mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-22 11:54:35 +01:00
fix(ui): group failed-query Sentry reports by status and query name (#21475)
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work. > - The browser app can report errors to Sentry. A final query error goes to Sentry when it is not an outage, a denial, or a missing item. > - Every failed API request throws an `ApiError` from the same frames in `api/client.ts`. > - Sentry groups errors by stack. Thus all unrelated server failures go into one issue, titled "ApiError: Internal server error". > - The report has no route, no query name, and no status. Breadcrumbs and URLs are removed for privacy. A person cannot find which endpoint failed. > - This pull request tags each failed-query report with its HTTP status and the static name of its query key. It also sets the Sentry fingerprint from these two values. > - The benefit: each failing endpoint gets its own Sentry issue, and a person can connect it to the server error. ## Linked Issues or Issue Description **What happened?** One browser Sentry issue collected unrelated server 500 errors from different pages and environments. The events had no route or query name. A person could not connect an event to a server error. **Steps to reproduce** 1. Turn on browser error monitoring with a Sentry DSN. 2. Make two different queries fail with a 500 response, for example the issues list and the agents list. 3. Look at Sentry. Both events are in one issue with the title "ApiError: Internal server error". Neither event shows which query failed. **Expected behavior** Each failing query has its own Sentry issue. The event shows the HTTP status and the static query name. It shows no IDs, refs, or filters. **Environment** The hosted browser app, staging and production builds. I searched open pull requests for duplicates. I found no pull request that changes browser query-error grouping. ## What Changed - `ui/src/lib/browser-error-context.ts`: add `queryKindFor`. It keeps only the first query-key part, and only when it is a lowercase word list such as `issues` or `agent-commentary`. Other values become `other`, so an ID, a ref, an email, or a typed value is never sent. - `ui/src/lib/browser-error-context.ts`: add `buildQueryErrorContext`. For an error with an HTTP status, it sets the `query_kind` and `api_status` tags and the fingerprint `["api-error", status, queryKind]`. For an error without a status, it sets only the `query_kind` tag and keeps default stack grouping. - `ui/src/lib/sentry.ts`: add `captureQueryException`. It uses the same serialized queue as `captureBrowserException`. Both functions now share one private capture step. - `ui/src/lib/query-client.ts`: `reportError` now also receives the query key. - `ui/src/main.tsx`: the query client reports through `captureQueryException`. - Tests for key sanitizing, the fingerprint, the reporter argument, and the gate call. ## Verification - Red before the fix: the 6 new assertions fail on the old source (`queryKindFor`, `buildQueryErrorContext` and `captureQueryException` do not exist, and `reportError` gets no key). - Green after the fix: `cd ui && npx vitest run src/lib/sentry.test.ts src/lib/browser-error-context.test.ts src/lib/query-client.test.ts` passes 3 files and 78 tests. - The privacy test makes sure that `company-uuid` and `PAP-1` from later key parts do not go into the Sentry call. It also makes sure that UUID, ref, email, and long heads become `other`. - `ui` typecheck: no errors in the changed files. The local install was filtered to the UI package, so unrelated packages (`plugin-sdk` build output, runner dependencies) could not resolve. Required CI supplies the full typecheck. - I did not run browser e2e suites. This change does not touch them. ## Risks - Low risk. Only the grouping and tags of browser query-error reports change. The app does not change what it shows to users. - Sentry opens new issues for future query errors, one for each status and query name. The old catch-all issue gets no new events. - A query key with a dynamic first part becomes `other`. Those errors still group together, but they keep the status tag. > For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and discuss it in `#dev` before opening the PR. Feature PRs that overlap with planned core work may need to be redirected — check the roadmap first. See `CONTRIBUTING.md`. ## Model Used - Anthropic Claude Opus 5.5 (`claude-opus-5-5`), in Claude Code with tool use, repository inspection, and shell code execution. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have searched GitHub for duplicate or related PRs and linked them above - [x] I have either (a) linked existing issues with `Fixes: #` / `Closes #` / `Refs #` OR (b) described the issue in-PR following the relevant issue template - [x] I have not referenced internal/instance-local Paperclip issues or links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip` URLs) - [x] My branch name describes the change (e.g. `docs/...`, `fix/...`) and contains no internal Paperclip ticket id or instance-derived details - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [x] I have updated relevant documentation to reflect my changes - [x] I have considered and documented any risks above - [ ] All Paperclip CI gates are green - [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
1 parent
ef2a49b6df
commit
f58bbefc4e
7 files changed
+130
-13
No files matched your search
@@ -1,3 +1,5 @@
|
||||
import { errorStatus } from "@/api/errors";
|
||||
|
||||
export interface BrowserErrorDetails {
|
||||
boundary: "app" | "route";
|
||||
componentStack?: string | null;
|
||||
@@ -167,6 +169,37 @@ export function __resetPageDiagnostics() {
|
||||
lastRejection = undefined;
|
||||
}
|
||||
|
||||
const QUERY_KIND_PATTERN = /^[a-z]+(?:[-_][a-z]+)*$/;
|
||||
const MAX_QUERY_KIND_LENGTH = 64;
|
||||
|
||||
/**
|
||||
* The query key's static name, such as `issues`. Later key parts hold
|
||||
* companies, refs and filters, and are never read. A head with a digit,
|
||||
* uppercase letter or other character can be an ID or a typed value, so it
|
||||
* becomes `other`.
|
||||
*/
|
||||
export function queryKindFor(queryKey: readonly unknown[]): string {
|
||||
const head = queryKey[0];
|
||||
return typeof head === "string" && head.length <= MAX_QUERY_KIND_LENGTH && QUERY_KIND_PATTERN.test(head)
|
||||
? head
|
||||
: "other";
|
||||
}
|
||||
|
||||
/**
|
||||
* Every failed request throws from the same `ApiError` frames, so default
|
||||
* grouping puts all of them in one Sentry issue. Group an HTTP failure by its
|
||||
* status and query name instead, so each failing endpoint has its own issue.
|
||||
*/
|
||||
export function buildQueryErrorContext(error: unknown, queryKey: readonly unknown[]) {
|
||||
const queryKind = queryKindFor(queryKey);
|
||||
const status = errorStatus(error);
|
||||
if (status === null) return { tags: { query_kind: queryKind } };
|
||||
return {
|
||||
tags: { query_kind: queryKind, api_status: String(status) },
|
||||
fingerprint: ["api-error", String(status), queryKind],
|
||||
};
|
||||
}
|
||||
|
||||
/** A bounded snapshot, taken at the error rather than when the capture queue runs. */
|
||||
export function buildBrowserErrorContext(details: BrowserErrorDetails) {
|
||||
const componentStack = sanitizeComponentStack(details.componentStack);
|
||||
|
||||
Reference in new issue
Block a user