zlay
atproto relay in zig zlay.waow.tech
relay zig atproto
Star
Fork
Watch

Commits on codex/prefix-format-space
Record offline midstream process-kill recovery across all variants
committed
Kill replay midstream and verify delivered revisions without re-ingestion
committed
Prepare replay oracle for delivered-revision crash checkpoints
committed
Publish ten-minute paired live run with qualified observer evidence
committed
Explain the shared cold-start resolver limit from measured counters
committed
Chart resolver waiting alongside collection index time
committed
Separate authority resolver service time from pool waiting
committed
Record paired public pilot smoke and cold-start backlog
committed
Respect observation qualification in live comparison charts
committed
Analyze paired pilots with the external evaluator cohort
committed
Publish same-input offline codec screen with resource chart
committed
Compare pilots against exact shared reference DID cohort
committed
Wait for assigned addresses before probing public pilot readiness
committed
Order proxy handlers as TLS then HTTP and retain readiness diagnostics
committed
Coordinate external evaluation while retaining bounded pilot cleanup
committed
Separate public relay pilots from their evaluation VM
committed
Expose coherent log and compaction gauges without blocking scrapes
committed
Document the corrected full-relay evaluator and network budget
committed
Expose a bounded network budget for live relay windows
committed
Record isolated reference window and evaluator transport checks
committed
Qualify collector intervals and support isolated reference probes
committed
Publish matched-resource live controls and disqualify failed references
committed
Exercise grouped tables through offline relay replay and restart
committed
Record the network-bounded grouped relay run without claiming qualification
committed
Include control index size from the exported live inventory
committed
Require durable RocksDB index writes in the matched live control
committed
Reject incomplete live observation windows after exporting evidence
committed
Keep live queue charts from clipping growing backlogs
committed
Distinguish grouped and legacy relay series in live charts
committed
Record grouped live startup probe and dependency fingerprints
committed
Account for actual vCPU count in live relay analysis
committed
Freeze live verification samples and require complete delivery checks
committed
Record and verify table encoding in bounded Fly relay runs
committed
Allow opt-in grouped tables in the isolated relay
committed
Pin ZKV allocation and decoder improvements for relay evaluation
committed
Compare relay pages across reader revisions and formats
committed
Compare prefix formats with paired offline read traces
committed
Measure real relay collection pages on identical traces
committed
Select the grouped-prefix core in an isolated index experiment
committed
Document matched-input index disk measurements
committed
Add offline index space and crash recovery experiment
committed
docs: record the isolated short-replay canary and verification
committed
test: freeze replay samples and account for inactive repositories
committed
deploy: retain existing canary addresses across replacement
committed
logs: report effective replay retention and byte budget
committed
deploy: reuse an empty canary app when replacing its machine
committed
http: accept bounded proxy-added WebSocket headers
committed
deploy: diagnose public WebSocket handshakes before replay probes
committed
deploy: bootstrap auto-destroy canary without machine updates
committed
deploy: add an isolated small ZKV relay canary
committed
replay: support small disk windows without RAM history
committed
docs: compare live relay coverage on sustained CPUs
committed
bench: summarize paired live coverage and processing costs
committed
bench: budget binary transfer time before live measurements
committed
bench: support sustained CPUs and record VM scheduling pressure
committed
bench: evaluate live pilot coverage against reference relays
committed
metrics: expose opt-in live processing stage timings
committed
docs: record ten-minute live relay capacity evidence
committed
bench: consume full-host relay for ten minutes
committed
bench: warm the live relay before attaching and checking a fresh consumer
committed