4 Commits

Include renames

Author	SHA1	Message	Date
sneak	018c12a8d2	Count hits, misses, fetches and transforms as Prometheus counters (closes #311) ... The five counts the cache_stats table held are now Prometheus counters in memory, as fields of the cache. NewCache makes them and registers them on the registry it is given. pixad gives it Prometheus' default registry, which /metrics serves; each test gives it a new one. The Increment methods add to their cache's counters, so counting makes no database call. cache_stats, Cache.Stats, Service.Stats, CacheStats and the pending counts are gone, with their tests. The background goroutine now writes only access times, and its names say so: StartPendingAccessTimeWrites, StopPendingAccessTimeWrites and pendingaccesstimes.go. The two stop tests whose write under way was a count now use access times. Model: opus-5-5	2026-10-26 09:12:33 +00:00
clawbot	99735f479b	Answer image requests with 503 in maintenance mode (closes #71) ... maintenance_mode was only reported by the health check; every request was still served. One middleware in routes.go, applied to /v1/image/ and /v1/e/ only, now answers them with 503, a Retry-After of MaintenanceRetryAfterSeconds and the JSON error body while it is on. It calls Server.MaintenanceMode(), which had no caller. The health check stays 200 and reports maintenance_mode: the image's Docker HEALTHCHECK requests it, a 503 there would make the container unhealthy, and upaas marks a deploy failed when its container is unhealthy. The login and URL generator pages and /metrics keep working. Documented in README.md and config.example.yml. Model: opus-5-5	2026-09-29 11:01:30 +02:00
clawbot	56217cbf4a	Add the four settings the README documented but pixa lacked (closes #61) ... README.md documented access_control_allow_origin, upstream_fetch_timeout, upstream_max_response_size and downstream_timeout, but pixa did not know them, so a config following the README aborted startup. Each is now a setting with its PIXA_ variable, defaulting to the value that was fixed in the code: *, 30s, 50 MiB and 60s. Durations are Go duration strings and must be positive; the size is whole bytes, at most 1 GiB. The origin is * or one http or https origin written exactly as a browser sends it; anything else aborts startup. downstream_timeout sets both the server's write timeout and the per-request timeout. The owner approved the edits to existing tests. Model: opus-5-5	2026-09-29 08:49:18 +02:00
clawbot	e410146fb6	Rate limit login attempts per client address (closes #66) ... POST / had no limit, so the signing key could be guessed at no cost. It is now limited to 5 attempts per minute per client by a new RateLimit middleware on github.com/go-chi/httprate; an attempt over the limit gets 429 with Retry-After. It counts by the address the ClientIP middleware resolved through trusted_proxies (an IPv4-mapped address as its IPv4 address, IPv6 by its /64) and runs after the body-size and CSRF checks, so every attempt that reaches the key comparison is counted. README says that with the default trusted_proxies a client with a private address can choose its counted address, and how to close that. Model: opus-5-5	2026-09-29 01:03:37 +02:00