PM-5999 - fix filter by track in sys admin/challenge bulk ingestion

src/apps/admin/src/ai/topscout-rag/IndexedChallengesPanel.tsx

@@ -16,7 +16,6 @@ import {
     Button,
     IconOutline,
     InputSelect,
-    InputSelectOption,
     InputWrapper,
     Table,
     TableColumn,
@@ -32,28 +31,13 @@ import {
     IndexedChallenge,
 } from '../../lib/services/rag-index.service'
 
+import { INDEXED_TRACK_OPTIONS, INDEXED_TYPE_OPTIONS } from './ingest-options'
 import styles from './IndexedChallengesPanel.module.scss'
 
 const stopPropagation: MouseEventHandler = ev => ev.stopPropagation()
 
 const PER_PAGE = 10
 
-const TRACK_OPTIONS: InputSelectOption[] = [
-    { label: 'All tracks', value: '' },
-    { label: 'Development', value: 'Development' },
-    { label: 'Design', value: 'Design' },
-    { label: 'Data Science', value: 'Data Science' },
-    { label: 'Quality Assurance', value: 'Quality Assurance' },
-]
-
-const TYPE_OPTIONS: InputSelectOption[] = [
-    { label: 'All types', value: '' },
-    { label: 'Challenge', value: 'Challenge' },
-    { label: 'First2Finish', value: 'First2Finish' },
-    { label: 'Marathon Match', value: 'Marathon Match' },
-    { label: 'Task', value: 'Task' },
-]
-
 interface Filters {
     search: string
     projectId: string
@@ -390,15 +374,15 @@ export const IndexedChallengesPanel: FC<IndexedChallengesPanelProps> = props =>
                 <InputSelect
                     name='filterTrack'
                     label=''
-                    options={TRACK_OPTIONS}
+                    options={INDEXED_TRACK_OPTIONS}
                     value={filters.track}
                     onChange={handleFilterChange('track')}
                     tabIndex={0}
                 />
                 <InputSelect
                     name='filterType'
                     label=''
-                    options={TYPE_OPTIONS}
+                    options={INDEXED_TYPE_OPTIONS}
                     value={filters.type}
                     onChange={handleFilterChange('type')}
                     tabIndex={0}

src/apps/admin/src/ai/topscout-rag/IngestChallengesPanel.tsx

@@ -13,41 +13,18 @@ import {
     ingestChallengeInRag,
     WorkflowPollTimeoutError,
 } from '~/libs/shared'
-import { Button, InputSelect, InputSelectOption } from '~/libs/ui'
+import { Button, InputSelect } from '~/libs/ui'
 import FormToggleSwitch from '~/libs/ui/lib/components/form/form-groups/form-toggle-switch'
 
 import { Field, TextField } from './FormFields'
+import {
+    BULK_STATUS_OPTIONS,
+    BULK_TRACK_OPTIONS,
+    BULK_TYPE_OPTIONS,
+} from './ingest-options'
 import { IngestionRun, IngestionRunSummary } from './IngestionRunSummary'
 import styles from './IngestChallengesPanel.module.scss'
 
-/**
- * tc-ai-api treats tracks and types as free-form strings (its own config
- * documents them as informational, not enums), so these are conveniences for
- * the common cases — never validation.
- */
-const TRACK_OPTIONS: InputSelectOption[] = [
-    { label: 'Any track', value: '' },
-    { label: 'Development', value: 'Development' },
-    { label: 'Design', value: 'Design' },
-    { label: 'Data Science', value: 'Data Science' },
-    { label: 'Quality Assurance', value: 'Quality Assurance' },
-]
-
-const TYPE_OPTIONS: InputSelectOption[] = [
-    { label: 'Any type', value: '' },
-    { label: 'Challenge', value: 'Challenge' },
-    { label: 'First2Finish', value: 'First2Finish' },
-    { label: 'Marathon Match', value: 'Marathon Match' },
-    { label: 'Task', value: 'Task' },
-]
-
-/** Mirrors the workflow's own default status set. */
-const STATUS_OPTIONS: InputSelectOption[] = [
-    { label: 'Active + Completed', value: '' },
-    { label: 'Active only', value: 'ACTIVE' },
-    { label: 'Completed only', value: 'COMPLETED' },
-]
-
 interface BulkFormState {
     projectId: string
     track: string
@@ -221,7 +198,7 @@ export const IngestChallengesPanel: FC<IngestChallengesPanelProps> = props => {
                             <InputSelect
                                 name='track'
                                 label=''
-                                options={TRACK_OPTIONS}
+                                options={BULK_TRACK_OPTIONS}
                                 value={bulk.track}
                                 onChange={handleBulkChange('track')}
                                 disabled={bulkDisabled}
@@ -232,7 +209,7 @@ export const IngestChallengesPanel: FC<IngestChallengesPanelProps> = props => {
                             <InputSelect
                                 name='type'
                                 label=''
-                                options={TYPE_OPTIONS}
+                                options={BULK_TYPE_OPTIONS}
                                 value={bulk.type}
                                 onChange={handleBulkChange('type')}
                                 disabled={bulkDisabled}
@@ -243,7 +220,7 @@ export const IngestChallengesPanel: FC<IngestChallengesPanelProps> = props => {
                             <InputSelect
                                 name='status'
                                 label=''
-                                options={STATUS_OPTIONS}
+                                options={BULK_STATUS_OPTIONS}
                                 value={bulk.status}
                                 onChange={handleBulkChange('status')}
                                 disabled={bulkDisabled}

src/apps/admin/src/ai/topscout-rag/ingest-options.spec.ts

@@ -0,0 +1,51 @@
+/* eslint-disable import/no-extraneous-dependencies */
+import {
+    BULK_TRACK_OPTIONS,
+    INDEXED_TRACK_OPTIONS,
+} from './ingest-options'
+
+jest.mock('~/libs/ui', () => ({}), { virtual: true })
+
+describe('ingest-options', () => {
+    describe('BULK_TRACK_OPTIONS', () => {
+        // The v6 Challenges API resolves `tracks` against challengeTrack.abbreviation,
+        // and an unmatched value silently widens the filter to every track rather
+        // than erroring — so these are pinned rather than left to drift.
+        it.each([
+            ['Development', 'Dev'],
+            ['Data Science', 'DS'],
+            ['Design', 'Des'],
+            ['Quality Assurance', 'QA'],
+        ])('sends the %s abbreviation %s to the search API', (label, abbreviation) => {
+            expect(BULK_TRACK_OPTIONS.find(option => option.label === label)?.value)
+                .toBe(abbreviation)
+        })
+
+        it('keeps an unfiltered option that sends nothing', () => {
+            expect(BULK_TRACK_OPTIONS[0])
+                .toEqual({ label: 'Any track', value: '' })
+        })
+    })
+
+    describe('INDEXED_TRACK_OPTIONS', () => {
+        // These match stored chunk metadata, which holds the track's full name.
+        it.each([
+            'Development',
+            'Data Science',
+            'Design',
+            'Quality Assurance',
+        ])('filters stored metadata by the full name %s', name => {
+            expect(INDEXED_TRACK_OPTIONS.find(option => option.label === name)?.value)
+                .toBe(name)
+        })
+    })
+
+    it('deliberately uses different values for the two surfaces', () => {
+        // Guards against someone "aligning" the two lists: one talks to the
+        // search API (abbreviations), the other to the vector index (names).
+        const bulk = BULK_TRACK_OPTIONS.map(option => option.value)
+        const indexed = INDEXED_TRACK_OPTIONS.map(option => option.value)
+
+        expect(bulk).not.toEqual(indexed)
+    })
+})

src/apps/admin/src/ai/topscout-rag/ingest-options.ts

@@ -0,0 +1,68 @@
+import { InputSelectOption } from '~/libs/ui'
+
+/**
+ * Track/type vocabularies for this page.
+ *
+ * These deliberately differ between the two panels, and the difference is not
+ * cosmetic:
+ *
+ *  - **Ingest** filters reach the v6 Challenges API (bulk ingestion paginates
+ *    `searchChallengesTool`), which resolves `tracks`/`types` against the
+ *    `challengeTrack`/`challengeType` tables' **abbreviation** column. Sending
+ *    "Development" there matches no row, so the filter silently widens to
+ *    "every track" instead of erroring.
+ *  - **Indexed Challenges** filters query stored chunk metadata, where
+ *    ingestion recorded the track's full **name** (`toFreeFormName(challenge.track)`
+ *    in challenge-ingestion-workflow.ts). Sending "Dev" there matches nothing.
+ *
+ * So: abbreviations going out to the search API, full names coming back from
+ * the index. Do not "align" the two lists.
+ */
+export const BULK_TRACK_OPTIONS: InputSelectOption[] = [
+    { label: 'Any track', value: '' },
+    { label: 'Development', value: 'Dev' },
+    { label: 'Design', value: 'Des' },
+    { label: 'Data Science', value: 'DS' },
+    { label: 'Quality Assurance', value: 'QA' },
+]
+
+/**
+ * Challenge types are resolved by abbreviation too, but unlike tracks the
+ * abbreviations here are unconfirmed — they are rows in `challengeType`, not
+ * constants in any repo. These values are the type *names*; if bulk ingestion
+ * ignores a type filter, this list is the first place to look.
+ */
+export const BULK_TYPE_OPTIONS: InputSelectOption[] = [
+    { label: 'Any type', value: '' },
+    { label: 'Challenge', value: 'Challenge' },
+    { label: 'First2Finish', value: 'First2Finish' },
+    { label: 'Marathon Match', value: 'Marathon Match' },
+    { label: 'Task', value: 'Task' },
+]
+
+/** Mirrors the workflow's own default status set. */
+export const BULK_STATUS_OPTIONS: InputSelectOption[] = [
+    { label: 'Active + Completed', value: '' },
+    { label: 'Active only', value: 'ACTIVE' },
+    { label: 'Completed only', value: 'COMPLETED' },
+]
+
+/**
+ * Indexed-challenge filters match stored metadata, so these carry the track's
+ * full name — see the note on BULK_TRACK_OPTIONS.
+ */
+export const INDEXED_TRACK_OPTIONS: InputSelectOption[] = [
+    { label: 'All tracks', value: '' },
+    { label: 'Development', value: 'Development' },
+    { label: 'Design', value: 'Design' },
+    { label: 'Data Science', value: 'Data Science' },
+    { label: 'Quality Assurance', value: 'Quality Assurance' },
+]
+
+export const INDEXED_TYPE_OPTIONS: InputSelectOption[] = [
+    { label: 'All types', value: '' },
+    { label: 'Challenge', value: 'Challenge' },
+    { label: 'First2Finish', value: 'First2Finish' },
+    { label: 'Marathon Match', value: 'Marathon Match' },
+    { label: 'Task', value: 'Task' },
+]