 Message Date
bvandeusen 681d7777f3 fix(build): install CPU-only torch in ml image (drops ~5.6GB CUDA layer) 2026-05-15 21:04:26 -04:00
bvandeusen 37e97d52ee chore: bump runtime targets to Python 3.14 and Node 22
Forward-looking pin per operator direction ("build for the future and not
rebuild the past"). Touches everywhere a version is named so all
downstream artifacts (Dockerfiles, ruff config, package.json engines)
agree.

Python 3.14 (released Oct 2025) is the current stable. Node 22 is the
most recent LTS line still receiving updates (Maintenance LTS since
Oct 2025); Node 24 (released Apr 2026) goes Active LTS in Oct 2026 and
will be the natural next bump.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-14 07:52:15 -04:00
bvandeusen ec03297382 feat: add Dockerfile.ml — separate image for the ml-worker role
ML deps (torch, transformers, onnxruntime, opencv) are ~4GB and stay out
of the regular web/worker image. Models self-heal into /models on start
(implemented in FC-2). HuggingFace cache vars point inside /models so
weight downloads are persistent across restarts.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-14 07:42:14 -04:00