fix(search): validate Lucid queries and isolate stale candidates

Author: waleedlatif1

apps/sim/lib/api/contracts/mothership-assistant-tools.ts

@@ -4,8 +4,10 @@ import { LIVE_SEARCH_PROVIDER_IDS } from '@/lib/sim-search/live/provider-catalog
 export const liveSearchProviderSchema = z.enum(LIVE_SEARCH_PROVIDER_IDS)
 export type LiveSearchProvider = z.output<typeof liveSearchProviderSchema>
 
-export const NOTION_SEARCH_TERMS_REQUIRED =
-  'Notion requires search terms. Add keywords or a concise question.'
+export const SEARCH_TERMS_REQUIRED = {
+  notion: 'Notion requires search terms. Add keywords or a concise question.',
+  lucid: 'Lucid requires search terms. Add document-title keywords or a literal shape-text query.',
+} as const
 
 /**
  * Native queries one call may send to the same provider account. Alternatives run as separate
@@ -64,11 +66,11 @@ export const nativeSearchQuerySchema = z
           message: `${input.provider} kind must be one of: ${kinds.options.join(', ')}.`,
         })
     }
-    if (input.provider === 'notion' && !input.query)
+    if ((input.provider === 'notion' || input.provider === 'lucid') && !input.query)
       context.addIssue({
         code: 'custom',
         path: ['query'],
-        message: NOTION_SEARCH_TERMS_REQUIRED,
+        message: SEARCH_TERMS_REQUIRED[input.provider],
       })
   })
 export type NativeSearchQuery = z.output<typeof nativeSearchQuerySchema>

apps/sim/lib/mothership/generated/sim-assistant-tools.generated.ts

@@ -22,8 +22,10 @@ export const liveSearchProviderSchema = z.enum([
 ])
 export type LiveSearchProvider = z.output<typeof liveSearchProviderSchema>
 
-export const NOTION_SEARCH_TERMS_REQUIRED =
-  'Notion requires search terms. Add keywords or a concise question.'
+export const SEARCH_TERMS_REQUIRED = {
+  notion: 'Notion requires search terms. Add keywords or a concise question.',
+  lucid: 'Lucid requires search terms. Add document-title keywords or a literal shape-text query.',
+} as const
 
 /**
  * Native queries one call may send to the same provider account. Alternatives run as separate
@@ -82,11 +84,11 @@ export const nativeSearchQuerySchema = z
           message: `${input.provider} kind must be one of: ${kinds.options.join(', ')}.`,
         })
     }
-    if (input.provider === 'notion' && !input.query)
+    if ((input.provider === 'notion' || input.provider === 'lucid') && !input.query)
       context.addIssue({
         code: 'custom',
         path: ['query'],
-        message: NOTION_SEARCH_TERMS_REQUIRED,
+        message: SEARCH_TERMS_REQUIRED[input.provider],
       })
   })
 export type NativeSearchQuery = z.output<typeof nativeSearchQuerySchema>

apps/sim/lib/sim-search/live/application.test.ts

@@ -287,20 +287,34 @@ describe('authorized live retrieval', () => {
       }
     )
   })
-  it.each([
-    { startDate: '2026-08-01T00:00:00Z' },
-    { source: ' notion ', startDate: '2026-08-01T00:00:00Z' },
-    { sortBy: 'newest' as const },
-    { sortBy: 'oldest' as const },
-  ])('rejects a Notion-only live listing with %j', async (bound) => {
-    await expect(
-      searchLiveKnowledge.execute({
-        principal,
-        input: { ...input, query: ' \t ', filters: { source: 'notion', ...bound } },
+  describe.each(['notion', 'lucid'] as const)('%s requires terms before dispatch', (provider) => {
+    it.each([
+      { startDate: '2026-08-01T00:00:00Z' },
+      { source: ` ${provider} `, startDate: '2026-08-01T00:00:00Z' },
+      { sortBy: 'newest' as const },
+      { sortBy: 'oldest' as const },
+    ])('rejects a provider-only listing with %j', async (bound) => {
+      await expect(
+        searchLiveKnowledge.execute({
+          principal,
+          input: { ...input, query: ' \t ', filters: { source: provider, ...bound } },
+        })
+      ).rejects.toMatchObject({ code: 'validation' })
+    })
+    it('rejects an empty native query even with a date bound', async () => {
+      await expect(
+        searchLiveKnowledge.execute({
+          principal,
+          input: {
+            ...input,
+            query: 'topology',
+            filters: { startDate: '2026-08-01T00:00:00Z' },
+            nativeQueries: [{ provider, query: ' \t ' }],
+          },
+        })
+      ).rejects.toMatchObject({
+        issues: expect.arrayContaining([expect.objectContaining({ path: [0, 'query'] })]),
       })
-    ).rejects.toMatchObject({
-      code: 'validation',
-      message: 'Notion requires search terms. Add keywords or a concise question.',
     })
   })
   it.each([undefined, 'google_drive'])(

apps/sim/lib/sim-search/live/application.ts

@@ -10,8 +10,8 @@ import {
   type LiveSearchAccountStatus,
   liveSearchProviderSchema,
   type NativeSearchQuery,
-  NOTION_SEARCH_TERMS_REQUIRED,
   nativeSearchQueriesSchema,
+  SEARCH_TERMS_REQUIRED,
   workspaceSearchFiltersSchema,
 } from '@/lib/api/contracts/mothership-assistant-tools'
 import { canonicalJson, fingerprint, instantScopePart } from '@/lib/api/cursor-binding'
@@ -344,8 +344,12 @@ export const searchLiveKnowledge = defineAuthorizedKnowledgeUseCase({
     )
       throw new OrchestrationError('validation', 'Invalid live search query or result limit')
     const filters = input.filters
-    if (!queries && filters?.source === 'notion' && !input.query.trim())
-      throw new OrchestrationError('validation', NOTION_SEARCH_TERMS_REQUIRED)
+    if (
+      !queries &&
+      (filters?.source === 'notion' || filters?.source === 'lucid') &&
+      !input.query.trim()
+    )
+      throw new OrchestrationError('validation', SEARCH_TERMS_REQUIRED[filters.source])
     if (
       filters?.startDate &&
       filters.endDate &&

apps/sim/lib/sim-search/live/lucid-mcp.ts

@@ -41,7 +41,10 @@ function resource(value: string): { id: string; kind?: string } | undefined {
   }
 }
 
-async function metadata(client: ManagedSearchMcpClient, id: string): Promise<NativeDocument> {
+async function metadata(
+  client: ManagedSearchMcpClient,
+  id: string
+): Promise<NativeDocument | undefined> {
   if (!UUID.test(id)) invalid('Invalid Lucid document identity.')
   const row = object(await client.call('lucid_get_document_metadata', { document_id: id }))
   const url = resource(string(row.viewUrl))
@@ -58,7 +61,7 @@ async function metadata(client: ManagedSearchMcpClient, id: string): Promise<Nat
     typeof row.lastModified !== 'string' ||
     !Number.isFinite(Date.parse(row.lastModified))
   )
-    invalid('Lucid document metadata is incomplete, changed, or no longer readable.')
+    return undefined
   return {
     id,
     kind: url.kind,
@@ -99,6 +102,7 @@ export async function searchLucidMcp(
         'Lucid document search requires a document UUID or Lucid URL and a literal text query.'
       )
     const document = await metadata(client, scope.id)
+    if (!document) invalid('Lucid document metadata is incomplete or no longer readable.')
     if (
       (scope.kind && document.kind !== scope.kind) ||
       !products.some((product) => product === document.kind)
@@ -176,13 +180,16 @@ export async function searchLucidMcp(
   const limit = Math.max(1, Math.min(MAX_CANDIDATES, input.limit))
   const documents = await mapWithConcurrency(candidates.slice(0, limit), 3, async (candidate) => {
     const document = await metadata(client, candidate.id)
-    if (document.kind !== candidate.kind) invalid('Lucid document product changed during search.')
+    if (!document || document.kind !== candidate.kind) {
+      dropped = true
+      return undefined
+    }
     return document
   })
   const capped = candidates.length > limit || result.results.length >= 200
   const localDates = hasDateBounds(input.filters) || Boolean(dateSortDirection(input.filters))
   return {
-    documents,
+    documents: documents.filter((document) => document !== undefined),
     hasMore: capped,
     partial: dropped || capped || localDates,
     message:
@@ -191,7 +198,7 @@ export async function searchLucidMcp(
         ? ' Dates use current modification timestamps; sorting and the end-date filter cover only the retrieved candidates, not the entire account.'
         : '') +
       (capped ? ' The candidate limit was reached; narrow the title query.' : '') +
-      (dropped ? ' Unsupported search results were excluded.' : ''),
+      (dropped ? ' Unsupported or no-longer-readable search results were excluded.' : ''),
   }
 }
 
@@ -225,6 +232,7 @@ export async function readLucidMcp(
   reference: Pick<NativeDocument, 'id' | 'kind' | 'revision'>
 ): Promise<NativeDocument> {
   const before = await metadata(client, reference.id)
+  if (!before) invalid('Lucid document metadata is incomplete or no longer readable.')
   if (
     (reference.kind && reference.kind !== before.kind) ||
     (reference.revision && reference.revision !== before.revision)
@@ -304,6 +312,7 @@ export async function readLucidMcp(
   }
   const after = await metadata(client, before.id)
   if (
+    !after ||
     after.revision !== before.revision ||
     after.kind !== before.kind ||
     after.modifiedAt !== before.modifiedAt ||

apps/sim/scripts/test-search-lucid-e2e.ts

@@ -93,9 +93,15 @@ protocol.setRequestHandler(CallToolRequestSchema, async (request) => {
   if (mode === 'tool-error') return failed()
   if (name === 'lucid_get_document_metadata') {
     metadataReads++
+    if (args.document_id === ID && mode === 'candidate-error') return failed()
+    if (args.document_id === ID && mode === 'candidate-rate')
+      return { ...failed(), content: [{ type: 'text' as const, text: 'Rate limit reached' }] }
     if (mode === 'revoked' && metadataReads > 1) return failed()
     return result({
-      documentId: mode === 'metadata-id' ? OTHER_ID : args.document_id,
+      documentId:
+        mode === 'metadata-id' || (mode === 'stale-candidate' && args.document_id === ID)
+          ? OTHER_ID
+          : args.document_id,
       title: TITLE,
       product: mode === 'spark' ? 'lucidspark' : 'lucidchart',
       viewUrl:
@@ -119,7 +125,12 @@ protocol.setRequestHandler(CallToolRequestSchema, async (request) => {
     assert(
       Object.keys(args).every((key) => ['query', 'product', 'last_modified_after'].includes(key))
     )
-    const rowCount = mode === 'search-cap' ? 200 : 1
+    const rowCount =
+      mode === 'search-cap'
+        ? 200
+        : ['stale-candidate', 'candidate-error', 'candidate-rate'].includes(mode)
+          ? 2
+          : 1
     return result({
       query: args.query,
       results:
@@ -387,6 +398,31 @@ try {
     padding += 'x'
     await assert.rejects(read, /512 KiB/)
   })
+  await check(
+    'A stale candidate cannot discard another independently readable document',
+    async () => {
+      mode = 'stale-candidate'
+      const page = await search()
+      assert.deepEqual(
+        page.documents.map((document) => document.id),
+        [OTHER_ID]
+      )
+      assert.equal(page.partial, true)
+      assert.match(page.message ?? '', /excluded/i)
+    }
+  )
+  for (const [failure, status] of [
+    ['candidate-error', 'unavailable'],
+    ['candidate-rate', 'rate_limited'],
+  ] as const) {
+    await check(`Candidate provider failure ${status} remains terminal`, async () => {
+      mode = failure
+      await assert.rejects(
+        () => search(),
+        (error: unknown) => error instanceof NativeSearchError && error.status === status
+      )
+    })
+  }
   await check('Capped title search discloses coverage without inventing a cursor', async () => {
     mode = 'search-cap'
     const page = await search()