gpu-turnstile

6 Commits

Author	SHA1	Message	Date
mram	cf9be55a6c	Make poll intervals and operational timeouts configurable ci / test (push) Successful in 47s Details ci / docker (push) Failing after 30s Details New env vars: UNLOAD_POLL_INTERVAL, HISTORY_POLL_INTERVAL, PROBE_TIMEOUT, FREE_TIMEOUT, WARM_TIMEOUT, SHUTDOWN_TIMEOUT, PROMPT_CAPTURE_LIMIT. Defaults unchanged; invalid values fail fast at startup.	2026-09-20 19:21:25 +02:00
mram	0f950f3134	CI: use gitea.actor as registry login username ci / test (push) Successful in 48s Details ci / docker (push) Skipped Details Gitea authenticates registry pushes by token alone; a single REGISTRY_TOKEN secret suffices.	2026-09-20 18:55:46 +02:00
mram	73fb8ac3ad	CI: log in to registry with REGISTRY_USERNAME/REGISTRY_TOKEN secrets ci / test (push) Successful in 47s Details ci / docker (push) Skipped Details The automatic GITEA_TOKEN has no write:package scope, so docker login to git.rambossek.at failed with unauthorized.	2026-09-20 18:24:54 +02:00
mram	5a0e181293	CI: build release images only from vX.Y.Z tags ci / test (push) Successful in 47s Details ci / docker (push) Failing after 23s Details Tests still run on every push. The docker job is gated on ref_type=tag, validates the tag is a strict vX.Y.Z semver, and publishes :<tag> plus :latest. No branch images (sha-* / main latest) anymore. v0.1.0	2026-09-20 18:12:00 +02:00
mram	db11da8307	Rename to gpu-turnstile; swap port roles The proxy now listens on the standard service ports (Ollama :11434, ComfyUI :8188) and the actual services move one port up (:11435, :8189). Metric prefix is now gpu_turnstile_.	2026-09-20 18:10:22 +02:00
mram	065c294a96	gpulock: GPU arbitration proxy for Ollama + ComfyUI Implements SPEC.md: two listeners, one writer-preferring two-mode lock, Ollama unload before image jobs, ComfyUI history polling + VRAM free, optional model warm-up, healthz/metrics endpoints, streaming-safe reverse proxies, Dockerfile and Gitea Actions CI.	2026-09-20 18:05:20 +02:00