* Deploys wait for the Mac's image; image-follow builds master's head; CI image bakes ffmpeg/graphviz; dashboard edits refuse missing dashboards, tabs and widgets
- release.sh: the deploy's image step waits up to wait_minutes for the image, and stops early once master moves on
- make image-follow: builds only master's newest head, cancelling a superseded build
- Dockerfile.deps bakes ffmpeg/graphviz; ci.sh's fallback install is bounded
- query-preview test polls for CodeMirror's time-sliced highlighting
- EditRefusal: DashboardGone/TabGone/WidgetGone join EditReadOnly in the route types
- CLI: disambiguate EventsSearchOpts record updates
* log-list test: poll for the load-newer reveal scroll, which lands on a later frame under load
* image-follow: an image already built for the merged tree counts
* hpack: list migration 0278 in monoscope.cabal
* Workspaces: one build per base, restore from object storage on a miss, builds report how the checkpoint arrived
* Workspaces: restore-then-build reads as one case
* Workspaces: number the run base migration after master's 0275
* Typed follow-ups: GitConnError, /api/v1 problem envelope declared per route, typed GET fragments
Pkg.Git local validation returns GitConnError instead of Text. /api/v1 refusals and
Servant's own rejections answer one {"error": ApiProblem} envelope, declared per
route via Refuses for OpenAPI/MCP; the CLI prints its message. Remaining
RespHeaders (Html ()) fragments (Telemetry, Issues, CodeContext, Dashboards,
Settings, LogExplorer) return typed view values.
* Review fixes: closed ProblemCode shared with the CLI, split git origin errors, MCP shows refusal messages
ApiProblem moves to monoscope-shared with a closed ProblemCode enum; ApiError
owns every code. Servant rejections no longer guess a field from message text.
Refuses takes one status per combinator. Origin/endpoint/connection errors are
separate types; snippets report ConnectionInvalid. Metric fragments share one
record; the code-mappings picker route returns the fragment it swaps.
- shared/src/Web/Operations.hs declares every /api/v1 operation once: MCP tool
name, agent access, description, and its CLI command, arguments and listing
columns. AgentCatalog.apiOperations and the CLI's per-resource command types,
parsers and dispatch now derive from it; a generic parser and runCall replace
them, and ResourceKind is gone. Typed Method/Verb, Input, Arg and Listing keep
invalid argument shapes unrepresentable.
- Agent local tools: single-use KQL/tool helpers folded into executeLocal; tool
names in errors come from the operation. get_code_context no longer builds a
throwaway cache (fetchSourceEvidence takes Maybe Cache).
- Derived names: constructorName (Generic) for actionName/operationName,
slackReadName via encodeEnumSC (SearchPersonal -> SearchPublicAndPrivate),
StreamFrame ToJSON via deriving-aeson.
- CLI charts on granite (round ticks, exact x range, width-exact) and tables on
table-layout (widest column absorbs overflow, single-line header rule).
- retry replaces hand-rolled backoff in retryOnDeadlock, retryOnPoolExhaustion
and the CLI's withRetry; agoText replaces prettyTimeShort; http-types urlEncode
in the CLI; text-metrics levenshtein replaces editDistance.
Net about -970 lines.
* CLI: write stdout/stderr as UTF-8 so help survives a C locale
Under LC_ALL=C (minimal Linux images) GHC's handles are ASCII, so help
pages with — or → printed the usage line and exited 1. The binary help
sweep now runs under LC_ALL=C and asserts the non-ASCII lines.
* CLI: metrics query --assert fails when the response does not decode
A response that isn't a metrics result skipped the assertion and exited 0,
so a CI gate built on --assert passed silently. It now exits 1.
* CLI specs: one spec of self-contained lifecycles
Merge CLIBinarySpec, CLIE2ESpec and CLILifecycleSpec into CLI.CLISpec: help
usage, error and list envelopes as tables, binary-vs-live-server walk, one
in-process lifecycle per workflow. withCliBin moves to CLI.Harness. The fake
router honours issues list filters and drops the /log_explorer and
/chart_data arms the CLI no longer calls.
* Log-pattern bulk actions accept ack like issues do
* CLI spec: one binary runner, per-example temp dirs for as-code yaml
* Drop dead telemetrySchemaJson; give the CLI binary spec a UTF-8 locale
The stripped test env dropped CI's LANG=C.UTF-8, so GHC's stdout fell back to
ASCII and help text containing — or → exited 1.
Slack, web conversations, routines and native review/fix jobs share the capability catalog, current authority checks, run ledger and operation receipts. Runs, charts and saved reports open inside the conversation inspector.
Consolidates duplicate regression suites and documentation while retaining the product contracts, native recovery evidence and migration baseline transition.
Validation: all eleven exact-input local checks pass, including 1,216 integration examples (25 pending VM cases), 2,105 doctests, 315 unit examples, 16 CLI examples, 1,041 UI tests and 172 browser tests (one skip). Separate native/KVM acceptance is recorded in the review. The exact Linux AMD64 image was built locally on this Mac with OrbStack/Rosetta.
* Fix dashboard variable loading, CLI test events, and onboarding checks
- Dashboards: charts init from a data-chart-config attribute instead of
inline scripts; resolved constants travel with the swapped tab panel;
variable picker filters via hx-live and syncs tagify without refetching.
- CLI: send-event/telemetrygen exchange a login session for the project's
ingestion key (Ingestion effect, Credential sum type) and exit nonzero
unless the span was exported.
- API: auth yields a typed ApiPrincipal (Authenticated | DemoReader);
/api/v1/ingestion-key refuses the demo bypass.
- KQL isempty/isnotempty compare as text so numeric args don't coerce ''.
- Onboarding: no toast on automatic integration polls; CLI quick test
sends one event; docs panel copy menu uses a native popover.
* Update page and e2e assertions for chart config attributes and lazy docs
* Auto-format code with fourmolu
---------
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
CI ran `hlint src/`, so monoscope-shared and monoscope-cli — 22 modules, including
the KQL parser and the binary we ship to users — were never linted. shared/src was
already clean; the CLI had eight issues, one of them error severity.
All mechanical, all Relude preferences the rest of the codebase already follows:
T.pack over a String where toText does, T.unwords/T.unlines where Relude's are in
scope, the Data.Set alias the hlint config pins, and a lambda that is a <=< .
cabal test monoscope-cli:cli-tests passes 16/16 and the binary links.
`monoscope logs search POISON_ROW_DROPPED` — the example in the command's own
--help and in docs/cli.md — has been exiting with
invalid query: Unknown field "POISON_ROW_DROPPED"
since field validation landed in e7ad369. Nothing was wrong with the search
itself: `buildSearchParams` rewrites a bare term into `body has "X" or summary
has "X"` before it goes on the wire, and that is valid. But `validateEventsOpts`
validated `opts.query` — the raw term, before the rewrite — so the parser was
asked to resolve a search term as a column and the CLI exited before the rewrite
ever ran.
Validate the text that will actually be sent. Confirmed in ghci against this
tree: parseQueryToAST "chunkevt" is Left, and the rewritten form is Right.
Server-side is unaffected: the API only ever received the rewritten query, so
this was never visible through the web UI or a direct API call.
The two halves drifting apart is the actual defect, so the doctests on
rewriteBareQuery now assert that its output parses, not just what it looks like.
"Acknowledged" promised nothing. It moved a row out of the Inbox and kept
sending alerts, so the only way to actually stop being paged was to archive —
which also meant "this isn't worth fixing". The word had no meaning you could
act on, and the Acknowledged tab was where issues went to be forgotten.
How the others resolve it: Sentry's Archive takes an explicit "until" (a
duration, or until it happens again N times) and stops notifying for it.
Datadog keeps monitor muting time-boxed and separate from status. PagerDuty's
ack is "I'm on it" with a timeout that re-pages if nothing happens.
The common thread is that silence is a promise with an expiry. So rather than
adding a second verb next to acknowledge, acknowledge *becomes* the silence
control — one concept, with a window:
Open needs triage notifies
Acknowledged someone owns it, until <when> silent, then back to Inbox
Archived not actionable silent, stays hidden
- apis.issues gains acknowledged_until (0125). A far-future sentinel — the same
one monitors use for an indefinite mute — encodes "until it regresses".
It supersedes cooldown_until, which said the same thing but only for
log_pattern_rate_change and only for a fixed 24h.
- The notification paths honour it. claimIssueNotification, the new-endpoint
claim, and the error-subscription sweep all now skip acknowledged and
archived issues; a lapsed ack stops blocking immediately rather than waiting
on the sweep. This is the actual bug: none of them looked at ack state before.
- Detectors consult isSilenced before firing a fresh issue for a target, so an
ack means "don't tell me about this again", not "hide the old row".
- expireLapsedAcks (in the 10-min notification sweep) returns timed acks to the
Inbox with an ack_expired activity event, so one can't rot out of sight.
- UI: the acknowledge control is a split button — one click silences until it
regresses, the caret offers 1h/4h/8h/1d/3d/1w. Rows and the detail header
show the time left; each tab says in one line what it means for notifications.
Bulk actions now offer only the transitions that make sense from the tab
you're on. Ack and archive were <a> without href: not keyboard-operable.
They are buttons now, and the icon-only row actions have accessible names.
- API/CLI: POST /issues/:id/ack?duration_minutes=, issues bulk duration_minutes,
monoscope issues ack --for MINUTES. Absent duration = indefinite, which is
what ack did before, so existing callers keep their semantics.
The duration menu and its "4h left" label are shared with monitor muting, so
"for how long?" is answered the same way in both places.
Found by actually running the commands end-to-end:
- The waterfall keyed the tree on the row UUID, but parent_id holds a *span*
id (the server aliases context___span_id to latency_breakdown). Every span
therefore looked like a root and nothing was ever indented. EventRow gains
spanId and the tree keys on it. The test fixture had encoded the same
misunderstanding, so it passed throughout — it now carries both ids with the
real linkage.
- logfmt emitted the summary's ANSI colour codes and did not escape quotes or
newlines, so a log body containing either split one event across lines. The
summary is stripped of styling and values are escaped.
- events get --tree hardcoded a 90d scan, which the store can refuse outright
("merge-on-read dedup exceeded its per-query limit"). --since narrows it;
the default is unchanged.
Removes three forks of the same logic and turns on the tests that would have
caught them:
- CLI.Chart.seriesFrom generalises seriesFromMetrics over any header/grid pair,
so CLI.Dashboard.widgetSeries and its private isTimeCol go away. The two had
already drifted (one recognised _timestamp, the other didn't). CLI.Commands's
third copy of the time-header list goes with them.
- Pkg.CLIFormat gains extractColIdxMap and renderSummaryCell; CLI.LogView and
CLI.Commands each had a byte-identical private copy of both.
- LogView measured ANSI width with T.filter (/= '\ESC'), which leaves the
'[2m' behind and under-sized the message column whenever colour was on. It
now uses CLI.Chart.visibleWidth, which the module already imported.
Dead weight: ChartStyle was never set to anything but Line, so the type, the
field and the unreachable canvas fill are gone; clamp/clampD collapse into
Data.Ord.clamp; two zipWith/foldl' plumbings threaded an index nobody read.
apiDashboardData stops returning (widget, window) tuples — the window comes
from the request, which also means a dashboard with no widgets reports its
range instead of null.
The doctest runner now covers cli/CLI (768 -> 820 examples). Turning it on
found two of my own examples asserting things that were never true, one that
would not compile, and two list literals that were ambiguous under the
runner's OverloadedLists. Four hspec cases duplicating doctests are deleted.
`errors == true` fails to plan on TimeFusion (the column is Utf8View, so the
boolean comparison has no common type). status_code=="ERROR" is the filter
that actually works against both stores.
optparse's plain footer reflows its string, so every command's Examples block
collapsed into an unreadable wall of run-on commands that could not be
copy-pasted. examplesFooter lays the lines out explicitly.
Also folds the CLI's two copies of "how wide is the terminal" into
CLI.Table.termWidth, which already guards the stdin probe.
Two things a terminal-first user reaches for that were missing:
- `monoscope status` — one screen answering "is anything wrong right now":
throughput, errors, busiest services, open issues, alerting monitors. A panel
whose query fails says so inline; an unavailable panel and a quiet one are
not the same thing.
- `monoscope open <target> [id]` — the web UI link for a trace, issue,
dashboard, log query or the project, opened or just printed. The host comes
from /api/v1/me (new `host_url` field) rather than being guessed from the API
URL, so self-hosted and split API/UI hostnames both work.
Writing status surfaced three real bugs in `summarize <agg> by <field>` with no
time bin — the shape behind `chart --type bar` and the
`summarize count() by resource.service.name` example in our own docs. It failed
for every caller:
1. The grouping column was never projected, so the query returned a bare column
of numbers with nothing saying which group each belonged to, and no decoder
accepted it. It now emits (group, aggregate), matching the binned branch.
2. /api/v1/metrics defaulted to the DTMetric decoder, which needs a leading
timestamp. Charts.decoderFor now infers the decoder from the query shape:
no `by` → scalar, `by bin(...)` → timeseries, anything else → label/value.
3. `| sort by count_ desc` was silently dropped (the branch hand-rolled its
ORDER BY), and once aggregates were cast to text for the decoder it would
have sorted them lexicographically. ORDER BY now resolves aggregation
aliases back to their expressions, so it sorts numerically.
Also: test-dev and the unit/integration suites needed the new packages on their
dependency lists.
docs/cli.md gains sections for `chart`, `dashboards render`, the trace
waterfall, --format, and the dashboard-render JSON envelope; the install
section now states the no-prerequisites guarantee and what enforces it.
monoscope-skills learns the same commands, with the agent-facing framing:
render a dashboard to find broken or empty widgets, sort a trace's spans by
duration rather than reading the waterfall, and bound `tail` with `timeout`
so a tool call cannot hang.
Two bugs found while writing the examples:
- chart ignored --json and always drew; it now honours the output mode.
- a failed query came back as an empty result set with `error` set, which the
CLI rendered as "no data in range" and exited 0. It now prints the server's
message and exits non-zero in every mode.
The CLI no longer needs the frontend build, proto codegen, or the eight
Homebrew packages the macOS job used to install — none of that is reachable
from monoscope-cli. Removing openapi3 from the shared package (a CLI has no use
for schema generation; the ToSchema instances move to Pkg.DeriveUtils and the
new Web.WireSchemas) takes the stripped -O2 binary from 107MB to 97MB.
- cli-release.yml: builds monoscope-cli:exe:monoscope only, with -O2
-split-sections plus --gc-sections/-dead_strip and strip.
- scripts/check-cli-linkage.sh fails the build if the binary links libpq,
librdkafka, grpc, protobuf or a compression library — the no-prerequisites
promise is now enforced rather than assumed. It also rejects /opt/homebrew
paths, and the macOS job repoints Homebrew's zlib at /usr/lib and re-signs,
which would otherwise have shipped a binary that only runs on a Mac with brew.
- cli-tests now run in the main CI workflow; make cli-build/cli-test/
cli-release-build for local use.
Measured but not taken: dropping brick/vty saves only 4MB, which does not pay
for losing the arrow-key project picker at login.
- One event per line (time, severity, service, duration, message, trace id)
instead of a bordered table, which is what you actually scan during an
incident. --format line|table|logfmt picks the shape; --json/--yaml still win.
- events get --tree draws a waterfall: spans indented by depth with a bar
showing where each sat inside the trace, so "what took the time" is visible.
Spans whose parent is missing render as roots rather than disappearing.
- tail gets --format, --interval and --since (backfill), and now overlaps its
poll window so a row written a second late is not skipped forever. --grep
matches the message, span name and service rather than the raw joined row.
- New cli-tests suite covers the renderers as pure [Text]: widths, empty and
flat series, ANSI-safe padding, waterfall shape, dashboard band layout.
Adds GET /api/v1/dashboards/{id}/data: the server resolves constants,
variables and every widget's query up front and returns the 12-column grid,
so a non-browser client draws the dashboard without re-implementing any of it
or firing a request per widget.
CLI side:
- CLI.Chart draws braille line charts, block bar charts, sparklines and stat
tiles as pure [Text], with nice y-ticks, time-formatted x labels and a legend.
- CLI.Dashboard maps the grid onto the terminal width, so widgets that sit side
by side in the browser sit side by side here; group widgets nest a sub-grid.
- monoscope dashboards render <id> [--tab] [--widget] [--var k=v] [--watch]
- monoscope chart '<kql>' [--type line|bar|stat|table] [--watch]
- metrics chart now goes through the same renderer instead of a bare sparkline.
Widget.chartQuery moves the browser's default aggregation server-side, and
getTerminalSize is no longer called when stdout is not a terminal (it reads
stdin, which threw EOF under a pipe).
The monoscope binary linked lib:monoscope, which drags in libpq, librdkafka,
grpc, protobuf, snappy, lz4 and zstd. Anyone installing just the CLI had to
install all of them first.
- New monoscope-shared package holds the DB-free modules the server and CLI
genuinely share (wire types, KQL grammar, HTTP effect, formatting).
- New monoscope-cli package owns CLI.*; it depends on monoscope-shared only.
- Pkg.Deriving splits the pure DerivingVia wrappers out of Pkg.DeriveUtils
(which re-exports them, so no call site changes).
- Web.Wire homes Paged / device-flow / FacetValue away from their DB-bound
modules.
- runCLILifecycle moves into the integration suite, which can see both packages.
The produced binary now links only libSystem, zlib, ncurses, iconv and ffi.
* feat(cli): monitors/dashboards-as-code + in-process CLI lifecycle tests
- Move cli/ into the library (CLI.Main.cliMain); exe is a thin Paths wrapper.
Enables runCLILifecycle: real optparse parse + command pipeline in-process,
HTTP routed to handlers via runHTTPtoServant, stdout captured (silently).
- monitors apply was not an upsert: an id-less file created a new row on
every apply. Add POST /api/v1/monitors/apply (natural key: alert title,
mirroring dashboards' file_path), wire CLI applyOne to it.
- Add monitors yaml ID dump (inverse of monitorFromInput) so dumps round-trip
through apply; same contract dashboards already had.
- Test router: thread verb + query params through writes (PUT/PATCH/POST were
indistinguishable, mute --for lost its duration); add monitors, dashboards,
events, facets, metrics, share routes + DELETE routing.
- CLILifecycleSpec: monitors-as-code and dashboards-as-code lifecycle walks
(apply idempotency, yaml round-trip, mute/unmute/star/duplicate/delete).
* feat(cli): single-envelope chunked search, OTLP/HTTP ingestion, scalar metrics fix + lifecycle tests
- Chunked search no longer leaks slices as NDJSON: slices fetch newest-first
with the remaining --limit budget, rows dedupe by id at boundaries, and the
result renders through the same single-envelope path as unchunked searches.
has_more/cursor come from the stopping slice. --no-chunk removed;
--chunk-hours 0 disables slicing.
- OTLP/HTTP ingestion (POST /v1/traces, /v1/logs, application/x-protobuf)
reusing the gRPC processing pipeline with x-api-key header auth. The
hs-opentelemetry exporter is http/protobuf-only, so send-event/telemetrygen
had no working collector-less path; configureOtelEnv now also honors a
pre-set OTEL_EXPORTER_OTLP_ENDPOINT. Handlers injected into Routes.server
from System.Server (proto-lens orphan IsLabel instances clash with
generic-lens labels if Routes imports OtlpServer).
- Scalar metrics queries (summarize with no by-clause) crashed on float8/Int
decode under the default DTMetric pivot; queryMetrics now infers DTFloat —
fixes the documented 'metrics query ... --assert' CI-gate flow.
- Lifecycle tests: investigation pipeline (search → get → context →
share-link resolving unauthenticated via shareLinkGetH), chunked
single-envelope contract (ordering, dedupe, --limit/has_more), metrics
--assert exit codes, and send-event shipping a span over the real OTLP/HTTP
wire into a live Warp instance (withLiveServer).
* Auto-format code with fourmolu
* lint: bring src/CLI under hlint zero-hint policy
Moving cli/ into src/ exposed it to 'hlint src' for the first time.
toText over T.pack, putTextLn, pass, universe, fromLeft, whenNothingM_,
Relude decodeUtf8 (already lenient), S alias for Data.Set, eta reductions.
* Auto-format code with fourmolu
* refactor(cli): address PR #428 review feedback
- warn (printDebug) + document when chunked-search dedup falls back to
full-row JSON keys because the response lacks an id column
- note why DashUnstar can't use runLifecycle (server route is DELETE
/star, no POST unstar endpoint)
- convert IssueList/EndList/MemberAdd positional Maybe fields to records
* fix(make): TEST_MATCH was silently matching nothing
GHCi :main does not strip mid-word quotes, so --match="PATTERN" passed
literal quote chars to hspec and every filtered run reported
'0 examples, 0 failures'. Pass --match PATTERN unquoted; for patterns
with spaces put the quotes in the variable: TEST_MATCH='"CLI lifecycle"'.
---------
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
- cli: AuthCmd never called resolveMode, so isJsonOutput stayed False and
the C7 non-interactive guard never refused the device flow when piped
- LogSpec: processMessages now returns Either WriteFailure (acks, poison);
length on the Either was always 1
- ApiV1Spec: get_schema serves Schema.deriveSchema over the live column
set, not the static telemetrySchema
Neutralises descriptive references to another vendor in comments and
user-facing copy. Functional behaviour is unchanged.
- cli/CLI/Table.hs, cli/CLI/Core.hs, cli/CLI/Commands.hs: comment polish
around the table renderer + bareword search.
- src/Web/MCP.hs: trim vendor list in tool-naming convention comments.
- src/Pages/Anomalies.hs: rephrase breadcrumb-source comments. The wire
attribute keys themselves (sentry.breadcrumb.category/type/message)
are SDK-defined identifiers and stay — those are what we parse.
- src/Pkg/ErrorFingerprint.hs: drop the analogy in normaliser comments.
- src/Pages/Onboarding.hs: rewrite the FAQ answer (also fixes two typos
and an awkward sentence).
- docs/mcp.md: simplify the verb-first naming-convention sentence.
CLI
- Mode model: drop --agent and --output FORMAT; flags are now --json | --yaml | --table, plus auto-detect (TTY → table, pipe → JSON). MONOSCOPE_FORCE_COLOR overrides.
- Pretty tables: new Sentry-style Unicode box-drawing renderer (cli/CLI/Table.hs) with per-cell CellStyle/Severity colouring, terminal-width truncation, and a pagination cue under list output. Listings (issues, monitors, dashboards, api-keys, teams, members, endpoints, log-patterns) flow through it via runListVia.
- Bare-string queries → full-text: foldFiltersIntoQuery rewrites a bareword/phrase to `body has "X" or summary has "X"` so `monoscope logs search POISON_ROW_DROPPED` works instead of 42703.
- --service repeatable: 0/1/N values expand to noop / == / `in (...)`; parser already supports `in`.
- --level normalises to upper-case.
- Auto-chunk long --since: events|logs search splits windows > --chunk-hours (default 1h) into slices anchored on a single now, streams NDJSON per slice, short-circuits on --first/--id-only, --no-chunk opts out.
- events get --field FIELD / --show-body: project subset of the event JSON without forcing jq.
- --limit alias added across listings (alongside --per-page / --top).
- Friendly gateway errors: 502/503/504 + Cloudflare HTML are collapsed to one-liners with a "narrow --since" hint; one jittered retry on transient 5xx + ResponseTimeout.
- Strict stderr in non-table mode: printError emits NDJSON {"error": {...}} on stderr; ANSI suppressed everywhere when stdout isn't a TTY.
- --debug sets X-Debug:1 header so the server can include raw SQL/Hasql exception text in error.details.
- renderAPIError decodes structured server errors and shows just the one-liner.
Server
- Schema introspection (C1): startup queries information_schema.columns for otel_logs_and_spans, derives the dotted-form attribute set, seeds Pkg.Parser.Expr.flattenedOtelColumnsRef (via setFlattenedOtelColumns). flattenedOtelAttributes reads from the cache; falls back to flattenedOtelAttributesBuiltin if introspection fails.
- /api/v1/schema reads from the live cache via Models.Telemetry.Schema.deriveSchema, which decorates each live column with the hand-coded FieldInfo when present and drops dotted-form Schema.hs entries that don't match a real column.
- Facets fallback (C2, new Web.FacetsFallback module): when apiFacets gets an empty result for a requested field, run a bounded SELECT col, COUNT(*) FROM otel_logs_and_spans … GROUP BY col LIMIT 50 on the whitelisted column. Tags response with "source": "fallback". Module is split out so Hasql.DynamicStatements imports don't perturb Generic derivation in Web.ApiHandlers.
- Structured KQL errors (A2): queryEvents returns JSON-shaped 400s with {error: {code, message, field?, suggestion?, details?}}. Detects column-not-exist for both Postgres (column "X" does not exist) and TimeFusion (Schema error: No field named X).
- Severity field aliases: severity.text, severity.severity_text, level all canonicalise to severity.severity_text.
* ingestion: TF-strict commit, per-partition offsets, row-level bisection
The 2026-06-01 TimescaleDB outage exposed several ways the consumer
advances Kafka offsets past failed batches, losing hours of customer
data. Commit semantics, the dual-write asymmetry, and the fire-and-forget
DLQ all combined to step over any failure that wasn't a hard exception.
- kafkaService now commits per (topic, partition) via
commitPartitionsOffsets. Only groups whose batch succeeded -- or whose
poison content was successfully DLQ'd -- contribute offsets. Failed
groups leave their partitions uncommitted; lag growth becomes the
outage signal. DLQ is no longer subscribed by the live loop.
- pubsubService mirrors: poison -> DLQ-then-ack; transient -> no ack.
- publishToDeadLetterQueue returns Either Text () so the caller can gate
commit on DLQ success.
- bulkInsertOtelLogsAndSpans bisects on non-transient Hasql failures,
isolating a single poison row in O(log N) attempts so a bad row no
longer wedges the partition.
- bulkInsertOtelLogsAndSpansTF requires TF success to return normally;
PG-only failure is logged via PG_WRITE_FAILED_TF_OK and tolerated (PG
is being phased out).
- New Pkg/IngestError classifier replaces the brittle string-match
isUnrecoverableError. Unknown errors default to PoisonBatch
(DLQ-actionable) rather than TransientInfra (silent stall).
* Auto-format code with fourmolu
* ingestion: inline transient-error check, tighten review nits
PR-review polish on top of the durability changes:
- Inline the Hasql-transient check at the two failure-handler call sites
in Pkg/Queue.hs and delete Pkg/IngestError.hs. Two 1-line predicates
with two call sites in one file wasn't worth its own module.
- Drop the now-trivial allTopics = kafkaTopics alias; inline into
consumerSub and drop the redundant all_topics startup-log field.
- publishToDeadLetterQueue: collapse the case lefts results to
maybe (Right ()) Left . listToMaybe . lefts.
- Lower the bisection depth cap from 12 to 10 (still 1024 rows, well
above chunkSize=700). Log BISECT_DEPTH_EXHAUSTED before re-raising
so the theoretical stall is visible if it ever happens in practice.
- Split bulkInsertOtelLogsAndSpansTF's (_, Left eTf) wildcard into
explicit (Right _, Left eTf) and (Left ePg, Left eTf) arms so a
both-failed event logs both errors instead of swallowing the PG side.
- Replace the CLAUDE.md cross-reference in the dual-write doc with an
inline note; CLAUDE.md is not in the repo and would rot.
* Auto-format code with fourmolu
* ingestion: extract isTransientException, drop commitTPs IORef
Second round of review polish on the durability changes:
- Data/Effectful/Hasql: export `isTransientException :: SomeException
-> Bool`. Collapses the four `maybe False isTransientHasqlError
(fromException e)` repetitions across Queue.hs and Telemetry.hs to
`Hasql.isTransientException e`.
- Pkg/Queue.hs:kafkaService: drop the `commitTPs` IORef +
`recordSuccess` writer; restructure the topic-group loop as
`fmap concat $ forM ...` that returns the per-batch TopicPartition
list directly. `tpsFor` becomes a small pure helper. Removes
`newIORef`/`modifyIORef'`/`readIORef`/`liftIO` ceremony.
- Pkg/Queue.hs:publishToDeadLetterQueue: replace `listToMaybe . lefts`
with `partitionEithers + listToMaybe`, matching the existing idiom
used at the `pollMessageBatch` site in the same file.
- Pkg/Queue.hs: add an inline comment by the (now-absent) poll
threadDelay site noting why no backoff lives there, so future
readers don't reintroduce one as a "fix".
* ingestion: concatMapM, lefts, $>, trim doc
Idiomatic touch-ups on top of the durability changes:
- Pkg/Queue.hs:kafkaService: replace `fmap concat $ forM` with
`concatMapM` from Control.Monad.Extra. Lift the per-group action
to a `let processGroup = ...` binding so the layout closes cleanly.
- Pkg/Queue.hs:publishToDeadLetterQueue: drop the partitionEithers
pair-with-ignored-second; use `lefts` directly.
- Models/Telemetry/Telemetry.hs:bulkInsertOtelLogsAndSpans: switch
the POISON_ROW_DROPPED branch from `0 <$ action` to `action $> 0`
so it reads left-to-right.
- Data/Effectful/Hasql.hs: trim the isTransientException doc comment
to a single WHY line.
* Auto-format code with fourmolu
* ingestion: hoist snippet build, document edge cases, guard sentinels
Address three correctness/observability issues and two conciseness
nits surfaced in review:
- Models/Telemetry/Telemetry.hs:bulkInsertOtelLogsAndSpans: hoist
otelRowSnippet above the bisection recursion so each row's snippet
is built (and any OTEL_UNPARSEABLE_IS_REMOTE log fires) exactly once
even when bisection recurses O(log N) levels. Bisection now operates
on a vector of (record, snippet) pairs.
- Models/Telemetry/Telemetry.hs:bulkInsertOtelLogsAndSpans: docstring
notes the non-TF-mode duplicate-row edge case when a mid-bisection
transient propagates and Kafka redelivery re-mints UUIDs. The
hypertable has no UUID-only unique constraint we can ON CONFLICT
against; under TF-strict semantics PG failures never trigger retry,
so the bug only surfaces in tests / legacy.
- Pkg/Queue.hs:publishToDeadLetterQueue: collapse the explicit
results-binding + lefts/listToMaybe chain to
`fmap sequence_ $ forM ...`; same first-Left semantics, less syntax.
Docstring notes the duplicate-in-DLQ-on-partial-failure tradeoff.
- Pkg/Queue.hs:kafkaService:tpsFor: filter out negative offsets in
the partition-commit map so a librdkafka sentinel (OffsetEnd = -1,
OffsetInvalid) can't rewind the consumer to the start. Shouldn't
appear in pollMessageBatch output but making the assumption explicit.
* ingestion: clarify DLQ aggregation, drop unreachable guard, comment <*>
Review polish, fourth pass:
- Pkg/Queue.hs:publishToDeadLetterQueue: switch back to explicit
`results <- forM ...; pure (sequence_ results)` instead of
`fmap sequence_ $ forM ...`. Same semantics, reads left-to-right
and matches the explicit-bind idiom used elsewhere in this file.
- Models/Telemetry/Telemetry.hs:bulkInsertOtelLogsAndSpans: drop the
unreachable `go _ pairs | V.null pairs` guard. The outer guards
V.null records; the single-row branch handles 1; recursive calls
always split a >=2 vector — so 'go' never sees an empty vector. A
comment notes the reasoning.
- Models/Telemetry/Telemetry.hs:bulkInsertOtelLogsAndSpans: comment
the sequential `<*>` semantics so a future refactor doesn't quietly
introduce parallelism that would let the right half attempt rows
after the left half throws (changing the transient + depth-exhaust
semantics).
Note: HM.fromListWith was considered for tpsFor to match the rest of
Queue.hs, but Kafka.Types.PartitionId has Ord but not Hashable, so
Data.Map.Strict stays.
Tracking deferred tests in #403.
* ingestion: trim docstring, where-bind rawInsert chunks, uncurry tpsFor
Three review nits:
- bulkInsertOtelLogsAndSpans: trim the ~25-line docstring to a 5-line
summary. The non-TF-mode edge case and the snippet-hoisting rationale
are PR-discussion material, not per-call doc.
- bulkInsertOtelLogsAndSpans:rawInsert: move chunkSize / chunks /
chunkStmt from `let ... in case` to a `where` clause so the body
reads case-first.
- kafkaService:tpsFor: deduplicate the per-partition K.TopicName topic
by mapping `uncurry (K.TopicPartition (K.TopicName topic)) . second
K.PartitionOffset` over Map.toList. Same result, no constructor
repetition.
Skipped: HM.fromListWith for tpsFor — Kafka.Types.PartitionId has Ord
but not Hashable (confirmed via :info), so Data.Map.Strict stays.
* ingestion: whenLeft_ tfThread, name firstFailure, guard bisect cap
- bulkInsertOtelLogsAndSpansTF tfThread: switch the Left/Right case
to whenLeft_ \e -> do ... (already 8+ call sites in the codebase).
Drops two lines and matches local idiom.
- publishToDeadLetterQueue: name the aggregated result 'firstFailure'
with an inline comment so the sequence_ short-circuit semantics
(first Left wins, subsequent failures only in the dlq log above)
is explicit at the call site.
- bulkInsertOtelLogsAndSpans: relax docstring to tie the bisect cap
(10 levels → 1024-row ceiling) to messagesPerPubsubPullBatch /
kafkaBatchSize rather than chunkSize. Add a runtime
BISECT_CAP_MAY_NOT_ISOLATE log on inputs above 1024 so on-call sees
the bound being approached before BISECT_DEPTH_EXHAUSTED fires on
a real poison row.
* ingestion: simplify firstFailure / tpsFor / processGroup
- publishToDeadLetterQueue: drop the intermediate `firstFailure` binding;
inline as `pure $ sequence_ results`. The short-circuit comment now
sits above the pure.
- kafkaService: drop `concatMapM` import (used exactly once); use
`concat <$> mapM processGroup ...`. Merge the two adjacent let blocks
inside the do (byTopic + processGroup) and lift `tpsFor` to the outer
`where` clause — it doesn't close over anything inner.
- kafkaService:tpsFor: rewrite as a list comprehension. Easier to read
than the point-free uncurry+second form.
- bulkInsertOtelLogsAndSpans: drop the unnecessary `Relude.when`
qualifier — Relude is imported unqualified.
* Auto-format code with fourmolu
* ingestion: concatMapM, demote bisect-cap log, pattern-guard rawInsert
- kafkaService: switch `concat <$> mapM` back to `concatMapM` from
Control.Monad.Extra (extra is already a build-dep).
- bulkInsertOtelLogsAndSpans: demote BISECT_CAP_MAY_NOT_ISOLATE from
logAttention to logInfo. It is a capacity advisory, not an error;
on-call shouldn't get paged during normal high-volume ingestion.
- bulkInsertOtelLogsAndSpans:rawInsert: switch the single-chunk vs
multi-chunk branch from `case chunks of` to a pattern guard
(`| [single] <- chunks = ... | otherwise = ...`). Shorter and uses
the LambdaCase/ViewPatterns extensions already enabled.
* ci: fix hlint warnings in SchemaCatalog
- upsertSummary: `tryAny ... >>= either (const pass) (const pass)`
collapses to `void $ tryAny ...` (hlint Use whenLeftM_).
- Drop now-unused `Data.Text qualified as T` import.
* cli: adapt to hs-opentelemetry-api API changes
New hs-opentelemetry-api added arguments to two signatures:
- shutdownTracerProvider :: TracerProvider -> Maybe Int -> ... (timeout)
- TracerOptions :: Maybe Text -> [ExceptionHandler] -> TracerOptions
Pass Nothing / [] at the existing call sites.
---------
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
- Replace ad-hoc liftIO getCurrentTime + manual diff with Log.timeAction in
processBatchPipeline and processList; drop redundant local bindings.
- Collapse per-call runLogT scaffolding in kafkaService into a single
LogT/localData context so client_id/group_id/topic ride along automatically.
- Group kafka poll results by topic before dispatch so multi-topic polls get
the right ce-type and attributes per batch.
- Move kafkaProducerProps under getOrInitKafkaProducer (only caller).
- publishToDeadLetterQueue now logs publish failures instead of silently
discarding them; threads appLogger through pubsub/kafka callers.
- Simplify Start.startApp tracer bracket.
* fix(events): default search to exact predicate match, scope expansion to descendants
Events search used to return every span in any matching trace — siblings,
parents, and uncles included — so a query like `kind=="server" AND
sdk_type=="JsExpress"` would also pull in pg.query, pg-pool.connect, and
internal spans that failed the predicate. Two fixes:
1. Default to exact match. New `with_children` flag (POST body field on
/api/v1/events/query, query param on GET /api/v1/events) opts back into
trace context. CLI exposes `--with-children`. UI handler still passes
True to preserve the LogExplorer waterfall.
2. When expansion is on, return only descendants of matched rows. The
broad SQL fetch in selectChildSpansAndLogs is preserved for performance
(one indexed scan on the partitioned otel_logs_and_spans), and a new
keepDescendantsOf BFS trims the candidate set to actual sub-trees of
the predicate hits. The function previously returned the whole trace
despite its name.
Caught a latent bug along the way: selectChildSpansAndLogs built its
column-index map from getProcessedColumns output, which retains
"<expr> as <alias>" strings — so lookups for "latency_breakdown" /
"parent_id" silently returned Nothing. Piped through listToColNames so
the alias-only keys match.
* Auto-format code with fourmolu
* fix(hlint): drop lambda in keepDescendantsOf BFS filter
* review: tighten keepDescendantsOf
- Drop Data.Map.Strict; reuse the already-imported HashMap (HM is O(1)
avg, and avoids a new import surface for one helper).
- Fix latent bug: rows missing a span_id (latency_breakdown) used to pass
the visited-set filter unconditionally via `maybe True`, leaking
unidentifiable rows into descendants. Flip to `maybe False` — drop
them, since we can't dedup or descend from them.
- Switch BFS to prepend-then-reverse to avoid O(n²) `acc <> newKids` /
`rest <> newSids` chains that materialise quadratic spines under lazy
evaluation.
- Cache `V.toList seedSpanIds` (used twice).
- Use `<$ guard` for the CLI flag — matches the surrounding style for
Maybe-shaped param entries.
- Clarify the `count` vs `logsData length` assertion in the LogSpec
trace-tree test.
* test(log-explorer): real cursor-pagination regression test
The existing 'should paginate through multiple pages using cursor' test
fetches page 1 and asserts hasMore=True, but never actually fetches
page 2 — so a broken cursor would silently pass. It also ingests 1000
messages all at frozenTime, so any ordering bug across the cursor
boundary would be hidden by ties.
New test walks pages 1 → 2 → 3 from 5 rows ingested at distinct
timestamps with `| limit 2`, asserting:
* each page returns the exact expected slice (idx [1,2], [3,4], [5])
* the -0.001s cursor nudge actually skips the boundary row (no
overlap between consecutive pages)
* hasMore flips False on the last page
Identifies rows by querying `attributes.page.idx` from the DB after
extracting `latency_breakdown` (= context___span_id) from each result
row — LogResult doesn't surface attributes directly.
* test(log-explorer): guard cursor advancement and pageIdsFor shape
* review: address keepDescendantsOf review notes
- Rename bfs → go (the queue prepended new nodes, making it DFS in
practice; correctness is unaffected but the name was misleading).
- HM.findWithDefault [] over fromMaybe [] (HM.lookup ...) for the
parent→children lookup, matching the library's intended API.
- Drop two over-explanatory inline comments inside keepDescendantsOf
(the maybe False rationale and the prepend-build-then-reverse note);
Haddock already covers what's load-bearing.
- Document in Haddock that seeds are not included in the output —
callers already hold them in requestVecs.
- Note on the V.null seedSpanIds early return: kept to skip the SQL
round-trip; comment why so the redundancy with keepDescendantsOf's
guard reads as intentional.
---------
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Numbers, bools and arrays were being stringified in the events search
JSON envelope, and the summary column came back as a JSON-encoded string
of `field;style⇒value` markup. Agents/jq consumers had to coerce every
cell. Now rawRows preserve types end-to-end, empty strings collapse to
null, and summary/latency_breakdown are run through cleanSummaryValue.
Adds two new commands:
- `send-event -m TEXT` — sends a real OTel span from scripts/CI; supports
--level, --service, --tag, --extra, --resource flags
- `telemetrygen --kind=trace|log|metric` — generates synthetic load at a
configurable rate; replaces the external `go install telemetrygen` workflow
Both commands derive the OTLP gRPC endpoint from the configured API URL
and authenticate automatically via the stored API key.
Shared helpers (withOtelProvider, configureOtelEnv, sendSpan) eliminate
duplication between the two commands.
Updates README, docs/getting-started.md, docs/DOCKER_COMPOSE_README.md,
and the onboarding modal to use the new commands instead of the external
binary with manual API key headers.
* fix: surface OTel exception events in KQL search and issue stack_trace
- KQL: `attributes.exception.{type,message,stacktrace,escaped}` now
COALESCEs the flattened column with a jsonb_path probe over span events
(`event_name=="exception"`). Previously every search returned 0 because
hs-opentelemetry records exceptions as span events, not flat attributes.
- API: `apiIssueGet` synthesises a span-chain "logical stack" when the
source `exception.stacktrace` is empty (true for 100% of monoscope's
prod exceptions — GHC backtraces aren't captured by default).
- Pages.Telemetry: guard `renderSpanRecordRow` integer divisions against
divide-by-zero (empty service group / all-instant trace) — root cause
of the recurring `ArithException` in the issue detail page.
- CLI: bump HTTP response timeout 30s -> 5min so long KQL searches
don't surface as `ResponseTimeout` errors.
* Auto-format code with fourmolu
* feat: improve CLI events get and KQL field aliases
- events get: add --at TIMESTAMP flag for direct O(1) lookup via new
GET /api/v1/events/{id}/time/{ts} endpoint (uses timeseries partition
key instead of 90d range scan)
- events get: extend default lookback from 24H to 90d so older spans
resolve without --at
- KQL: map span_name → name, service → resource___service___name,
trace_id → context___trace_id in WHERE clauses (these are SELECT
aliases, not real columns; querying them returned HTTP 400)
- CLI: add client-side KQL validation via validateQueryOrDie; bad queries
now show Megaparsec's precise error instead of an opaque server 400
- queryMetrics: fix error body (was "Invalid signature; \"...\""
instead of "Invalid query: ...")
* fix: correct synthStackFromSpans service lookup and cleanup
- synthStackFromSpans: replace wrong nested resource["service"]["name"]
lookup with Telemetry.spanServiceName (uses flat "service.name" key,
matching the SQL and the existing helper). Service labels were always
empty before this fix.
- Remove now-unused AesonText import from ApiHandlers
- enrichIssue: add TODO comment noting the N+1 query cost and the
correct long-term fix (store synthetic stack at ingestion time)
- apiEventGet: move to after apiFacets with proper haddock; fix
orphaned apiFacets comment that was running into apiEventGet's doc
- Expr.hs: document outputFieldAliases SELECT/WHERE trade-off and the
exceptionFlattenedFields/flattenedOtelAttributes overlap invariant
- Commands.hs: clarify why validateQueryOrDie is not needed for tail/
context (those commands have no user-supplied KQL field)
* review: address code-review feedback
- cli: validate --at ISO-8601 timestamp eagerly; an unparseable value now
prints an error and exits non-zero instead of silently falling back to a
90d range scan that hides the typo.
- cli: fix Haddock — collapse two stacked '-- |' markers on
validateEventsOpts into one block so haddock no longer drops the second.
- api: enrichIssue decodes via the typed RuntimeExceptionData record
instead of pattern-matching raw JSON, so a renamed/removed field is a
compile error rather than a silent "" default.
- parser: derive exceptionFlattenedFields from flattenedOtelAttributes so
the "every entry must also live under attributes.exception.* there"
invariant is enforced automatically.
- api: hoist the empty-spans guard to the top of synthStackFromSpans so
the sortOn/map allocation is skipped on the empty case; replace
'maybe "" (\\v -> …)' with foldMap to match codebase style.
- tests: add Web.ApiHandlersSpec covering synthStackFromSpans
(ordering, error marker, missing service/name/span_id, empty list);
extend ParserSpec with output-field-alias and exception-COALESCE
rewriting cases.
* fix(hlint): use sortOn from Relude in synthStackFromSpans
Relude already re-exports sortOn so the explicit List. qualifier is
redundant. The hlint check on the previous commit was already flagging
this; my edit kept the same call shape and inherited the warning.
* docs and readme
* fix: update tests to reflect trace_id alias resolution and resource format
- Update isnotnull(trace_id) test to expect context___trace_id in WHERE
- Fix resource field format in ApiHandlersSpec to use service.name flat key
- Add bracketed service assertion to synthStackFromSpans test
- Simplify span context access and trim verbose comments in ApiHandlers
* fix(telemetry): atMapText falls back to flat dotted keys
getNestedValue now tries the literal joined key before splitting on '.',
so resource maps with flat 'service.name' keys resolve correctly. Also
inlines exceptionFlattenedFields, tightens CLI event-id validation, sets
a 5m wreq timeout for long aggregates, and refactors the span-order
test to compare line indices.
* fix(test): import findIndex in ApiHandlersSpec
* fix(test): replace Data.List.findIndex with list comprehension
---------
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Claude <noreply@anthropic.com>
* refactor: extract mkPageCtx page-handler bootstrap
Replaces 25+ inlined BWConfig + sessionAndProject + ask AuthContext
sequences across Pages/* with a single mkPageCtx helper that returns
(Session, Project, BWConfig) seeded with the always-set fields
(sessM, currProject, config). Net: -102 LoC.
* refactor: extract toastError helper for validation/error responses
Collapses the recurrent 'addErrorToast msg Nothing >> addRespHeaders payload'
shape across Pages/Projects.hs and Pages/Dashboards.hs. Net: -5 LoC plus
+5 LoC for the helper and re-export = breakeven, but eliminates a duplicated
two-line pattern at every site.
* refactor: extract withCardFrame for widget renderer card shell
Collapses the duplicated outer flex-wrapper + optional widget header +
flex-1 fill + surface-raised card shell shared between renderLogsWidget,
renderTraceTable, and renderTable into a single helper.
* refactor: derive FieldTypes/FieldCategoryEnum JSON via WrappedEnumSC
Replaces manual ToJSON/FromJSON instances and the parseFieldTypes,
parseFieldCategoryEnum, fieldTypeToText helper functions with a single
deriving-via clause that piggybacks on the existing WrappedEnumSC wrapper.
Net: -50 LoC.
* refactor: distill CLI; reuse server types (Schema, Paged, MetricsData)
Replace ad-hoc AE.Value field-name probing in the CLI with typed parsing
against the server's existing response types so wire-shape changes
surface as compile errors:
- filterSchema now takes Models.Telemetry.Schema.Schema
- normalizeList decodes via Web.ApiTypes.Paged
- metrics renderers/assertion take Pages.Charts.Types.MetricsData
Also: factor validateDurationOrDie/validateAndNormalizeKind helpers,
fold three case-of-OutputMode triplets through renderWith, derive
ServiceRow/ConfigGetJson/AuthStatusJson/TraceSummary JSON via CustomJSON.
* Auto-format code with fourmolu
* docs: add agentic-pipeline showcase to README
Show the discover -> search -> context -> triage chain with the stable
JSON envelope each command exposes, so a reader scanning the README sees
why the CLI is agent-friendly before getting to the Claude Code Skills
section.
* refactor(cli): apply hlint fixes (errors + warnings + suggestions)
Drops 2 errors and 7 warnings from `hlint cli` to 0:
- Resource.hs: `>> pure n` -> `$> n` for the apply-batch counter
- Commands.hs: `ExceptT . pure` -> `hoistEither`; manual `maybe ... Right` -> `maybeToRight`
- Config.hs: `either (const x) id` -> `fromRight x`
- UI.hs: `T.unlines` -> Relude's `unlines`
- Validate.hs: drop redundant `void`
Plus higher-impact suggestions:
- `data` -> `newtype` for ServicesCommand, ConfigGetOpts, ServicesListOpts, FormEditorName
- `fmap (fmap f) x` -> `f <<$>> x` in 6 list/schema/facets handlers
- `Set.singleton` -> Relude's `one`
- `pure ()` -> `pass` in UI event handlers
- `\_ -> action` -> `const action`
- `(\c -> Map.lookup c idxMap)` -> `(`Map.lookup` idxMap)`
Build green: `All good (8 modules)`. The two remaining hlint hints
(`Data.Set as Set` alias, `universe` import) belong in `.hlint.yaml`
rather than the code.
* fix(test): unwrap IssueL through .base.id in ErrorPatternsSpec
IssueL is a list-view wrapper around Issue (`{ base :: Issue, ... }`),
so the `id` field is reached through `.base`. Four sites in
ErrorPatternsSpec called `issue.id` directly, which type-errored after
IssueL gained its current shape. Matches the pattern used in
src/Pages/Anomalies.hs (`errL.base.id`).
* fix(cli): accept uppercase --since suffixes (1H, 24H)
The platform's TimePicker emits uppercase ("1H", "24H") — the same
strings the CLI itself injects as defaults via buildSearchParams /
runServicesList — but the validator only accepted lowercase, so a user
who copy-pasted the displayed default got rejected client-side. Fold
the suffix to lowercase before checking the whitelist; preserve
original casing in the returned value.
Add a regression test in CLIE2ESpec asserting `--since 1H` is accepted.
* fix(cli): four review nits — flag-aware duration error, single decode, doc/comment fixes
1. validateDurationOrDie now takes the flag name, so --window failures
read "--window must match …" instead of "--since must match …".
Three call sites updated (--since x2, --window x1).
2. aggregateServices keeps its `Map.lookup "service"` but with a comment
pinning the contract: the server aliases resource___service___name to
the short "service" column in colIdxMap (Pages.LogExplorer.Log.allCols).
Added an e2e regression in CLIE2ESpec that asserts services-list still
returns {services, count} so a column rename can't silently zero the
output.
3. docs/cli.md agentic-incident-workflow: replace .items[] with .data[]
in the issues-list and log-patterns-list jq pipes — they'd break
against the now-normalised list envelope.
4. runEventsContext --summary used to walk logsData twice (once in
normalizeEventsResponse, once in withTraceSummary). Introduce a
`DecodedEvents` record that carries idxMap/rows/raw, decode once at
the call site, and thread it through both functions.
* perf(cli): back services list with /api/v1/facets instead of 10k-event scan
Previously runServicesList fetched up to 10,000 events and aggregated
client-side because /api/v1/metrics returned empty for `summarize ... by`.
The facets endpoint (precomputed in apis.facet_summaries via a background
job) is one indexed lookup — sub-millisecond regardless of project size.
- Drop the events-scan fallback; switch to GET /api/v1/facets?field=resource.service.name
- ServiceRow loses the `last_seen` field (facets don't track per-value
recency). The remaining {name, events} is what the use-case actually
needs ("what services exist? how busy are they?").
- e2e regression in CLIE2ESpec already covers the {services, count}
envelope; comment updated to pin the facets-backed contract.
- docs/cli.md and monoscope-skills/README.md output-shape tables drop
last_seen.
* chore: trailing review nits — unused asks, V.take, normaliseList warning, debug-comment
Server side:
- dashboardsGetH: drop the appCtx ask; bw.config.liveReloadDashboards
carries the flag now that mkPageCtx threads config through.
- teamAlertsGetH: drop appCtx + sess + project, all genuinely unused
bindings.
(dashboardGetH / dashboardTabGetH / dashboardTabContentGetH /
unifiedMonitorOverviewH still bind appCtx — they read appCtx.env.* which
isn't on bw.config, so the binding is required.)
CLI:
- takeFirstEvent: V.take 1 arr instead of take 1 (toList arr); avoids the
Vector→List→Vector round-trip.
- Resource.normalizeList split into normalizeListE returning Either, with
Left carrying the raw payload on shape mismatch. runList now flags the
unrecognised case via printDebug so contributors see when the server
envelope drifts.
- Core.renderParams: comment pinning the "debug-only, NOT URL-encoded,
redact if you ever pass auth via query params" contract.
* fix(cli): CI doc uses .data, emitFirstEventId avoids list round-trip
- docs/cli.md CI/Automation example said `jq '.items | length'` —
inconsistent with the agentic-workflow section above and broken since
normalizeList moved everything to `.data`.
- emitFirstEventId now uses `arr V.!? 0` instead of `head (toList arr)`
— same V.take optimisation as takeFirstEvent, no intermediate list
allocation.
Bug #2 from the review (renderEventsTable in runEventsContext receiving
raw `val`) is intentional: the table renderer reads `cols`/`logsData`
out of the raw envelope to produce a dense table; the JSON path uses
the named-field `enriched` form. Passing `enriched` to the table
renderer would render an empty table because the normalised shape
doesn't carry `cols`/`logsData`. The split is by design — table mode is
column-oriented, JSON mode is record-oriented.
* fix(cli): six review items — agent-mode truthy check, --first table parity, single decode, derived facets, forall pinning
3. isAgentMode no longer fires on `CI=` (empty) or `CI=false` — many dev
shells export CI unconditionally; treat empty/"false"/"False"/"FALSE"/"0"
as off.
4. --first parity in table mode: takeFirstRow trims the raw envelope's
logsData to one row, so table output matches the JSON path's
takeFirstEvent slice instead of silently dumping all rows.
5. Free-floating Haddock above DecodedEvents merged into the actual
docstring (the orphan would have triggered a -Wmissing-signatures-ish
warning eventually).
6. runEventsSearch now decodes once and threads DecodedEvents through
normalizeDecoded, matching the pattern in runEventsContext. No more
double walk of logsData on the JSON path.
7. parseServiceFacets decodes via Models.Apis.Fields.FacetValue
(FromJSON via DAE.CustomJSON) — drops 6 lines of hand-rolled
KM.lookup/AE.String/AE.Number unpacking and surfaces server wire
changes at compile time.
8. withMetricsData now forwards aeson's Error message through
printDebug so a server envelope drift is visible under --debug
instead of vanishing into the raw-JSON fallback.
Plus: explicit `forall es a b` on apiGetJson/apiPostJson/apiPutJson/
apiPatchJson so TypeApplications at call sites
(`apiPostJson @_ @_ @AE.Value`) survive future constraint reordering by
fourmolu/hlint. Without this, alphabetical resort of
`(AE.FromJSON b, AE.ToJSON a, Environment :> es, ...)` flips the
implicit type-var order and the third `@AE.Value` lands on `es` instead
of the body type.
Run fourmolu over the touched files; #10 (double blank lines) is
auto-fixed.
* chore: round-three review nits — null instead of "", helpCases coverage, dead code
3. normalizeDecoded now emits `null` for missing fields (row shorter
than colIdxMap) instead of silently coercing to "". Agents can now
tell "field present but empty" from "field missing".
4. Comment on the --debug setEnv site explaining the single-shot, pre-
thread-spawn placement (so future readers know it's not a race).
6. emitCompletion lifts to IO once at the entry instead of per-branch.
7. CLIBinarySpec helpCases extended with `facets`, `schema`, `services
list`, `events {search,context,tail}`, `metrics {query,chart}`,
`auth login`, `completion` — every top-level subcommand now has a
--help smoke test.
8. dashboardGetH/dashboardTabGetH: replace appCtx.config.hostUrl with
bw.config.hostUrl now that mkPageCtx threads config through; appCtx
binding stays because appCtx.env.apiKeyEncryptionSecretKey isn't on
bw. dashboardTabContentGetH untouched — no mkPageCtx, no bw.
Style:
- Drop dead extractNumericRows from Pkg.CLIFormat (no callers since the
metrics path moved to MetricsData).
- Export validateDurationFor so the doctest examples (which reference it)
resolve at doctest runtime.
Re #2 (parseServiceFacets via FromJSON) and #10 (double blank lines):
already landed in commit 2a54bd0b.
Re #5 (renderParams allowlist): keeping the comment-only enforcement.
The CLI doesn't put credentials in query params today (Bearer + project
go via headers), and a type-level distinction would touch every API
helper for a hypothetical leak. The doc note is sufficient guard for
contributor review.
* fix(cli): trace-summary error count, KQL injection guard, small wins
Core bug:
- withTraceSummary used `cell "severity"` to count errors per trace, but
the events colIdxMap (Pages.LogExplorer.Log.allCols) doesn't expose a
"severity" column — only the boolean "errors" flag. Result: every
trace summary reported error_count=0 regardless of traffic. Now reads
cell "errors" and treats "true" as 1. Also added a defensive fallback
for service ("service" <|> "resource.service.name") in case a future
column-set rename hides the alias.
KQL safety:
- runEventsGet escapes `\` and `"` in opts.eventId before splicing into
the query string. Event IDs aren't UUIDs (synthetic-… span ids,
log_pattern integers, …) so format validation is wrong; escaping
defends against injection regardless of shape.
Small wins:
- isAgentMode: fold suffix to lowercase once via `map toLower v` rather
than enumerating "false"/"False"/"FALSE".
- filterSchema: `Map.take n filtered` instead of
`Map.fromList (take n (Map.toList filtered))` (containers ≥ 0.6.3).
Doctests note:
- The earlier "doctest comment" the review flagged was already removed
in a prior commit. Pure validators are still covered end-to-end via
CLIE2ESpec ("--since 1xyz", "--kind banana") which exercise the real
binary. Wiring CLI/* into the lib doctest target would need moving
the modules into src/ — out of scope for this round.
Pushed back:
- AuthStatus method ADT: the strings "env"/"token" appear in two
adjacent lines of one function. A two-constructor ADT + custom ToJSON
is more boilerplate than the leak it prevents (no other consumer).
- Map merge tuple readability: a record would cost two lines and one
data declaration to save zero compile errors. The merge fits on one
line; leaving it.
* chore: clarify intent — `mempty` empty bodies, capFacets exhaustiveness comment
- Pages/Projects.hs: replace seven `toastError "…" ""` and three
`addRespHeaders ""` with `mempty`. Reads as "no body" rather than
"empty string body" — the underlying type is `Html ()` whose Monoid
identity is the empty Html, same wire effect.
- cli/Main.hs `capFacets`: comment on the trailing wildcard clause so
the (Just n) + non-Object fallthrough is explicitly documented vs the
Nothing case the first clause already handles.
Re facets in CLIBinarySpec helpCases: already added in 76ebf71c.
* chore: clear hlint warnings across src/
BackgroundJobs.hs: 6× displayException + 2× fromException were
resolving to Control.Exception.* (because the file did
`import Control.Exception (ErrorCall (..))` which exposed the
Control.Exception namespace to GHC's resolver and tripped
.hlint.yaml's "use Relude version" rule). Drop the unqualified
import; ErrorCall now goes through `Control.Exception qualified as CE`
so the only `Control.Exception` reference is the explicit qualifier
and unqualified displayException/fromException resolve through Relude.
Pages.Anomalies.hs:
- NE.head -> head (Relude)
- two whenJust lambdas -> point-free (toHtml composition)
- when (... `notElem` ...) -> unless (... `elem` ...)
Pages.Settings.hs: sortOn fst -> sortWith fst (no caching needed for
projection on first component).
Pages.Telemetry.hs: MapS.singleton k v -> one (k, v) (Relude polymorphic
container constructor).
Pages.LogExplorer.Log.hs: two `\j -> (j, x)` lambdas -> tuple sections
(TupleSections is part of GHC2024, already on).
Verified clean: `hlint src/BackgroundJobs.hs src/Pages/Anomalies.hs
src/Pages/Settings.hs src/Pages/Telemetry.hs src/Pages/LogExplorer/Log.hs`
reports "No hints".
* Auto-format code with fourmolu
* fix(cli): forward Paged decode error, lenient body decode, unambiguous null
3. normalizeListE now returns Either (rawValue, reason) — when an Object
fails @Paged@ decoding the aeson Error message is captured and
surfaced through runList's --debug log. Contributors see *what*
shape changed instead of a generic "did not match envelope".
normalizeList stays pure with `either fst id . normalizeListE`.
4. httpExToError now uses TE.decodeUtf8With lenientDecode for both the
status reason and the response body. The body field of
StatusCodeException is the truncated raw bytes (≤1024B) and may be
gzip leftovers, HTML, or arbitrary upstream junk; strict decodeUtf8
would throw at the error-handling boundary, masking the real HTTP
failure with a UnicodeException.
6. isAgentMode: replace `not (null v)` with `v /= ""` to dodge the
Prelude.null vs Relude container-null ambiguity. v is String here.
Re prior style notes:
- Double blank lines between top-level bindings: produced by fourmolu
format pass; project convention is enforced by the formatter.
- AuthStatusJson / ConfigGetJson: kept as records — the CamelToSnake
derivation pins the wire shape (an inline AE.object [...] silently
drifts when fields are added).
- runConfigGet OutputMode asymmetry vs init/set: noted; init takes no
output path (interactive form) and set has no payload to format.
Follow-up if it surfaces in user feedback.
- normalizeList = either fst id . normalizeListE (was either id id;
shape changed when Left started carrying the reason).
- CLI/UI.hs `currentValue = T.strip . unlines . BE.getEditContents`:
verified — `unlines` here is the Relude re-export, which is polymorphic
over the IsList/Foldable instance. Compiles clean.
* fix(test): unwrap IssueL through .base.issueType in AnomaliesSpec
Same pattern as the ErrorPatternsSpec fix in f0ab87d7: IssueL wraps
Issue in `base`, so the issueType field lives at `c.base.issueType`,
not `c.issueType`.
* fix(test): add http-types to integration-tests build-depends
CLIE2ESpec imports Network.HTTP.Types.Status (statusCode) for the
reachability probe but http-types wasn't on the test target's deps.
The library has it (line 173), the test suite didn't inherit it.
* fix(test): drop System.Environment lookupEnv import — Relude already exports it
* fix(cli/test): try @SomeException → tryAny, plus seven review nits
Build break: UnliftIO.Exception.try's forall is (m, e, a); the @SomeException
TypeApplication landed on m and tripped a kind error. Switch to tryAny
(MonadUnliftIO m => m a -> m (Either SomeException a)) — same semantics,
no type-arg juggling. Drops the redundant Data.List (isInfixOf) import
along the way (Relude already has it).
Review items:
- AuthStatus: extract printStatus helper for the env/token branches —
3 lines repeated → 1 helper.
- apiFacets: replace addUTCTime (-86400) with negate nominalDay (Data.Time
re-export). No more 86400-second magic.
- capFacets: drop the now-redundant `Nothing` clause; the wildcard at the
end already covers it (and `Just n` + non-Object).
- parseDurationMs: T.toLower the input via a view pattern so --watch 5M
is treated like --watch 5m, matching validateDurationFor's case-insensitive
suffix check.
- reqOpts: collapse the broken multi-line `& W.header "Accept" .~ [...]`
formatter artifact into a single parenthesised line so subsequent `&`
chain elements align under W.defaults.
- Resource.runList: route the unrecognised-envelope reason through
printError (always-on stderr) instead of printDebug — agents
expecting {data, pagination} need to see when the server returns
something else.
- DecodedEvents: decode count / has_more / cursor at construction
instead of carrying the raw KeyMap. Three lookups eliminated, intent
is local to decodeEvents.
Re minor nits noted:
- isAgentMode/isDebugMode re-read env on each call: kept; called at
most a handful of times per invocation, no hot path.
- Data.Char (toLower): kept the explicit import — Relude does re-export
it but the explicit form documents which `toLower` we mean (vs
Data.Text.toLower which operates on Text).
* fix(test): restore Data.List (isInfixOf) — Relude does not re-export it
* fix(test+ci): two CLI test failures, default e2e suite to the public demo project
CI failures:
1. CLIBinarySpec "auth status exits and shows status message" — when CI
is set, isAgentMode auto-enables and `auth status` emits the JSON
envelope ({authenticated, method, api_url, project}). The test
predicate only matched plain-text "Not authenticated" / "Authenticated".
Extend it to also accept the JSON shape ("authenticated" key).
2. CLIBinarySpec "completion --help" exited 1. The completion subparser
declared a positional SHELL argument but didn't attach `<**> helper`,
so `--help` got consumed as the SHELL value and fell through to the
"unknown shell" branch. Adding helper makes optparse intercept --help
and emit usage with exit 0.
Default e2e suite to prod demo project:
CLIE2ESpec.getE2EConfig now defaults @MONOSCOPE_API_URL@ to
https://api.monoscope.tech and @MONOSCOPE_PROJECT@ to the public demo
00000000-0000-0000-0000-000000000000. Only @MONOSCOPE_API_KEY@ has no
safe default (read keys for other projects would leak). The pendingWith
message names the missing variable specifically so it's obvious how to
unlock the suite.
Wired up in CI: pullrequest.yml's integration-tests step now reads
MONOSCOPE_API_KEY from secrets.MONOSCOPE_DEMO_API_KEY. Repo owner needs
to mint a read-only key on the demo project and store it in repo
secrets — then every PR/push runs the suite against prod. Forks (no
secret) still get the CLIBinarySpec smoke tests.
docs/cli.md updated with the new default-driven workflow.
* fix(test): assert specific merge edge instead of fragile global count
ErrorPatternsSpec test 12 ("Full LLM pipeline") asserted
`length mergedTypeErrors == 1` after merging one (childId, canonId)
pair. The query counted ALL TypeError patterns with canonical_id set
across the project, so any earlier test that merged a TypeError made
the count drift up. CI now sees 2 because some upstream change in the
LLM/notify pipeline produced an additional pre-merged pattern by the
time test 12 runs.
Switched to a behaviour-level assertion: query the specific childId's
canonical_id and require it equals canonId. This tests what
`assignErrorsToCanonical` actually does, with no dependence on the
global state earlier tests left behind.
* refactor(cli): lens-based shape ops, guards, precomputed indices
- takeFirstEvent / takeFirstRow rewritten with Data.Aeson.Lens
(`AL.key … . AL._Array %~ V.take 1`). 14 lines of pattern-match-and-
reinsert collapsed to two one-liners. AL is qualified to avoid
collision with the `key` record fields on ConfigSet/GetOpts.
- foldFiltersIntoQuery: case (T.null …, T.null …) → pattern guards.
Reads top-down instead of as a tuple-match puzzle.
- normalizeListE: AEKM.member → has (AL.key …) lens accessor; the
branch matches on the value directly, no need to destructure
AE.Object obj first.
- extractColIdxMap: round → floor on the Scientific. Column indices are
already-integral non-negative ints; floor documents that intent and
drops banker's rounding semantics that don't apply here.
- parseDurationMs: drop the bare-number-as-seconds fallback. validateDurationFor
rejects unsuffixed values upstream, so the only paths that reach
parseDurationMs have a recognised suffix; a typo now returns the
5s default loudly instead of silently treating "5" as five seconds.
- runMetricsChart: validateDurationOrDie "--watch" before entering the
forever loop, so a bad --watch string fails at flag parse rather
than burning HTTP requests every 5s.
- withTraceSummary: precompute svcIdx / errIdx / tidIdx once instead
of re-Map.lookup'ing the same column path per row. Saves one Map
lookup per cell per row — measurable on large
`events context --summary` payloads.
Build deps: cli exe gains `lens-aeson` (already in lib deps for the
server's Data.Aeson.Lens usage). hpack regenerated monoscope.cabal.
---------
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
All CLI resource paths were missing the /api/v1 prefix, causing 404s
(e.g. GET https://api.monoscope.tech/issues instead of /api/v1/issues).
- cli/CLI/Resource.hs: resourcePath now returns /api/v1/<slug>
- cli/Main.hs: all explicit paths (/issues, /me, /project, /share,
/schema, /api_keys, dashboard widgets) prefixed with /api/v1
- src/Pkg/TestUtils.hs: extend runHTTPtoServant to route GET/POST for
all api/v1 paths (issues, endpoints, log_patterns, me, project) to
real handlers; support POST/PUT/PATCH/DELETE dispatch
- test/integration/CLI/CLISpec.hs: add cliPost helper and CLI resource
command tests that exercise /api/v1/* paths through the HTTP router
- test/integration/Web/ApiV1Spec.hs: document path list as CLI contract
- Two-pass spike detection: run detectLogPatternSpikes twice in tests
(first pass pends, second pass fires) to match the algorithm
- Doctest imports: qualify with package name (-XPackageImports) and
filter --color/--jobs/--match flags unsupported by doctest runner
- Channel gating: add countingSendGated in Settings for slack/discord/
whatsapp/pagerduty; wire channel setup in BotTestHelpers; clear
disabled_channels in test teardown
- AnomaliesSpec: use frozenTime instead of getCurrentTime for determinism;
add missing summary column to INSERT
- LogPatternsSpec embedding test: isolate pre-existing patterns via
merge_override before running patternEmbeddingAndMerge
- ErrorPatternsSpec: exclude PatternEmbeddingAndMerge from background
jobs run to allow explicit embedding control in test 12
- BusinessFlowsSpec: add ON CONFLICT DO UPDATE for idempotent test setup
- Fix Slack report assertion: T.toTitle transforms "TEST" → "Test",
so assert "🧪 Test:" instead of the original "🧪 TEST"
- BackgroundJobs SQL: cast cooldownHours to ::int in make_interval call
- PatternMerge: add ORDER BY id to getUnembeddedLogPatterns for stable
ordering under the 500-row limit
- Distill Settings.hs: merge countingSend/countingSendGated into one
helper with Maybe Text gate argument
- Distill TestUtils.runAsBase: replace case result with >>= either
- Distill CLIBinarySpec: inline single-use combined let binding
- Add golden files for Slack conversations API responses
Adds CLI subcommands and JSON handlers covering /me, /project,
/endpoints, /log_patterns, /issues, /teams and /members, plus shared
paginated envelope and generic bulk result parameterised by id type.
Extracts share-event persistence into Models.Apis.ShareEvents and
introduces a SnakeJSON type alias to collapse repeated deriving-aeson
boilerplate.
- CLI error rendering via renderAPIError (HTTP code + body), drops raw Show output
- decodeBody no longer coerces empty body to "null"; surfaces explicit error
- httpExToError includes URL/method context across all branches
- runApplyResource counts per-file failures; exits non-zero only at the end
- Collapse redundant CLI wrappers; add withCfgMode, resourceIdPath helpers
- readYamlOrJson branches on .json extension for clearer parser errors
- Rename apiPost to apiPostUnauth to flag its device-code bootstrap role
- BulkFailure serialises as {id, error}; add Eq for test assertions
- Integration tests for bulk failure shape, unknown action 400, monitor 404
- CLI help-case test asserts invoked subcommand name appears in output
- Add --version subcommand test
- docs/cli.md: fix apply wording, drop anomalies bulk row, sync 48h TTL marker
- Fix anomalies bulk UUID encoding + archive via new archiveAnomaliesAndIssues helper
- Fix mkPaged hasMore off-by-one; hide Paged constructor behind mkPaged
- Fix endpointsListH per_page count via Endpoints.countEndpointsByProject
- Collapse DashboardFull to nest DashboardSummary; drop mkSummary HasField shim
- Collapse monitor/dashboard/api-key lifecycle handlers with withRefetch helper
- Drop unused monitor list filter/since query params
- Scope Resource.runApply to dashboards (removes dead MonApply path)
- Add dashboards unstar CLI command; fix star help text to drop toggle claim
- Extract CLI runAPI + export withResult helpers to dedupe Main.hs dispatch
- Fix CLI decodeBody to surface parse errors on empty body instead of silent null
- Replace runStaticUUIDRef crash on empty pool with UUID.nextRandom fallback
- Tighten CLI jsonCall constraint to IOE only (HTTP travels via action param)
Public JSON REST API under /api/v1 for monitors, dashboards, api-keys,
events, metrics, anomalies, issues, log-patterns, endpoints, projects,
share links, and `me` — mirrors HTML handler behaviour but runs in
ATBaseCtx with API-key auth.
Handlers & types (Web.ApiHandlers, Web.ApiTypes):
- Thin JSON adapters over existing Models.*. A single `bulkExec`
dispatcher backs monitor/dashboard/log-pattern/anomaly bulk endpoints;
anomalies bulk routes through Anomalies.acknowlegeCascade so API
cascade semantics match the HTML path.
- Issues share a `mkSummary` HasField-polymorphic projection used by
both IssueSummary and IssueFull, and an IssueOp ADT that compresses
ack/unack/archive/unarchive into one runner.
- Paged envelope (items/total/page/per_page/has_more) + mkPaged.
- Log patterns and anomalies are thin aliases over issues filtered by
issue_type so one storage path backs three UI views.
Shared code consolidated:
- Monitors created via API parse KQL → SQL via Pkg.Parser, matching the
form path (previously logQueryAsSql was empty).
- ProjectApiKeys.encodeApiKeyB64 consolidates encrypt+base64 previously
duplicated across ApiHandlers, Pages.Settings, Pages.Projects.
- Dashboards.getDashboardByFilePath replaces an inline raw-SQL query in
the YAML-apply handler.
- Pkg.DeriveUtils.JsonValueSchema DerivingVia wrapper replaces five
manual ToSchema fallbacks.
- Data.Effectful.UUID.runStaticUUIDRef: shared IORef-backed UUID pool so
deterministic test UUIDs persist across handler invocations in a
single test. Pkg.TestUtils exposes runAsBase / trUUIDRef.
- Models.Apis.Issues.IssueType/IssueEvent now derive Bounded/Enum and
gain ToSchema + FromHttpApiData through WrappedEnumSC, flowing
through Servant capture/body params + OpenAPI without manual
instances. Models.Apis.Endpoints.Endpoint derives HI.DecodeRow.
CLI (cli/):
- Top-level commands wired against the new endpoints.
- Shared OutputMode (JSON/YAML/Table) + detectOutputMode + renderByMode
+ apiGet/apiGetJson/apiPost/apiPostJson/apiPutJson/apiPatchJson/
apiDelete helpers in CLI.Core.
- package.yaml for the `monoscope` exe: Paths_monoscope added to
generated-other-modules (needed for --version); base listed as an
explicit dependency; `Relude as Prelude` mixin dropped so the autogen
Paths_monoscope.hs `import Prelude` is unambiguous. All cli modules
already import Relude explicitly. Fixes a CI build failure on Linux.
Integration coverage: 18 tests in Web.ApiV1Spec covering OpenAPI spec,
auth, schema/monitors/events/metrics endpoints, monitor and dashboard
CRUD + lifecycle, API key CRUD, events body query, share links.