feat(genai): add AI span inspector - #2
Draft
Sun-ZhenXing wants to merge 56 commits into
Draft
Conversation
Sun-ZhenXing
marked this pull request as draft
July 30, 2026 03:34
…re tests (SigNoz#12383) primus bumped golangci-lint to v2.12.2, whose govet now runs the inline analyzer and whose sloglint is stricter. CI resolves primus.workflows@main, so every PR started failing lint the moment that landed. - reflect.Ptr is a deprecated alias carrying //go:fix inline, so it is now reflect.Pointer at all five call sites - metricsstatementbuilder imported golang.org/x/exp/slices, which carries //go:fix inline pointing at the stdlib; the analyzer cannot inline generics, so switch the import to stdlib slices as the directive intends - pkg/instrumentation/loghandler emits OpenTelemetry semantic-convention attributes (code.filepath, exception.type, ...), which are dotted rather than snake_case by definition. Renaming them would break every log consumer, so the keys move to constants in instrumentationtypes, which already held this kind of key -- and already defined code.function, so source.go was duplicating it. Six of the seven alias the semconv constants that define them; exception.code has no OTel equivalent. sloglint resolves a same-package constant back to its literal but skips a qualified one, so this needs no exclusion. CI reported 8 issues but capped at max-same-issues=3, hiding 2 more reflect.Ptr sites and 4 more sloglint ones. Separately, TestTimeout/WaitTillNoTimeoutForExcludedPath failed with "transport connection broken: http: CloseIdleConnections called". TestTimeout and TestCache issue requests through http.DefaultClient while a parallel subtest in response_test.go closes an httptest.Server, and httptest.Server.Close calls http.DefaultTransport.CloseIdleConnections. Both tests now use their own client and transport, and are closed via t.Cleanup. That makes Serve return ErrServerClosed on every run, so the require.NoError wrapping it is dropped -- it could never have held, and require runs t.FailNow off the test goroutine anyway. Bare Serve in a goroutine matches routerweb and render tests.
…pinned serving (SigNoz#12324) * feat(querier): wire clickhouseprometheusv2 for shadow comparison and pinned serving Stand up the v2 provider next to the default one and give the querier two flag-gated ways to exercise it, neither affecting default serving: - shadow: with use_prometheus_clickhouse_v2 on, every PromQL query re-runs on v2 after the response is sent; result diffs are logged. Bounded by a small per-process admission cap (skip, not queue, at the cap). - pin: the X-SigNoz-PromQL-Provider header serves the response from v2 directly, for side-by-side comparison by integration tests and support. Pinned requests bypass the cache in both directions. Typed storage errors (series/sample budgets) survive the engine's wrapping and surface as 4xx instead of internal 500s. PromQL results now carry ClickHouse scan stats. prometheus::provider: clickhousev2 also makes v2 the serving provider outright (no shadow in that mode - nothing to compare against). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * test(promql): replay the conformance corpus on both providers Every corpus case now runs twice — default provider, and pinned to clickhousev2 via the flag-gated X-SigNoz-PromQL-Provider header — each leg asserted against the same frozen expectations, each leg with its own known-divergences ledger enforced in both directions. The new known_divergences_v2.json is the rollout scorecard: the provider swap is measured by burning it to empty. The legs are never asserted against each other: both can sit within one rounding quantum of the expected value yet differ by up to two quanta at a rounding boundary, so a leg-vs-leg check would reintroduce the boundary noise the tolerance absorbs. Anchoring both to the same oracle over the same ingested bytes already localizes any disagreement to a provider. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…Noz#12385) - Embed statementbuilder.Config into querier.Config with mapstructure ",squash"; keys move to querier.skip_resource_fingerprint.* (env SIGNOZ_QUERIER_*). - Drop the standalone statementbuilder section and its config factory. - Pass statementbuilder.Config wholesale into NewLogQueryStatementBuilder.
… metric (SigNoz#12313) * chore: delta temporality metrics should always be considered as a sum metric * test: add integration tests * test: parametrise the non-reduced test for rate and increase both * test: add comment explaining last samples in test * chore: remove unneeded comments --------- Co-authored-by: Srikanth Chekuri <srikanth.chekuri92@gmail.com>
* fix: enforace the required tag on dashboard spec fields * test: add empty list and objects for required fields in integration tests
- Replace the search() stub: the logs condition builder fans a case-insensitive
match of the term across every searchable column (log columns, body/body_v2,
attribute + resource maps).
- Grammar: searchCall takes a valueList (parser regenerated), so the scoped form
search('term', body, resource) parses and narrows the fan-out to those contexts.
- The visitor emits FilterOperatorSearch and flags the statement scan-heavy; the
statement builder attaches a CostGuard the querier enforces via EXPLAIN ESTIMATE
against a per-shard budget, cumulative across buckets on the window-list path.
- body_v2 gets its own, lower budget (search_max_scan_rows_json_body): toString()
rebuilds every document and no skip index prunes.
- Logs-only: other signals reject search().
- Unit tests for the per-context fan-out SQL and both budgets; integration suites
run the same matrix over the legacy body and over body_v2.
* fix(alertmanager): resolve email SMTP settings from env * refactor(alertmanager): remove dead legacy config and tidy integration test helpers * test(alertmanager): scope smtp rotation test to env rotation only * chore: drop redundant test * chore: log and warn; do not persist global config
SigNoz#12384) getCaretContext resolved the active slot to a token's start index while keeping replaceEnd at the caret, so a caret parked in the whitespace before that token produced an inverted range (replaceStart > replaceEnd). dslCompletionSource passes that range to CodeMirror as CompletionResult from/to, and accepting a suggestion threw RangeError: Invalid change range 16 to 15 (in doc of length 36) inside view.dispatch. The same inverted range made spliceAtCaret duplicate the skipped character. The replaced range is defined as ending at the caret, so clamp it there and let the slot collapse to a plain insertion point.
…right place & move events out of constants (SigNoz#12359) * fix(pods): redirecting to the wrong doc for chart * refactor(events): move out of src/constants * test(use-log): fix issue with mock * docs(infra-monitoring): fix more broken doc path * fix(pods): disable sort for pod restarts
* chore: add api to retry migration for a dashboard * feat(dashboard): add retry migration action for legacy dashboards The legacy-dashboard dialog only offered the dashboard ID and a link to support. Now that the v1->v2 migration can be re-run on demand, let an editor trigger it from there and fall back to support only if it still fails. Retrying needs edit access (the endpoint is EDITOR-gated), so viewers keep the ID-and-support dialog unchanged. --------- Co-authored-by: Ashwin Bhatkal <ashwin96@gmail.com>
* Revert "chore: added infraMonitoringV2 feature flag (SigNoz#12031)" This reverts commit 1d441b6. * chore(infrastructure-monitoring-v2): remove feature flag
…gNoz#12382) A numeric coercion yields seconds since epoch, so max(timestamp) came back as 1758113657.04 and the third-party APIs "Last Seen" rendered as January 1970. Both rules key on the physical column type, since no FieldDataType denotes a timestamp. - a time column is never coerced: it reaches the aggregate in its native type and the driver returns a time.Time - a bare `timestamp` resolves to the intrinsic column alone; a same-named numeric attribute no longer joins the candidate union, where the mixed branches failed with "no supertype for DateTime64, Float64" - the exists guard leaves the String bucket: `timestamp <> ''` becomes a typed epoch-zero comparison, verified equivalent on ClickHouse 25.5 including a non-UTC server timezone - max/min/quantile/count keep working; sum/avg and the rate_* variants now fail at the database instead of returning a rescaled number, pinned by integration tests until they are rejected up front - tests/fixtures/traces.py wrote kind_string as the enum member name ("SPAN_KIND_CLIENT"), so any feature filtering on it matched nothing; it now maps the six kinds to the exporter's form ("Client") - Last Seen assertions pin the encoding and the exact instant, rather than accepting either an RFC 3339 string or epoch millis - drop the /overview/domain integration tests and the fixture surface that served only that route; the endpoint is no longer used
Add metadata endpoint checks for openstack (covers OVHcloud), oracle, alibaba, akamai (linode), ibm, and scaleway, and the FLY_APP_NAME environment variable for fly.io. OpenStack must come before AWS since it exposes an EC2-compatible endpoint.
…2397) * fix(version): detect docker under private cgroup namespaces Docker defaults containers to a private cgroup namespace on cgroup v2 hosts (20.10+), which remaps /proc/self/cgroup to "0::/" and defeats the runtime-name sniff, so compose installs report deployment.mode=unknown. Check the container manager marker files (/run/.containerenv, /.dockerenv) and the container env var first, mirroring systemd's detect_container, and fall back to mountinfo when the cgroup path is namespaced away. * fix(version): drop the docker case from the container env check docker never sets the container variable; /.dockerenv already covers it
* feat: support llm trace list and span list * fix: take perf into consideration * fix: more tests * fix: more cleanup * fix: cleanup and more tests * fix: add resource fingerprint cte * fix: edge cases and correct cost key * fix: update integration test * fix: update openapi * fix: address comments * fix: address comments * fix: fix tests * fix: add back the flag in metadata * fix: remove source and change to builder ai query * fix: updated openapi * fix: minor cleanup * fix: refactor as requested * fix: address comments * fix: address comments * fix: remove tracefield. explicit rejection * fix: remove comment * fix: remove accidentally added file * fix: add NewFactory * fix: refactor integration tests * fix: address comment * fix: use assert * fix: use assert and condense comments
…sh (SigNoz#12402) * fix(k8s-expanded-row): ensure values are correctly aligned Ref: https://app.notion.com/p/signoz/Alignment-in-rows-inside-a-group-wrt-column-headers-is-not-correct-3b1fcc6bcd1980dd910bff84d1defe47?v=65cfcc6bcd1982a6bee688d8fd55420c&source=copy_link * fix(pods): ensure table aligment on group by Ref: https://app.notion.com/p/signoz/Alignment-in-rows-inside-a-group-wrt-column-headers-is-not-correct-3b1fcc6bcd1980dd910bff84d1defe47?v=65cfcc6bcd1982a6bee688d8fd55420c&source=copy_link * fix(infra-monitoring): not allowing name columns to be sorted Ref: https://app.notion.com/p/signoz/No-option-to-order-resources-by-name-3b1fcc6bcd198019a729d2575731387e?v=65cfcc6bcd1982a6bee688d8fd55420c&source=copy_link * fix(infra-monitoring): cannot sort name columns when group by is enabled Ref: https://app.notion.com/p/signoz/Cannot-use-sort-when-group-by-is-added-3b2fcc6bcd19806d9514d27e4da26c35?v=65cfcc6bcd1982a6bee688d8fd55420c&source=copy_link * fix(infra-monitoring): sort key was not correct Ref: https://app.notion.com/p/signoz/Review-all-sortable-columns-3b2fcc6bcd1980c4b267f3e35d4e875f?v=65cfcc6bcd1982a6bee688d8fd55420c&source=copy_link * feat(infra-monitoring): add descriptions for each threshold of entity progress Ref: https://app.notion.com/p/signoz/For-usage-metrics-the-info-icon-can-explain-the-colours-right-there-instead-of-just-saying-to-go-to-3b1fcc6bcd1980fc81a4f2b8179541b9?v=65cfcc6bcd1982a6bee688d8fd55420c&source=copy_link * fix(infra-monitoring): explorer button not allowing the back button Ref: https://app.notion.com/p/signoz/After-click-to-go-to-metrics-explorer-on-chart-we-cannot-go-back-3b1fcc6bcd1980d7b52fd7dbe0b3e29f?v=65cfcc6bcd1982a6bee688d8fd55420c&source=copy_link * fix(infra-monitoring): polluted navigation history Fixes the following issue: - Go to namespaces - Select “generator” and open the details - Close the details - Go to pods - Group by namespaces Expand “generator”, this will cause it to be redirected to namespaces category Ref: https://app.notion.com/p/signoz/wrong-redirect-when-navigating-between-categories-3b2fcc6bcd1980e996fedd639ce72b01?v=65cfcc6bcd1982a6bee688d8fd55420c&source=copy_link * fix(infra-monitoring): get min max not detecting 1month as valid interval Ref: https://app.notion.com/p/signoz/1month-is-not-supported-to-be-received-as-relativeTime-on-metrics-explorer-3b2fcc6bcd1980c0bd17f1e3d869e987?v=65cfcc6bcd1982a6bee688d8fd55420c&source=copy_link * refactor(infra-monitoring): update table statistics for better auto page size Ref: https://app.notion.com/p/signoz/change-the-baseline-values-for-calculated-page-size-3b2fcc6bcd19807caafbdedb725b6ab7?v=65cfcc6bcd1982a6bee688d8fd55420c&source=copy_link
* fix(dashboard): count panel stats from the v2 spec The panel counters walked a top-level `widgets` array and returned early when the key was missing. A v2 dashboard stores only `metadata` and `spec`, with panels as a map under `spec.panels`, so every v2 row hit that early return and all `dashboard.panels.*` stats stayed at zero. `dashboard.count` was unaffected — it is a row count. Read the v2 spec instead: count the panels under `spec.panels` and take each panel's signal from its query envelope, reusing the typed read path and `QueryEnvelope.GetSignal`. Signal-less queries (promql, clickhouse sql, formulas) count towards the panel total only. v1 rows are no longer parsed for panel stats and contribute to `dashboard.count` alone. * test(dashboard): drop the constant name arg from the stats query helper statsBuilderQuery only ever received "A", which go-lint flags via unparam. A panel holds a single query, so the name never mattered to the assertions; the composite test still names its sub-queries through statsBuilderQuerySpec. * refactor(dashboard): move v2 panel stats to a perses_ file All v2 code lives in perses_-prefixed files until the v1 code goes away. Pure move of the stats block out of dashboard.go, tests alongside it. * fix(dashboard): count create-v2 stats off the postable spec CreateV2 already holds the postable dashboard, so decoding the storable back into a v2 dashboard just to count its panels was a needless type conversion on the create path. Split the panel walk into addPanelStats over a DashboardSpec, and add NewStatsFromPostableDashboardV2 for the create path. The storable variant keeps its signature for the periodic collectors, which only have rows.
* feat(alert-channels): add google chat channel type, defaults and url validation Prefills the title and description templates verbatim from the backend's DefaultGoogleChatReceiverConfig, and extracts the slack templates into SlackInitialConfig so the form can restore them when the type changes. * feat(alert-channels): add google chat option and form fields * feat(alert-channels): wire google chat create, edit and test calls * test(alert-channels): cover the google chat channel form * fix(googlechat): use standard markdown in default templates for cardsV2
* fix: remove if condition for json parser * fix: update integration tests
…2415) * refactor(metrics-explorer): move use page size out of infra monitoring * refactor(query-builder): update code to reference v2 instead of v1 * refactor(infra-monitoring): delete v1 code
…metric-reduction-feature` path (SigNoz#12413) * chore: using local tables in place of distributed tables in infra-monitoring reduced metric path * test: infra-monitoring list endpoints on the metrics reduction path * chore: address review comments
…Noz#12305) Bumps [github.com/google/cel-go](https://github.com/google/cel-go) from 0.28.0 to 0.29.0. - [Release notes](https://github.com/google/cel-go/releases) - [Commits](cel-expr/cel-go@v0.28.0...v0.29.0) --- updated-dependencies: - dependency-name: github.com/google/cel-go dependency-version: 0.29.0 dependency-type: indirect ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
…z#12222) Bumps [google.golang.org/grpc](https://github.com/grpc/grpc-go) from 1.80.0 to 1.82.1. - [Release notes](https://github.com/grpc/grpc-go/releases) - [Commits](grpc/grpc-go@v1.80.0...v1.82.1) --- updated-dependencies: - dependency-name: google.golang.org/grpc dependency-version: 1.82.1 dependency-type: indirect ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* feat: enable FGA for dashboards and their public config * test: add integration test for dashboard FGA * fix: fix permissions for public dashboards, pinning, views * fix: allow viewers to manage views * fix: remove edits to the public dashboard line
…, Auth0, and PgBouncer datasources (SigNoz#12098) * feat(onboarding): add HCP Vault, OpenTelemetry eBPF, Langflow, Cohere, and Auth0 datasources Add UI onboarding configurations (logos + datasource entries) so these tools appear in the Add Data Source onboarding flow. SVG logos optimized with svgo. Closes signoz.io#3614 (HCP Vault) Closes signoz.io#3626 (OpenTelemetry eBPF / OBI) Closes signoz.io#3680 (Langflow) Closes signoz.io#3702 (Cohere) Closes signoz.io#3562 (Auth0) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(onboarding): add PgBouncer datasource Add PgBouncer (PostgreSQL connection pooler) metrics onboarding entry, reusing the existing PostgreSQL logo. Closes signoz.io#3553 (PgBouncer) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore(onboarding): update SVG logos and category of datasources --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: Jugal Kishore <jugal@signoz.io>
…WebUI, DSPy datasources (SigNoz#12363)
…2440) A tab that outlives a deploy requests hashed assets the new build no longer has. `lazyRetry` already recovers from this by reloading once, so the resulting errors are noise — they spike on every deploy and each one burns a Session Replay (`replaysOnErrorSampleRate: 1.0`). ### Sentry `ignoreErrors` Filters the whole class. Four patterns because the same failure is worded differently per source: | Pattern | Source | |---|---| | `Unable to preload CSS for` | Vite's own thrown `Error`, identical everywhere | | `Failed to fetch dynamically imported module` | Chromium | | `error loading dynamically imported module` | Firefox | | `Importing a module script failed` | Safari | `ignoreErrors` is applied as an event processor (`@sentry/core` `eventFilters.js`), so the event is dropped before transport. Replay's error flush hooks `afterSendEvent`, which never fires for a dropped event — so this stops the replay burn too, not just the issue count. Trade-off, stated plainly: stale-asset failures now produce no Sentry signal at all, including the case where the reload doesn't fix it. A genuinely broken deploy has to be caught from asset 404 rates rather than from Sentry. ### `lazyRetry` Behaviour is unchanged. One guard added: `setSessionStorageApi` returns `false` when sessionStorage is blocked (iframe, storage disabled), and the retry flag can't persist. The reload was previously issued anyway, so every failed import reloaded forever with no way out. It now reloads only when the flag was actually written.
…tatements (SigNoz#12325) > **Stack** (review in order; each PR's diff is against its predecessor): > 1. SigNoz#12323 `v2-read-path` — v2 native read path (leaf package) > 2. SigNoz#12324 `v2-wiring` — wiring, shadow/pin rollout machinery, dual-leg conformance > 3. SigNoz#12325 `v2-transpiler` — PromQL→ClickHouse transpiler + classification golden > 4. SigNoz#12093 `issue-4293` — the /prometheus API move (breaking slice, last) ### What The performance half of the v2 provider: an allowlist compiler (`classify`/`rewrite`) that evaluates proven PromQL shapes entirely inside ClickHouse on the `timeSeries*ToGrid` aggregate functions (CH ≥ 25.6), so one row per output series comes back instead of every raw sample. Everything not provably equivalent falls back to the engine over the PR-1 querier; a transpilable subtree under a non-transpilable node runs hybrid (subtree materialized as synthetic series, engine on top). `TryExecuteRange` slots into the PR-2 serve/shadow paths (until now engine-only) through the new `prometheus.RangeExecutor` capability interface — the provider stays unexported and pkg/querier keeps holding `prometheus.Prometheus`; providers without the capability (v1) simply never transpile. The capability folds into the main interface once v1 is removed. Highlights (docs/contributing/prometheus.md carries the full correctness story): - Range functions map to verified grid aggregates; `increase` is `rate × range` exactly (same extrapolated delta, factor algebra). - Instant selectors reproduce stale-marker shadowing with a three-aggregate compare — skipping stale rows in WHERE would resurrect the sample the marker buried. - `*_over_time` at range = k·step aggregates whole step buckets (`groupArrayInsertAt` + slide) — no per-window fan-out, no prefix-sum differencing. - **Window-sliver filtering** (the headline perf commit, folded here): when the window is narrower than the step, only window/step of the timeline can influence any grid point; a lattice predicate in WHERE cuts the aggregate's input by the coverage ratio — measured 74s/28GiB → 16s/4.3GiB on a 36k-series 1w rate, and a 2.67B-sample case that exceeded 150GiB now completes in 19s/17GiB. Over sliver-filtered rows the last-style gates lift (instant selectors and `last_over_time` transpile at window < step), and disjoint-window `*_over_time` forms drop the divisibility gate. - Scalar-op pipelines apply in Go, slot by slot — same float64 ops, same order the AST dictates. Two guards land with it: - **Classification golden** (`classification_golden_test.go` + `testdata/classification_golden.json`): freezes the route (full/hybrid(n)/fallback + reason) of every conformance-corpus expression, one line each — 317 expressions: 132 full, 39 hybrid, 146 fallback. The test also requires each expression to route the same on every corpus grid; if a classifier change ever makes the route grid-dependent, the test fails and the key must grow. Routing is its own correctness surface — silently falling back costs the pushdown, silently transpiling an unproven shape risks wrong numbers; both now show up in review as a golden diff, with the corpus suite's v2 leg judging the numbers. - **Workload coverage reporter** (`TestClassifyCorpus`, env-gated): classifies a JSON-lines corpus of real dashboard/alert queries and buckets fallbacks by reason, to steer future allowlist work. **What the dual-leg suite caught on its first transpiled run** (evidence the PR-2 guard works, worth stating in review): - The classifier read a duration expression's offset (`x offset step()`) as zero and transpiled it — offset expressions parse *without* the experimental-parser flag, so they reach production. 20 corpus cases served silently wrong numbers. Fixed by refusing `OriginalOffsetExpr` / `RangeExpr` / `StepExpr` at classification (engine evaluates them exactly); regression cases added, golden regenerated (30 routings flipped to fallback). - Name-drop assembly treated temporally-disjoint same-labelset twins as separate series: `-{job="api"}` spanning `http_requests`/`http_errors` returned a 400 the engine would not raise, and hybrid `-metric_a or -metric_b` returned duplicate `{}` series. The engine's actual rule is: assemble the matrix by labelset, merging elements that never share an evaluation timestamp; error only on a same-timestamp conflict. Both the full-plan path (`mergeSameLabelsetSeries`) and the hybrid post-strip path (`mergeMatrixByLabelset`) now reproduce it, with unit tests pinning the corpus scenarios. - 12 remaining divergences, all one class, recorded in `known_divergences_v2.json` with causes: the engine aggregates with Kahan compensated summation (sum, sum_over_time) and an overflow-free incremental mean (avg); ClickHouse's `sumForEach`/`avgForEach`/`arraySum` are naive, so ±1e100 cancellation returns 0/residue and near-max-float64 `avg` overflows to ±Inf. Burn-down note: `sumKahanForEach` for the cancellation class; the overflow class needs an incremental-mean aggregate ClickHouse doesn't have. ### Alternatives considered and discarded - **General PromQL→SQL translation.** An allowlist inverts the failure mode: an overlooked construct becomes a fallback instead of a wrong number. Every shape on the list was validated slot-for-slot against the vendored engine on live data before entering it. - **ClickHouse's own PromQL dialect** (ClickHouse#57545, `dialect='promql'`). Emits the same grid functions, but currently covers only rate/irate/delta/idelta/last_over_time, has no fallback engine, and ties us to their TimeSeries table engine. We use the same primitives with our own classifier and our own exactness gates. - **Prefix-sum differencing for `*_over_time` windows.** Large-minus-large cancellation drifts past the shadow tolerance on counter-sized values; direct per-slot combination of at most W bucket partials adds the way the engine adds. - **Fanning each sample into every window that covers it.** Multiplies rows by W — billions of rows for a long range over a short step; the bucketed form's row count is series × buckets, the size of the output. - **Handling staleness by filtering stale rows in WHERE (instant units).** Resurrects the older real sample the marker was written to bury; hence the last-overall vs last-non-stale timestamp comparison. - **Transpiling @-modifier and default-resolution subqueries.** Their evaluation grid depends on server runtime settings the transpiler cannot see; they stay on the (exact) engine path. ### Test plan - `go test ./pkg/prometheus/clickhouseprometheusv2` — transpiler unit tests (SQL forms, classification, scalar ops, subquery grids), golden. - `pytest integration/tests/promqlconformance/` — the v2 leg now exercises transpiled serving for every routable corpus case; ledger unchanged (empty). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Fable 5 <noreply@anthropic.com> Co-authored-by: Pandey <vibhupandey28@gmail.com>
Bumps `clickhouse-sql-parser` to v0.5.5, fixes the false rejection that was left over once it landed, and closes three holes in the same validator that the first two changes brought to light. ## The bump **Reserved keywords as expression operands** ([SigNoz#305](AfterShip/clickhouse-sql-parser#305)). `interval` was fixed in v0.5.4, but the same defect affected 36 other keywords once the column appeared as an operand rather than bare. Sweeping 94 candidates against ClickHouse 26.8.1.337, only `on` still rejects — and ClickHouse runs that too. This one was live: `sum(limit)` on a metric label. **Panic on an unparseable `DEFAULT` expression** ([SigNoz#306](AfterShip/clickhouse-sql-parser#306)). Both known cases return a parse error now instead of dereferencing nil. The `recover` in `ErrIfStatementIsNotValid` stays — it guards the next one of these, not these two. [SigNoz#307](AfterShip/clickhouse-sql-parser#307) also allows `CAST` in a table function's argument list. ## Table functions are only table functions in a table position The parser types a call inside a table function's argument list as a `TableFunctionExpr` as well, so the generator allow list only ever cleared a generator whose argument was a literal. Every real dashboard computes its row count — `numbers(greatest(1, intDiv(end_ns - start_ns, step_ns) + 1))` — and every one was refused, on `intDiv` rather than on `numbers`. `TableExpr.Expr` is the only table position a SELECT can reach, so the allow list asks that instead. Of the four places the parser builds a `TableFunctionExpr`, two are `CREATE TABLE` paths rejected as not-a-SELECT before the walk starts, one is `parseTableArgPrimaryExpr`, and one is the `FROM`/`JOIN` path that wraps into a `TableExpr`. ## Three holes that were already open Skipping argument position is only safe if nothing there can read, and that turned out not to be true — not because of this change, but independently of it. **Reading functions.** `file` is both a table function and a scalar function, and the validator never inspected scalar calls at all. On `main` today, `SELECT file('/etc/passwd')` is accepted and returns the file. A numeric wrapper passes ClickHouse's type check, so the row count alone is an oracle: `numbers(length(file(x)))` yields one row per byte. The same applies to the 42 dictionary accessors, which can be backed by HTTP, ODBC or another database, to `catboostEvaluate`, and to the introspection functions. All are now refused by name wherever they appear, under `clickhouse_sql_reading_function`. **`x IN db.table`.** ClickHouse reads this as `x IN (SELECT * FROM db.table)`, and a qualified name on the right of `IN` parses as a `Path`, not a `TableIdentifier` — so `SELECT * FROM t WHERE a IN system.users` bypassed the internal-database rule entirely. Now checked, including the `GLOBAL IN` and `NOT IN` forms. **Quoted generator names.** The allow list matched on the formatted name, which carries the quoting, so ``SELECT * FROM `numbers`(31)`` was refused. It now reads the identifier the way the internal-database branch already did. ## Effect Replaying 72 distinct shapes of production `clickhouse_sql` that the validator currently rejects: **64 pass, up from 59 on v0.5.4**. Two came from the bump, three from the table-position change, and those three are 379 of the 1390 sampled occurrences. The three new rules add no false positives to the corpus. Of the eight left, four are correct rejections (`system` reads, `SHOW TABLES`), one is a dashboard variable rendering as the literal `<no value>`, one is SQL ClickHouse also rejects, and two are an open upstream gap. ## Tests `TestErrIfStatementIsNotValid_ShouldPassButFails` is back, holding what remains: three forms of a parenthesised left operand of a set operator, and `on` as a column name. It also stopped panicking — `errors.Asc` dereferences the error it is given, so a case starting to pass took the suite out with a SIGSEGV instead of reporting. Both refusal tables now share one harness, bounded by the same timeout the passing table uses. Known gap: no input is currently known to panic the parser, so the `recover` has no test exercising it.
#### Description - Replace the multi-section PR template (change type, risk assessment, changelog, checklist) with four concise headings: Description, Issues closed, Screenshots, Additional Information. - Add `.claude/rules/` with agent rules for comments (repo-wide, Go, Python) and pull requests. - Ignore `.dev/` and `.claude/worktrees/` in `.gitignore`.
#### Description - Remove `.github/workflows/docs.yml`, which labeled `feat:` PRs with "docs required" and failed the check until "docs shipped" was added. - The job is not a required status check on `main`, and nothing else references the workflow or its labels.
…z#12342) ## Summary - Saved views now persist a versioned, typed spec (`schemaVersion` + `spec{compositeQuery, selectedFields, display}`) instead of a bare composite-query blob plus an opaque, frontend-owned `extraData` string -- mirroring the pattern dashboards already use for their v2/perses schema. - `/api/v1/explorer/views` keeps working exactly as before: a thin conversion layer translates to/from the legacy wire format, including folding `extraData`'s ad hoc JSON into the typed spec and back for backward compatibility. - A one-time migration rewrites existing rows into the new shape and drops the now-unused `extra_data`/`category`/`tags` columns. ### Scaffolding decisions - Using v2 for new handlers instead of renaming old handlers to something else for these reasons - keep the diff minimum for easier reviews, avoiding any git history or last updated at change in old route registration. - Keeping the conversion to old saved view type in handler itself rather than `savedviewtypes` package to keep it un-exported and not let them be available anywhere else to be used. It also enables `savedviewtypes` to be independent on query-service models. - Modified the existing handler and it's interface to include the v2 methods instead of adding another handlerV2 since apiserver already had handler wired in, so don't want to pass on 2 version simultaneously. ### Breaking change - Any unknown key in the `ExtraData` will be rejected and dropped silently in the old APIs and give error in new version. - If there was any way to add tag or category in saved view earlier, that data will be lost. - Old APIs will not support the old QB request payload, only v5 format is supported. --- Closes SigNoz/engineering-pod#4651 Alternative discarded SigNoz#12208 --------- Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
## Summary Adds the semantic-convention evolution foundation: - vendors the OpenTelemetry schema and SigNoz overlay - generates Go and TypeScript family tables deterministically - exposes the Go resolver API for family members, current names, and historical names - adds generation checks and unit tests Related to SigNoz#6143. ## Stack 1. SigNoz#12441 — Foundations (base: main) 2. SigNoz#12442 — Phase 1 query (base: SigNoz#12441) 3. SigNoz#12443 — Phase 1 services (base: SigNoz#12442) 4. SigNoz#12444 — Phase 1 closure gate (base: SigNoz#12443) 5. SigNoz#12445 — Phase 2 signals (base: SigNoz#12444) 6. SigNoz#12446 — Phase 3 migration UX (base: SigNoz#12445) 7. SigNoz#12447 — Phase 4 rollout (base: SigNoz#12446) **Current layer:** SigNoz#12441 ## Testing - `go test ./scripts/semconv` - `go test ./pkg/types/telemetrytypes/semconv` - `make semconv-check` ## Risk and rollback This layer is additive apart from CI generation checks. Roll back by reverting this PR; no stored telemetry is changed.
…#12456) ## Summary Adds an integration-test matrix that pins the deliberate keyless-row contract for filter operators, independent of any feature work: - **Negative operators are a set complement over all rows.** A row that does not carry the key at all must match `!=`, `NOT IN`, `NOT LIKE`, and `NOT CONTAINS`. Users opt into presence explicitly with `AND key EXISTS`. - **Positive operators carry an implicit existence guard** (`FilterOperator.AddDefaultExistsFilter`), so keyless rows never false-positive against sentinel defaults. - **`EXISTS` / `NOT EXISTS` partition rows exactly** by key presence, and `!= x AND EXISTS` is the documented composition for "present and not x". - **Numeric attributes inherit the map-default sentinel**: a missing key reads as `0`, so `num != 5` includes keyless rows while `num != 0` excludes them. This conflation is deliberate and pinned by name (`numeric_neq_zero_sentinel_conflation`) as the reference point for any value-expression change. Coverage: 46 cases — one shared matrix over traces and logs (resource and attribute contexts), metric labels (series without the label), the numeric sentinel, and the EXISTS composition. The contract, matrix, seed data, and assertions live together in one file so the contract reads top to bottom. ## Why These semantics were enforced only implicitly by the operator list in `AddDefaultExistsFilter`, with no test naming the intent. That gap allows an implementation change to alter negative-filter results silently and lets new tests calibrate expectations against the implementation instead of the contract. The attribute names used here are deliberately outside every semantic-convention family, so this file pins the base contract regardless of the semconv overlay state and serves as the oracle that family-field behavior (`queriertraces/13_semconv_evolution.py` in the semconv stack) must mirror. ## Testing - `uv run pytest --basetemp=./tmp/ --reuse integration/tests/queriercommon/06_keyless_semantics.py` — 46/46 passed against a stack built from main-based sources, and 2×46 against a long-lived shared stack, confirming the set-based assertions stay stable under environment reuse. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…sts (SigNoz#12460) #### Description - Add `.claude/rules/pytest.md` with conventions for the Python integration suite. The lead rule: **no `_`-prefixed helper functions in test modules** — inline the logic; repetition across tests is cheaper than indirection; genuinely shared machinery becomes a fixture. Fixtures live in `tests/fixtures/` only, never under `integration/tests/` — with one exception: SigNoz-level fixtures (a suite spinning up SigNoz with different envs via `create_signoz`/`create_migrator`) always belong in that suite's `conftest.py`. Plus: fixture-factory over indirect parametrization, skip at collection, config via explicit `--flags`, snake_case parametrize ids, and collection gotchas (`python_files` prefix matching, `--import-mode=importlib`). - Apply the no-`_helper` rule across `tests/integration`: all 40 module-level `_` helpers eliminated in dashboard, inframonitoring, promqlconformance, querier_json_body, querierlogs, queriermetrics, and queriertraces. Pure transforms and request wrappers were inlined at their call sites; case-table verifier callables became data flags with inline branches; the two 115-line resource-evolution mega-helpers folded into parametrized tests; shared machinery moved to `tests/fixtures/` as fixture-factories (`wipe_all_dashboards`, `load_pods_metrics`, `run_query_case`) registered via `pytest_plugins`. - Apply the py-comments rule across `tests/`: drop module docstrings that restate the filename, relocate the ones carrying real constraints next to the code they constrain, delete function/class docstrings that restate the identifier, trim narrated steps and Args/Returns boilerplate. - Fix camelCase parametrize ids in `queriermetrics/01_fill.py` (`fillGaps`/`fillZero` → `fill_gaps`/`fill_zero`). #### Additional Information - All 629 tests in the touched suites collect cleanly; `py-fmt`/`py-lint`/compileall pass. Runtime verification against the docker stack was not run.
…igNoz#12462) #### Description - Add the fixture-vs-function rule to `.claude/rules/pytest.md`: a fixture earns its indirection only by owning setup/teardown (`yield` + cleanup) or provisioning a resource; a stateless action or lookup is a plain importable function in the matching `tests/fixtures/` module taking `signoz`/`token` as ordinary arguments. - Apply it to the three fixture-factories introduced in SigNoz#12460 that have no lifecycle: `delete_all_dashboards` (renamed from `wipe_all_dashboards`) and `run_query_case` are now plain functions, their modules deregistered from `pytest_plugins`, and all call sites updated. - Generalize `Metrics.load_from_file` with a `label_substitutions` parameter (placeholder rewriting, e.g. `__START_TIME__` → runtime ISO string) and drop the bespoke `load_pods_metrics`, which duplicated the base-time rebase logic — `02_pods.py` now loads JSONL the same way as every other inframonitoring suite file. Follow-up promised in SigNoz#12460 (comment).
#### Description - `uv lock --upgrade` across the tests project: pytest 9.0.3→9.1.1, ruff 0.15.11→0.16.2, selenium 4.43→4.46, numpy 2.4.4→2.5.1, uvicorn 0.46→0.52.1, testcontainers 4.14.2→4.15.0, requests, sqlalchemy, websockets, and the rest of the transitive set (zstandard dropped as no longer required). - Ignore `PLR0917` (too-many-positional-arguments), newly enforced by ruff 0.16 — muted alongside the other `PLR09xx` complexity rules the project already ignores (193 pre-existing hits, all in test/fixture signatures). #### Additional Information - `py-fmt` (no reformats), `py-lint` (clean), and full integration-test collection (1782 tests) pass on the upgraded toolchain. Runtime verification against the docker stack was not run.
… the idp after keycloak login (SigNoz#12399) Deflakes the SSO login tests (`callbackauthn` and `basepath`). They all share the `idp_login` fixture, and after it clicked Keycloak's login button it could hand control back to the test too early — in two different ways ([example CI failure](https://github.com/SigNoz/signoz/actions/runs/30898802502/job/91957986941)). ### What The fixture used to wait for the login button to disappear and treat that as "login is done". Two things go wrong with that: 1. **The page can vanish while we're looking at it.** Asking "is the button still visible?" takes two round-trips to the browser: find `kc-login`, then ask whether it's displayed. If Keycloak's redirect lands between the two, the second call is asking about a node that no longer exists. Selenium normally recognises that as a stale element and quietly retries — but Keycloak → SigNoz is a *same-site* hop (`localhost` → `localhost`), where the renderer survives the swap and that detection can miss. The raw chromedriver error (`Node with given id does not belong to the document`) then escapes and fails the test. That's the CI failure above. 2. **The button disappearing doesn't mean login finished.** It only means we left the login *page*. In the SAML flow Keycloak next serves a small auto-submitting page — still on the IdP — and *that* POST is what actually creates the user in SigNoz. So a test could go looking for the user before SigNoz had ever seen the callback, and fail with `User ... not found`. Reproduces locally on `test_idp_initiated_saml_authn`. The wait now checks what the tests actually need: **the browser has left the IdP host** (the hostname in the URL changed) *and* the login button is gone. ### Guardrails - Nothing is held across the navigation — the button is looked up fresh on every poll with `find_elements`, so "gone" is simply an empty list, never a question asked of a dying node. - Any browser error during a poll is read as "still navigating, try again" instead of failing the wait. - The wait sits through everything that's still on Keycloak (the `login-actions` hops, the SAML interstitial) and passes only once SigNoz has handled the callback and redirected — so the user exists by the time the test asserts on it. - Wrong credentials still fail loudly: Keycloak re-renders the login form on its own host, so the wait times out exactly as before. One change, in the shared fixture — the OIDC and SAML flows in both `callbackauthn` and `basepath` all go through it. ### Notes - Failure 1 needs the redirect to land inside a ~2–5 ms window of a poll that only runs every 500 ms, so it's effectively a loaded-CI-runner lottery — which is why it's rare and CI-only. Failure 2 shows up locally. - Unrelated to the PR it fired on (SigNoz#12382, query-builder only); the identical SAML test passed in the same run. ### Testing - Reproduced failure 1 outside pytest, with a probe driving real headless **Chrome for Testing 151.0.7922.71** (the exact build from the CI log) through a click → POST → same-site redirect that mimics the Keycloak login flow, with server think-time near the 500 ms poll boundary. Both waits ran verbatim, at their real polling rate: | Post-click wait | Logins | Failures | |---|---|---| | old (`EC.invisibility_of_element`) | 400 | **3 × the exact CI inspector error** | | new (left-the-IdP check) | 400 | **0** | - Cross-checked the mechanism against the selenium 4.40 source with a stubbed driver: the detached-node error does escape the old wait (it only catches stale/not-found), while the new one absorbs it and passes on the next poll. A bad-credentials control times out on both old and new, so failure detection isn't weakened. - Ran the full suites locally on the final fixture, with a fresh sqlite + wal store per suite (matching the failing CI leg): `basepath` 6/6, and all 36 SSO/domain tests in `callbackauthn` — including `test_idp_initiated_saml_authn`, which flaked with `User not found` on the old wait in the same setup. (The one local non-pass, `test_apply_license`, is unrelated: it asserts on wiremock's request journal and the reused license-mock container is never reset between runs — CI gets a fresh mock.) - `make py-fmt` / `make py-lint` clean. Fixes SigNoz/engineering-pod#5850
…ts (SigNoz#12455) The `telemetry.*.last_observed` stats took `max()` over client-supplied event-time columns, so a single row with a skewed or corrupt timestamp (a 2050-dated log, a `2^32−1`-second span) poisoned them indefinitely. ### What - Traces/logs `last_observed` now reads `max(inserted_at)` — the collector-stamped insert time added in SigNoz/signoz-otel-collector#875; metrics reads `inserted_at_unix_milli` (metrics migration 1007). - Each signal checks `hasColumnInTable` first and falls back to the previous expression, so tenants without the schema migration keep today's behavior and switch over automatically. ### Notes - `created_at` is unusable here: pre-migration rows evaluate its `now64(3)` default at read time, so `max(created_at)` always reads as "now". - Pre-migration rows read `inserted_at` as epoch, which `max()` ignores; the all-old case lands on the existing `Unix() != 0` skip-guard. - Future-dated garbage never TTLs out (TTL is keyed on the event timestamp), which is why the old stat stayed wrong once poisoned. ### Testing - Expressions validated against `clickhouse local`, including garbage rows (`2^64−1`, `9.3e18` ns) and empty/pre-migration tables. - `go build`, `go vet`, golangci-lint clean. Fixes SigNoz/engineering-pod#5864
… 109 (SigNoz#12469) ## Summary - `restructureSavedViewSpec` (migration 109) bulk-inserts legacy `saved_views` rows into the new `saved_view` table, which has an enforced `org_id -> organizations(id)` FK (sqlite runs with `foreign_keys=ON`). - Some tenants hit `constraint failed: FOREIGN KEY constraint failed (787)` on startup because a row's `org_id` didn't match any live organization -- e.g. an org deleted before the old table had a cascading FK, or an install where `org_id` was never backfilled (`015_update_dashboards_savedviews` only backfills it when there's exactly one org). - Fix: fetch live organization IDs up front inside the same transaction, and skip (with a `WarnContext` log, counted in `skipped`) any row whose non-empty `org_id` isn't among them -- same treatment as the existing empty-`org_id` skip. Followup to SigNoz#12342 Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
… overlaps counted attrs (SigNoz#12451) ## Pull Request --- ### 📄 Summary The infra-monitoring v2 clusters/namespaces list APIs 500 with ClickHouse error 179 (`MULTIPLE_EXPRESSIONS_FOR_ALIAS`) when the request groups by an attribute that is also a counted resource attribute (e.g. clusters grouped by `k8s.node.name` or `k8s.namespace.name`, namespaces grouped by `k8s.deployment.name`). Root cause: `getPerGroupDistinctCounts` aliases each `uniqExactIf(...)` count column with the bare attr key, which collides with the groupBy column alias for the same key. Fix: alias count columns as `__count_<attr>`. Row scanning is positional and the result map is keyed in Go from `attrNames`, so nothing downstream changes. Integration tests: clusters API grouped by `k8s.namespace.name` and namespaces API grouped by `k8s.deployment.name`, with exact per-group `counts` assertions (identity-tuple semantics). Both reproduce the 500 on the pre-fix build and pass on the fixed build. #### Issues closed by this PR Closes SigNoz/pulse-pod#245 ### ✅ Change Type _Select all that apply_ - [ ] ✨ Feature - [x] 🐛 Bug fix - [ ] ♻️ Refactor - [ ] 🛠️ Infra / Tooling - [ ] 🧪 Test-only --- ### 🧪 Testing Strategy - Tests added/updated: Yes — integration (`04_namespaces.py`, `05_clusters.py`): new groupBy-on-counted-attr cases + per-group `counts` assertions on existing cases - Manual verification: Yes — replayed the failing production query with the fixed aliasing against ClickHouse - Edge cases covered: groupBy overlapping a counted attr; same-named deployment across namespaces counted as distinct entities --- ###⚠️ Risk & Impact Assessment - Blast radius: Infrastructure Monitoring — clusters & namespaces list APIs (counts query) - Potential regressions: None — SQL alias rename only; scanning is positional and result map keys are unchanged - Rollback plan: Revert this commit --- ### 📝 Changelog | Field | Value | |------|-------| | Deployment Type | Cloud / OSS / Enterprise | | Change Type | Bug Fix | | Description | Fixed a 500 error in Infrastructure Monitoring clusters/namespaces APIs when grouping by an attribute that is also part of the resource counts (e.g. node, namespace, or deployment name). | --- ### 📋 Checklist - [x] Tests added or explicitly not required - [x] Manually tested - [ ] Breaking changes documented - [x] Backward compatibility considered
## Pull Request --- ### 📄 Summary Adds a `filterByPodStatus` secondary filter to the v2 infra-monitoring list APIs (pods, nodes, namespaces, clusters, deployments, statefulsets, jobs, daemonsets). Pod status is a derived kubectl-style value (`k8s.pod.phase` + status reasons, resolved via `argMax`), not a real label, so it can't go through the normal query-builder filter. This PR resolves the full-scope status keyset up-front and intersects it with the metadata + ranked groups, keeping `total` and pagination correct. - Multi-select: the field is an array, pushed down as `WHERE lower(display_status) IN (...)` (OR within status, AND with the attribute filter). - When the optional status metrics were never ingested, the endpoint returns a non-blocking warning + empty page instead of silently filtering everything out. #### Screenshots / Screen Recordings (if applicable) N/A — backend + generated FE API types only; the UI is a separate change. #### Issues closed by this PR Part of SigNoz/engineering-pod#5778. --- ### ✅ Change Type - [x] ✨ Feature - [ ] 🐛 Bug fix - [x] ♻️ Refactor - [ ] 🛠️ Infra / Tooling - [ ] 🧪 Test-only --- ### 🐛 Bug Context N/A — not a bug fix. --- ### 🧪 Testing Strategy - Tests added/updated: - Unit test for the status push-down (`applyPodStatusFilter`, built with go-sqlbuilder). - Integration tests across all 8 entity APIs: list mode, grouped mode, validation, missing-metric warning, and multi-select union. - Manual verification: smoke-tested against staging data (single, multi, and grouped filters). - Edge cases covered: missing status metric → warning + empty; grouped mode keeps a group if ≥1 pod matches; multi-select returns the union of the selected statuses. --- ###⚠️ Risk & Impact Assessment - Blast radius: v2 infra-monitoring list endpoints only. - Potential regressions: none when the filter is unset (empty = off, fully additive). When set, an extra status query runs; it is gated behind the filter being present. - Rollback plan: revert the PR — no schema or data migrations involved. --- ### 📝 Changelog | Field | Value | |------|-------| | Deployment Type | OSS, Cloud, Enterprise | | Change Type | Feature | | Description | v2 infra-monitoring lists can now be filtered by pod status (multi-select). | --- ### 📋 Checklist - [x] Tests added or explicitly not required - [x] Manually tested - [x] Breaking changes documented - [x] Backward compatibility considered --- ## 👀 Notes for Reviewers - `filterByPodStatus` is optional and additive — no change to existing responses when omitted. - The status keyset is resolved once at full scope, then intersected — this is what keeps `total`/pagination correct despite status being a post-aggregation value. - OpenAPI spec + FE API types are regenerated (scalar → array); no hand-written FE.
…12471) <!--A few plain bullets saying what changed and why, for a reviewer skimming it - not a wall of text, not a restatement of the diff, not generated boilerplate.--> #### Description Disabling logs support for all GCP integration services. Reasons: 1. googlecloudpubsubpush receiver needs to reach Alpha stability, thus not included in contrib build -> can't suggest for logs 2. signoz otel collector includes this receiver in the build but it needs upgrade to v0.158.0 for a metrics related [fix](open-telemetry/opentelemetry-collector-contrib#49826) -> needs more testing hence can't suggest either Decision was taken to go ahead without logs for now - follow [ticket here](SigNoz/platform-pod#2901) for details. <!--Reference issues using `Closes #issue-number` to enable automatic closure on merge. --> #### Contributes to SigNoz/platform-pod#2901 <!--Anything reviewers should keep in mind while reviewing --> #### Additional Information The explanation above should be enough <!--Please delete paragraphs that you did not use before submitting.-->
…est (SigNoz#12475) #### Description - `make py-lint` is failing on `main` with `F821 Undefined name _load_pods_metrics` at `inframonitoring/02_pods.py:471`, which blocks every open PR. - `test_pods_filter_pagination_and_ordering` calls a helper that no longer exists. Replaced with `Metrics.load_from_file(get_testdata_file_path(...))`, matching the two other tests that seed `pods_phases.jsonl`. #### Additional Information - How it broke: SigNoz#12460 replaced the module-level `_load_pods_metrics` with a `load_pods_metrics` fixture, then SigNoz#12462 dropped that fixture in favour of calling `Metrics.load_from_file` directly. SigNoz#12278 branched before either landed and merged after, re-introducing one call to the long-gone helper. Git merged cleanly because the branches touched different lines, so nothing flagged it. - No behaviour change: the broken call passed no `start_time`, so no placeholder substitution happened; `load_from_file` with `label_substitutions=None` does the same earliest-to-`base_time` rebase. The replacement is byte-identical to the two sibling call sites for the same dataset.
#### Description - Removes `POST /api/v1/invite/bulk` — already deprecated, superseded by `POST /api/v1/invite`, and no callers left. - `Setter.CreateBulkInvite` stays; `CreateInvite` still delegates to it for the single-invite case. #### Issues closed by this PR Contributes to SigNoz/platform-pod#2667 #### Additional Information - OpenAPI spec and the generated frontend client are regenerated, not hand-edited. - The `integrationci / fmtlint` failure here is not from this PR — `make py-lint` is broken on `main`. Fixed separately in SigNoz#12475; this PR needs that merged (or a rebase on it) to go green. - First of three PRs splitting a v1 user-API cleanup. The other two also regenerate the spec and generated client, so whichever merges second needs the generators re-run.
#### Description
- Removes `GET`, `PUT` and `DELETE /api/v1/user/{id}` — all deprecated
and superseded by `/api/v2/users/{id}`, which the frontend already uses.
- Drops the dead code this leaves behind: the `SelfAccess` middleware
and `Claims.IsSelfAccess` (no callers left), the deprecated update
setters, and three `DeprecatedUser` helpers.
- Points the integration tests that deleted users at `DELETE
/api/v2/users/{id}`.
#### Issues closed by this PR
Contributes to SigNoz/platform-pod#2667
#### Additional Information
- Behaviour change: the removed `GET`/`PUT` were `SelfAccess`, the v2
equivalents are `AdminAccess`. Self-serve reads and updates go through
`/api/v2/users/me`, which is what the UI already calls — but worth a
second pair of eyes.
- `DELETE /api/v1/user/{id}` was the most widely reached of the three.
Please confirm nothing external (zeus) still calls it before merging.
- OpenAPI spec and the generated frontend client are regenerated, not
hand-edited.
#### Description - Invited members can now be given more than one role. The picker was single-select even though `POST /api/v2/users` has accepted a list of roles since custom roles landed. - Frontend only — nothing changed on the backend, the grant chain was already multi-role. - Onboarding analytics now emits `teamMembers[].roles` as a list, replacing the singular `role` key. #### Issues closed by this PR Closes SigNoz/platform-pod#2920 #### Screenshots / Screen Recordings #### Members Page https://github.com/user-attachments/assets/8b556549-789c-4ddf-b4af-5994eccc75f3 #### Onboarding Flow https://github.com/user-attachments/assets/a6dde5d3-e9b4-48c1-965f-352bb0a6a89f #### Additional Information - A row still requires at least one role to be considered valid. - Existing invite tests moved to `findByTitle` for role options, matching how `EditMemberDrawer` already drives the multi-select.
…ht, fix README wording
Sun-ZhenXing
force-pushed
the
feat-genai-ui
branch
from
August 9, 2026 12:53
9481abe to
0f15775
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
16 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Pull Request
📄 Summary
Screenshots / Screen Recordings (if applicable)
Issues closed by this PR
✅ Change Type
Select all that apply
🐛 Bug Context
Root Cause
Fix Strategy
🧪 Testing Strategy
📝 Changelog
📋 Checklist
👀 Notes for Reviewers