v0.3.1

Changelog

• 0e40474 Restyle the website and the plot page: Archivo and Plex Mono, looser headings, a shorter hero, flat charts with round axis ticks, a line saying whether the two clocks agree
• ccfbb20 mcpserver test: skip when either of the two runs was limited by the generator

v0.3.0

Changelog

• 999ea84 CI: allow a manual run
• 5ed9c72 Count a phase that ran even when the clock read it as zero
• bbb62c3 Docs: say Gmeter is an agent-friendly load tester, and document gmeter mcp
• 7f03757 MIT licence, a gRPC example, v3g on the roadmap
• 423ce05 Make the suite pass on Windows and macOS runners, not only on this Mac
• d30dd16 README: replace the finished roadmap checklist with what is coming next
• d65fafd Test: an HTTP/2 request to an HTTP/1 server may see a reset first
• 9daa876 Tests: assert that a phase ran, not that it took more than zero
• 9bfe510 Tests: do not pin the exact rate a search finds, and skip when the machine is too busy to trust timing
• 3cf7238 The Go package calls gRPC too: Config.GRPC, Result.Codes
• 9be4428 Website: say what k6 offers Go users, which is extensions
• 22032aa Windows builds, a Docker image, and a CI run on all three systems
• b82de24 gmeter mcp: serve the tools to an AI agent over stdio
• cba6844 mcp: a policy for what an agent may target and how much load it may ask for
• 28aa24e mcp: a server with a load_test tool, over HTTP and unary gRPC
• 1eb1970 mcp: find_capacity, check_slo, compare_runs, a run resource and progress

v0.2.0

Changelog

• 5dc18bd A public Go package: gmeter.Run, Result.Meets and FindCapacity
• dac2dd2 Client certificates: -cert and -key for mutual TLS
• 9e296e9 Per-request results and an HTML chart: -results and gmeter plot
• 4a7ff24 Website: install with Homebrew
• 86fcabf Website: plot, the Go package and client certificates; settle in-flight requests in the agent stop test

v0.1.0

Changelog

• e7f0d39 Add -endpoint-threshold to gmeter diff, and refuse thresholds that cannot fail
• 77caa21 Add -endpoints to gmeter replay: a per-endpoint breakdown of the results
• bad3f7d Add -error-threshold to gmeter diff
• f2e9cef Add -http2 to run, find and replay
• 12ed581 Add -insecure and -cacert for https targets
• 159e06e Add -phases to runs and replays, on by default
• f14a391 Add -spin flag, save it in the run info, and flag it in diff
• 730ce2e Add -stages, -arrivals and -seed
• 5335ace Add README
• 5e9c0fe Add Recorder: histogram readable while its owner writes
• f08610f Add gmeter diff, run info in the JSON, and credential redaction
• 5fc3ad5 Add gmeter find: search for the highest rate that meets a latency target
• 1213aaf Add gmeter replay: send a traffic log at its recorded pace or a fixed rate
• 1f25fa9 Add a one-page website in docs/, for GitHub Pages
• 0201b16 Add fixed-memory HDR-style latency histogram
• cbb2c62 Add gmeter -version and a release pipeline for Homebrew
• b3264e6 Add gmeter CLI
• 71e3dda Add gmeter agent: run a share of a test for a controller
• 4b2d669 Add histogram Dump/FromDump and FractionAbove
• 58fc9c2 Add hybrid sleep-then-spin pacing
• 3fe19e9 Add open-loop pacer with drift-free schedule
• bcd4b15 Add open-loop runner with corrected and service latency
• 53ac4b0 Add replay package: read JSON Lines and access-log traces
• 3ac4db4 Add run comparison: relative change, share-slower and like-for-like notes
• ae6f0db Add text and JSON reporting
• 696b5cd Add the shared-cause line to the endpoint gate, and document it
• 281c976 Add the slowest phase to the endpoint gate's verdict
• d8b897f Count bad statuses per stage; export the timer-lag threshold
• 6b35484 Document -group and per-endpoint wait p99
• 5ffbf0b Document -spin; regenerate the README samples with the new pacing
• c7cd691 Document HTTP/2, with a real comparison against HTTP/1.1
• bfa5fb4 Document gmeter find
• bec73d0 Document gmeter replay
• 9109e12 Document and test the status comparison for replays
• c2e8281 Document gmeter grpc, and point at -plaintext when TLS meets none
• 8aad639 Document how much memory a replayed trace takes
• a7ef971 Document load profiles, Poisson arrivals and the late tolerance
• cce8f0a Document phases in diff and per endpoint; v2 is complete
• 8b993f9 Document the per-endpoint breakdown
• ff7ba45 Document the per-endpoint diff
• 550033e Document the phases, and add a benchmark of what tracing costs
• 5a258df Fix literal \n in gmeter find progress and usage output
• 901ffa2 Let a run make any call instead of an HTTP request
• a5a06e2 Load-test a unary gRPC method with gmeter grpc
• bd210bd Make the website's opening grab attention
• 3a0b38f Open the website with a plain intro to what Gmeter is
• 47280de Pad Recorder to cache lines to stop false sharing
• 66741ea Polish the website: drop glows, heavy weights and off-grid spacing
• 4fca890 Record a per-window timeline for every run
• 5690f72 Release CI: skip the two timer-precision tests with -short
• 18364ac Rename ProgressEvery to SampleEvery
• 918264f Replay: keep a trace in about half the memory
• 79f68a4 Replay: keep the status each logged request got
• 01121a9 Replay: name each request's endpoint and cap how many are reported
• a912e6e Replay: name endpoints by -group rules
• adc5ab4 Replay: warn when a short trace is cycled
• 0ac9ab9 Report per-window latency stats in progress snapshots
• c7170fb Report stages, and stop counting microsecond lateness as late
• df24cc4 Report: "where the time goes" table and "phases" in the JSON
• 1d846fa Report: compare phases in a diff
• 100d04d Report: compare phases per endpoint in a diff
• ba6a269 Report: compare the endpoints of two runs in gmeter diff
• 1d76e84 Report: compare the error rate of two runs
• a2a65f3 Report: count responses by protocol, explain protocol errors, and note a protocol change in diff
• ecc10f0 Report: expose both runs' error rates
• 9019f97 Report: judge each endpoint's error rate
• 32b4d0e Report: note when most endpoints rose together
• 20c66d7 Report: per-endpoint table and "endpoints" in the JSON
• 3609fbf Report: say how many responses differ from the log
• be3ddd6 Report: say what to do about TLS failures
• 0413483 Report: show each endpoint's wait p99, and how it moved in diff
• 5d33fec Report: where each endpoint's time goes
• 6402bf1 Run a test or a replay on several agents with -agents
• 1cbbb55 Runner: compare each response's status class with the logged one
• bc2ab0d Runner: fix a data race in the HTTP/1.1 TLS test
• 02caa64 Runner: keep each label's wait as a distribution
• 840b841 Runner: merge the results of several generators
• 2ac16e0 Runner: phases per label
• ed9dd35 Runner: replay recorded times and per-request method, target, headers and body
• f4b7729 Runner: report results per request label
• 7da35d5 Runner: say precisely when a label's corrected and service latency differ
• 6ddabfd Runner: send over HTTP/2 when asked, and count each response's protocol
• f4c5e4b Runner: split each response's corrected latency into phases
• 86d36d4 Runner: split each stage's time into phases
• 55fc84e Runner: stop the late-count test from measuring the machine
• 9c801f1 Runner: take a TLS configuration, and name TLS failures
• fff4afc Save timeline, run info and full histograms in the JSON summary
• 67b8345 Schedule profiles of stages with Poisson arrivals, and per-stage results
• b3e40f1 Send a run's result as JSON with its full distributions
• cf9faf7 Show generator health and warnings in the report
• a388da2 Show the latest window in the live progress line
• 890873a Split a test among agents and merge what they send back
• 71bbf14 Split the pacer's waiting code out of the schedule file
• 125a5b8 Track generator health: timer lag and late dispatches
• 511c923 Update README for v1a
• 4b64adc Update README for v1b
• 0fb779d find: say where the time grew between the answer and the first failure
• 2cb200e gRPC: count a call past its own deadline as a timeout, every time