Skip to content

Releases

0.1.0-alpha.4 - 2026-11-09

Pre-release

Choose a tag to compare

@github-actions github-actions released this 09 Nov 14:54

Release Notes

Breaking

  • Protocol, schema and configuration use runner-neutral names (ADR-0014, #153): the event runner.detected { runner, version, profile_supported } replaces maestro.detected; the artifact class runner_log replaces maestro_log; the agent-run and telemetry fields are runner_exit_code, runner_exit_signal, runner_version, runner_profile_supported, runner_cpu_pct, runner_rss_bytes, runner_cpu_avg and runner_rss_max; the exit class is runner_error; the metric scope is runner; the failure-fingerprint dimension is runner; the configuration key is artifacts.max_bytes.runner_log. The migrations carry these names from the start.
  • There is no compatibility before v1.0.0. The alpha releases cannot be upgraded: start with a fresh database, and upgrade the agents and the server together, since an alpha agent cannot talk to this server.
  • cargo xtask upgrade-rehearsal replays release tags from v1.0.0 on. With none yet it reports that there is nothing to rehearse and succeeds.

Added

  • Agent: Maestro's own takeScreenshot images and startRecording videos can be uploaded, under the new classes user_screenshot (PNG) and user_recording (MP4). Off by default: fleet-agent run --upload-user-media or FLEET_UPLOAD_USER_MEDIA=1 turns it on, and then only a failed flow's files are uploaded, each linked to the step that made it. Their pixels are not redacted, so the switch is the operator's decision (#133).
  • Server: migration 0033 adds the two classes to the artifact allowlist and to the per-project policy, and the retention class user_media (14 days by default), seeded for every project. Each class has its own size cap (artifacts.max_bytes.user_screenshot, 20 MiB; artifacts.max_bytes.user_recording, 100 MiB), its own content type (image/png, video/mp4) and the usercontent serving rules: a recording is served inline, with ranges, and every read of either is audited (#133).
  • Dashboard: screen D shows a failed flow's own screenshots and recordings under the step that made them: an image inline, a recording in the browser's player, both from the usercontent origin. The dashboard's policy allows that origin in media-src (#133).
  • Docs: the security page's T22 row covers the opt-in and its controls; the runner, operator and user guides describe the switch, the classes, their caps and their retention (#133).
  • Dashboard: screen E lists the flow catalog, every flow the project knows, including flows that have never run (shown as "Never run", never as a pass). The agent sends the workspace's flows with their header tags in hello.catalog, and a stored plan names its flows; migration 0031 stores the tags on flow_definitions (#85).
  • Dashboard: screen E filters on tag, device model, OS version, runner, device, CI provider, commit, failure fingerprint and suite, as well as flow, branch, outcome and window. Every filter is in the address, so a saved view is a link (#85, with #41).
  • Dashboard: screen E shows a project trend over the filtered trusted runs: passed, failed and flaky counts per UTC day, and the median and 95th-percentile duration per day and over the window (#85, with #41).
  • Protocol: hello carries catalog (the workspace's flows, bounded at 500, with truncated) and each device sample in telemetry carries model and os_version; both are optional, and PROTOCOL_VERSION stays 1 (#85).
  • Server: GET /api/v1/projects/{project}/flows (the flow catalog) and GET /api/v1/projects/{project}/history/trend (the trend); the history takes the new filters (#85).
  • Server: one browser live socket carries several subscriptions. A subscribe may name its subscription with an id, and unsubscribe stops one; every frame names its subscription as sub. Each subscription is authorized on its own, and a refused one is answered alone while the socket stays open. A socket holds at most 32 subscriptions. A single subscribe without an id works as before. The protocol adds the unsubscribe request and the unsubscribed and samples frames, and the schemas are regenerated (#40).
  • Server: a committed telemetry checkpoint tells its project's subscribers that its runners' samples changed (samples frame), on every instance (#40).
  • Dashboard: one live socket serves the whole app. A page subscribes when it opens and unsubscribes when it closes, and a reconnect subscribes every page again, so moving between pages does not open another socket (#40).
  • Dashboard: the runner and device pages update live as new samples arrive and as their runs move on, without a reload. While a page is not receiving updates (reconnecting, or paused), its runner's liveness, its host sample, and its device's state and newest sample read Stale (#40).
  • Dashboard: a failed flow's page (screen D) and its row in the history (screen E) show the Maestro command that reruns that flow, selectable and copyable. The command is the run's own, with the flow's path last; env values and unrecognised flags are never shown, and the page names what was left out (#82).
  • Dashboard: an attempt's failed flows download as a plan manifest, from the flow page, the history rows of failed flows and the pipeline page. One shard per run with a failed flow; fleet-agent run --plan reruns exactly those flows. An attempt with no failed flow answers 404, and one whose failed flow has no path answers 409 (#82).
  • Server: GET /api/v1/projects/{project}/attempts/{attempt}/failed-flows serves that manifest as an attachment. The screen D and history rows carry reproduction (command, left_out, unavailable) (#82).
  • Agent: run.started carries reproduction, the words of the run's maestro test command with its env values as [REDACTED] and its unknown flags named only, built from the redacted argv. Migration 0029 stores them on agent_runs (#82).
  • Dashboard: Screen G, administration, read only (#139). A global administrator sees Administration in the navigation and reads /admin: projects, publisher trust policies, API keys (metadata only), role mappings, retention classes and windows, the artifact policy and the metric collection policy (both from the running configuration), agent versions against the protocol, ingestion health (refused events, reported losses and the lag of open runs), and /admin/audit, the audit history, newest first and 50 at a time. Everyone else is refused: the page's links are hidden and GET /api/v1/admin/… answers 403 (401 anonymously). The routes are in the authorization matrix. fleet_core::policy gains ViewAdministration, a global administrator's action; the audit history keeps ViewAudit. The operator guide's Administration section says where each area comes from. Forms to change these areas follow.
  • Dashboard: Screen G's changes (#139). A global administrator creates a project and renames it; creates an API key (its secret is shown once, on the page the form answers with), expires it or revokes it; gives a person a role on a project or on every project, and removes it; maps an identity-provider group to a role, changes it and removes it; and adds, edits, revokes and removes publisher trust policies. Each change is one transaction with its audit event (who, what, and the values before and after), and a refused change writes nothing. The forms post with the session's CSRF token, and a request from another site is refused. The same changes are on /api/v1/admin/… as JSON. Migration 0030 adds a source column to group_mappings and publisher_trust_policies (config or dashboard), lets a removed trust policy's tokens end with it, and indexes the audit history by time.
  • Dashboard: a configured trust policy or group mapping is locked on the page, and the start-up sync refuses a configured row that a page row already holds, naming it (#139).
  • Dashboard: Screen G's forms for retention, artifact policy and metric collection, per project (#139). A global administrator sets each retention class's window (0 keeps for ever; it applies to artifacts that become available afterwards), accepts or turns off each artifact class and its largest file, sets the daily bytes and files, and sets the calm and under-pressure sampling intervals (1 to 300 seconds). A blank field keeps the configured value, and an artifact override or daily quota can only be lower than the configured cap. Migration 0032 stores the overrides; the JSON routes are PUT /api/v1/admin/retention/{project}/{class}, …/artifact-classes/{project}/{class}, …/artifact-quota/{project} and …/metric-policy/{project}.
  • Dashboard: every change on these forms is audited in the transaction that makes it, with who made it and the values before and after. A refused change writes no event (#139).
  • Dashboard: a project created by the configuration (projects[]) is locked from renaming on the page: its row says from config, and the API refuses a rename with a reason that names the configuration. A project created on the page can still be renamed (#139).
  • Protocol: welcome.limits carries optional sample_interval_ms and pressure_sample_interval_ms: the project's own interval, else the server's ingest.sample_interval_ms and ingest.pressure_sample_interval_ms, else the agent's 5 s and 15 s. Both are omitted when unset, so PROTOCOL_VERSION stays 1 (#139).
  • Dashboard: a failure's First seen date shows beside its triage state on the flow page and in the history (#84).
  • Dashboard: every live page (overview, pipeline, flow, runner and device) has one polite status region that reads run events out in batches, at most once every five seconds: a run starting or ending, a flow failing or being interrupted, a pipeline's verdict changing. Log lines and counts are never read out (#149).
  • Agent: each emulator's qemu-system-* process is matched to its serial by its console port, and its CPU and resident memory are sent with the device's samples (emu_cpu_pct, emu_rss_bytes, #39).
    -...
Read more

0.1.0-alpha.3 - 2026-11-07

Pre-release

Choose a tag to compare

@github-actions github-actions released this 07 Nov 11:58
94dca31

Release Notes

Changed

  • Dashboard: the redesign's foundation: graphite and paper tokens, self-hosted IBM Plex, one Title Case status vocabulary ("Not started" replaces "Pending", a completed step reads "Passed"), monochrome provenance labels, a shared top bar with a live indicator on the pipeline page, and sections without cards.
  • Dashboard: the pipeline page's verdict is derived from its flows (Failed, Passed only with a plan total every flow accounts for, else No failures); readouts filter the flows, failures group by message, flows group by job and device with the plan's shard per device, and the activity shows only while the run is open.
  • Dashboard: the runner and device pages follow the redesign: a header with the name, a liveness pill and the age of the newest sample, one readout strip for host and Maestro CPU and memory, charts that state their window and resolution once, runs with verdicts from their exit status, and a live bar that shows the connection and can pause updates. Temperature is no longer charted, since no agent reports it (for a device too); the API still returns it as unavailable.
  • Dashboard: the Overview is redesigned. The readouts filter one Pipelines table whose verdicts are derived from the flows, "Needs attention" lists the last day's failed flows with their message, step, device and run, and the runners show their CPU and freshness.
  • Dashboard: history and compare follow the redesign: a filter band with underlined fields, one "Verified" for a table whose rows are all final, a stats panel that shows only the figures that say something, an answer first on compare ("Failures only in B") with a Swap button, and sections that state their provenance once.
  • Dashboard: the flow page is redesigned around the failure: steps are numbered as in the flow file (Maestro's setup commands fold into one unnumbered "Setup" row), a failure band names the failed step, its command and message (with a copy button), a history strip shows the flow's last executions, the verdict says why it is verified or provisional, and steps, evidence and logs sit in three panes.

Install fleet-agent 0.1.0-alpha.3

Install prebuilt binaries via shell script

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/moyaver/fleet/releases/download/v0.1.0-alpha.3/fleet-agent-installer.sh | sh

Download fleet-agent 0.1.0-alpha.3

File Platform Checksum
fleet-agent-aarch64-apple-darwin.tar.xz Apple Silicon macOS checksum
fleet-agent-aarch64-unknown-linux-musl.tar.xz ARM64 MUSL Linux checksum
fleet-agent-x86_64-unknown-linux-musl.tar.xz x64 MUSL Linux checksum

Verifying GitHub Artifact Attestations

The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:

gh attestation verify <file-path of downloaded artifact> --repo moyaver/fleet

You can also download the attestation from GitHub and verify against that directly:

gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo moyaver/fleet

0.1.0-alpha.2 - 2026-11-07

Pre-release

Choose a tag to compare

@github-actions github-actions released this 07 Nov 03:29
9fde28c

Release Notes

Added

  • Agent: a failed flow's device logcat is uploaded as an excerpt: the lines of the failing step's window (10 s before it to 2 s after it), redacted line by line, at most 512 KiB, keeping the newest lines and saying when older ones were cut. The window is placed on UTC by the log's own last line, so a log whose clock cannot be told from the failure's is not sent (#75).
  • Agent: an app that crashes during a flow is journaled as app.crashed (package and exception, from Maestro's crash-report.txt) on that flow. The flow's verdict stays Maestro's own (#75).
  • Server: the device_logcat artifact class, a text file with its own size limit (artifacts.max_bytes.device_logcat, 512 KiB) (#75).
  • Dashboard: screen D offers a failed flow's device logcat excerpt as a download, labelled "Device logcat excerpt" (#75).
  • Agent: at run end, GITHUB_STEP_SUMMARY gets a short Markdown summary: outcome counts, the flows that did not pass with their provenance and reason, and a dashboard link when FLEET_SERVER is set. A summary that cannot be written costs one notice; Maestro's output and exit status never change (#71).
  • Agent: the devices Maestro drives are sampled every 30 s with one read-only adb probe each: adb's state, whether sys.boot_completed is set, and the guest's CPU and available memory (guest, unknown when adb does not answer). The server stores them in device_samples and screen C serves them (#39).
  • Examples: a composite GitHub Action, examples/github-actions/fleet-agent, installs a pinned, checksum-verified agent release and runs a Maestro command under it (#71).
  • Dashboard: a flow's stats panel says how many of its failed runs a lost device excluded from flakiness (excluded on the flow's verdict; "2 runs excluded (device lost)"), and says nothing when there are none (#72).
  • Agent: a failing step's UI hierarchy (screen-hierarchy/step-NNN-….json) is uploaded as a failure_hierarchy artifact with its on-screen text values replaced by [REDACTED] (the tree and its attributes are kept); screen D lists it under its step, and the failing step's comes first (#74).
  • Agent: step events carry depth, parent and attempt where Maestro's log shows them (a step inside runFlow, retry or repeat, and the retry attempt that ran it). The flow page nests steps under their block and shows the attempt of a retried step, from GET /api/v1/projects/{project}/flows/{execution}, whose steps are now a tree with children (#38).
  • Server: per-IP limits on the pre-auth routes (agent token exchange, browser sign-in and callback) and per-credential limits on agent connections, answered with 429 Too Many Requests and Retry-After; an agent socket's events are paced, never dropped. Client IPs come from trusted_proxies (#35).
  • Server: viewer REST reads and live socket upgrades are limited per viewer credential (a viewer token or a browser session), answered with 429 Too Many Requests and Retry-After; the dashboard waits that long and keeps its last data instead of erroring (#35).

Changed

  • Dashboard: the pipeline's plan table reads a small plan endpoint (GET /api/v1/projects/{project}/attempts/{attempt}/plans) and refetches it only when a planned shard's runs change or the attempt ends, instead of the whole attempt snapshot on every live update; its own fetch error shows in the table (#115).
  • Server: the image takes wasm-opt, tailwindcss and wasm-bindgen from a cached build layer, pinned by checksum, so a source change no longer downloads them again and a stalled download resumes (deploy/docker).

Fixed

  • Agent: reconnects to a refusing server back off again. The jitter read the clock's low nanoseconds, which are always zero on macOS, so every delay was zero and the agent retried about every 1.5 ms (3,108 attempts in 5 s); jitter now comes from OS randomness, and the server's goaway reconnect jitter, which had the same bug, too (#90).
  • Agent: a failed write to journal.jsonl (disk full, I/O error) stops the journal like the cap does: one journal.truncated line names the error, the run's warning names the first missing seq, and fleet-agent inspect reports it. Before, the error was dropped silently (#112).
  • Dashboard: a flow's evidence keeps its state when its signed links renew (about every four minutes): the screenshot zoom and selection, and the log search, stay put instead of resetting, and the evidence card no longer rebuilds on every live update of the flow. A tooltip hides on Escape from anywhere while it is shown, not only from inside its trigger (#41).
  • Server: a run whose socket another instance holds shows online or stale from its persisted contact, as on the instance holding it, instead of always stale (#114).
  • Server: a worker's Maestro and agent figures are one series per agent run, raw and in the minute and hour aggregates, so runs that share a runner no longer merge or interleave; the runner's latest reading carries only the machine's figures. The dashboard's worker page draws one line per run (#42).
  • Agent: after Maestro exits, the agent keeps draining its output while the CI log is still taking it, and gives up on an output pump after 2 s with no reads and nothing held for the log, or 10 s after the exit at most, so a backed-up log no longer loses its tail and a grandchild that keeps writing cannot hold the exit (#124).
  • Server: the OpenAPI document describes a flow's nested step tree without recursing, so GET /api/v1/openapi.json no longer overflows the stack; StepView.children is an array of StepView references.

Install fleet-agent 0.1.0-alpha.2

Install prebuilt binaries via shell script

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/moyaver/fleet/releases/download/v0.1.0-alpha.2/fleet-agent-installer.sh | sh

Download fleet-agent 0.1.0-alpha.2

File Platform Checksum
fleet-agent-aarch64-apple-darwin.tar.xz Apple Silicon macOS checksum
fleet-agent-aarch64-unknown-linux-musl.tar.xz ARM64 MUSL Linux checksum
fleet-agent-x86_64-unknown-linux-musl.tar.xz x64 MUSL Linux checksum

Verifying GitHub Artifact Attestations

The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:

gh attestation verify <file-path of downloaded artifact> --repo moyaver/fleet

You can also download the attestation from GitHub and verify against that directly:

gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo moyaver/fleet

0.1.0-alpha.1 - 2026-11-06

Pre-release

Choose a tag to compare

@github-actions github-actions released this 06 Nov 09:30
9f89696

Release Notes

Added

  • Planner: the manifest records what an equal-count split of the same flows would have produced by the same estimates (equalCountEstimatedSeconds), so a plan can state its saving over that split (#81).
  • Dashboard: a pipeline with a Fleet plan shows each shard's planned and actual time (first flow start to last flow end) and the slowest shard's saving over an equal-count split (#81).
  • Server: with several instances on one database, a change committed through one instance reaches browsers on the others as a live delta (Postgres NOTIFY/LISTEN), not only after a resync (#34).
  • Agent: host samples carry the load average, CPU cores, swap and Linux pressure stall information (CPU, memory, I/O); screen C charts them and the worker page shows the core count, with "Not reported" for what an agent doesn't send (#42).
  • Fleet Planner (ADR-0013): fleet-planner plan splits a Maestro suite into CI shards. It finds flows the way Maestro does (config.yaml globs, tags, executionOrder), balances shards Longest Processing Time first from Fleet history (or an even split without it), writes an immutable manifest and a GitHub Actions matrix, and publishes the plan. fleet-agent run --plan runs one shard with the plan's workspace config, and refuses an empty shard or a command that would select flows itself.
  • The agent falls back to HTTPS batch transport after repeated blocked WebSocket upgrades, with server-authoritative resume and telemetry acknowledgments.
  • Generated protocol JSON Schemas, golden wire frames and a CI freshness check.
  • A pinned Maestro nightly emulator workflow and a captured absolute-path regression fixture.
  • Agent: cargo xtask bench-overhead measures the agent's overhead (I10) by running the fake-Maestro suite with and without it, reporting the duration delta, the agent's CPU time and its peak memory.
  • Agent: the Maestro version comes from the session log's own version line, and is recorded once per run as maestro.detected (no extra maestro --version process). The server keeps it on the run and in the run's condition fingerprint; a version outside Fleet's profile is warned about on stderr and recorded with profile_supported: false (#73).
  • M5 exit: screen D shows a failed flow's evidence from the usercontent origin: the failure screenshot in the screenshot viewer (tied to the step named in its file name), the Maestro log in the log viewer opened at the failing step of that flow (the step tree moves it; "Jump to the failing step" returns), command timings from commands.json, a link to the Maestro HTML report that opens sandboxed on the usercontent origin, downloads, refused uploads with their reasons, and honest empty states (nothing uploaded, no usercontent origin). The dashboard's CSP allows exactly the configured usercontent origin in img-src and connect-src. fleet-e2e proves acceptance criteria 7, 8 and 9 and invariant I7 in a real browser with the real agent. The step tree now sets aria-level, aria-posinset and aria-setsize (they were being written as attr:aria-level). User guide: "Investigating a failure".
  • M5: retention (issue #45). The leader runs telemetry rollups (one-minute and one-hour, idempotent, late-data aware; unavailable readings stay unavailable), drops expired partitions of raw samples, the journal and steps (raw samples only once rolled up), deletes expired rollups and audit events, expires artifacts by retention class (failure evidence is promoted when its flow finalizes as failed; legal hold keeps an artifact), and deletes blob objects only when no artifact refers to them and after a grace period; leftover temp files of crashed local uploads are swept (BlobStore::sweep_stale_uploads). Migration 0009; [retention] server settings; operator guide section. The metrics endpoints serve ranges over 24 hours from the rollups.
  • M5: screens E and F of the dashboard (History and Compare): GET /api/v1/projects/{project}/history (filters by flow, branch, outcome and time, bounded to 90 days, cursor pages, failure and condition fingerprints, flaky v1 verdicts naming the runs behind them), …/flows/{name}/stats (settled counts, p50/p90 duration of passes, the latest durations, failures grouped by fingerprint) and …/compare?a=&b= (flow-by-flow outcomes with explanations drawn from the sides' evidence only); the /history and /compare pages with URL-synced filters and attempt pickers; fleet-loadgen can set the commit, ref, architecture and failure pattern of a run.
  • M4 exit: the browser acceptance suite (fleet-e2e, m4_exit.rs) proves, in a real browser against the real server, that a developer can follow a full run without CI logs and that acceptance criteria 2, 4, 5, 6 and 14 pass (plus 12 for sign-in): the real fleet-agent wrapping the fake Maestro replays a captured failing run, which is followed from the overview to the failed flow page without a refresh; progress counts toward a Fleet plan's total and is marked uncertain without one; ten simultaneous runs stay correct, and the page stays responsive; signing in works end to end against a wiremock identity provider (PKCE, HttpOnly cookie, group refusal, sign-out). The user guide gains "Watching a run" (screens A–D, the provenance legend, uncertain progress, stale and disconnected).
  • M4: the dashboard: fleet-ui component kit, charts and swimlane timeline, log, screenshot and step views; screens A–D (overview, pipeline detail, runner and device detail, flow investigation) with live updates, resync after a server restart and stale and disconnected states; browser sign-in (OIDC code + PKCE, Postgres sessions, CSRF) and the authorize policy; REST reads for the four screens.
  • M5: failure fingerprints (flow identity + failing step + message with numbers, UUIDs, timestamps, durations, paths, serials and addresses removed, versioned v1:), condition fingerprints (device, Maestro major.minor, OS/arch, host class, CI provider; missing facts read unknown) and flaky v1 (confirmed: pass and fail on one commit and condition; suspected: flip rate ≥ 0.2 over the last 20 default-branch runs; inferred outcomes never count as passes) in fleet-core; migration 0007 and the ingest transaction store both fingerprints when a flow's outcome is final (ADR-0009).
  • M5: fleet-blob: the BlobStore contract (content-addressed per project; streaming put that verifies size and SHA-256, ranged get, head, idempotent delete, list for GC) with a local filesystem backend (atomic writes) and an S3-compatible one (object_store, path-style, custom endpoint; Garage is the reference), one contract suite run against both; fleet-blob::verify: size, SHA-256, magic bytes per artifact class and archive safety for zip and tar.zst (entry count, expansion limits, unsafe paths, links, devices, nested archives); [storage] server settings and the operator guide's storage section. No upload endpoints yet.
  • M3: fleet-loadgen runs simulated agents against a real server (--agents --devices --storm --duplicates), replacing mock-server; fleet-e2e proves the M3 exit criteria (50 agents × 4 devices, reconnect storm, server restart mid-run, fixture replay matching fleet-agent inspect, jobs and same-serial emulators kept apart); the Docker image (deploy/docker/Dockerfile, cargo-chef + distroless) and the Compose reference deployment (Postgres 18, Garage, Caddy); operator guide and protocol reference.
  • M3: POST /api/v1/agent/token exchanges a CI OIDC token (configured issuers only, RS256, JWKS cached and refetched on an unknown kid, aud/exp/nbf/iat checked, each jti once) for a short-lived agent token, matched by trust policies on immutable repository IDs (fleet-core::trust; pull_request_target and dynamic always refused), or exchanges an API key. Verified claims name the pipeline run and job. The agent does the exchange on GitHub Actions (#25), tunnels through HTTP(S)_PROXY with NO_PROXY and trusts FLEET_CA_FILE (#24). fleet-e2e runs the real agent against the real server.
  • M3: the live hub and /api/v1/live (snapshot, then deltas in version order per attempt, resync for slow subscribers, Origin check), REST snapshots with a consistent version (/api/v1/projects, …/attempts, …/attempts/{id}), /api/v1/openapi.json, liveness (online/stale/disconnected/completed) and leader-only sweeps that detach silent runs and abandon expired ones (open flows become unknown from server_inferred, jobs interrupted), and Fleet plans (POST /api/v1/plans, immutable, planned totals before any job starts; #21). Reads need a viewer token until browser SSO (M4). The plan manifest type moved to fleet-protocol::plan.
  • M3: the server ingests agents: /api/v1/agent/ws (fleet.v1+json: welcome with server-authoritative resume, ack after commit, event_rejected, seq_gap, fencing, pings, idle timeout, goaway + close 1012 on shutdown), the HTTPS batch fallback POST /api/v1/agent/batch, telemetry checkpoints with telemetry_ack, API-key auth, /livez and /readyz, and the serve, migrate, check-config, healthcheck and api-key create subcommands with figment configuration (TOML, FLEET_*, *_FILE).
  • M3: fleet-store with migrations 0001–0005 (Postgres 18) and the ingest transaction (epoch fencing, duplicate skip, ownership-checked upserts, claim-lattice outcomes, recorded gaps); fleet-core::outcome (suite and job outcomes); cargo xtask sqlx; CI runs the database tests against a Postgres 18 service.
  • Protocol: flow.ended may name its flow (flow), so a claim about a flow that never started (a skipped executionOrder flow) creates it.
  • M2: the fleet-agent CLI (run, inspect, flush, check, version): supervisor, live observer, spool, WebSocket transport, metrics, adb tracking, planner shards; fleet-loadgen mock-server; cargo-dist release pipeline.
  • M1: cargo xtask fixtures capture | import, 24 Maes...
Read more