fix(search): keep Confluence searchable for colon queries and say when nothing was searched

apps/sim/app/workspace/[workspaceId]/home/components/knowledge-search-results/knowledge-search-results.tsx

@@ -1,12 +1,13 @@
 'use client'
 
 import { useEffect, useMemo, useState } from 'react'
-import { Chip } from '@sim/emcn'
+import { Chip, ChipLink } from '@sim/emcn'
 import { useQueryStates } from 'nuqs'
 import { ActivityStatus } from '@/components/ui/activity-status'
 import type { WorkspaceSearchFilters } from '@/lib/api/contracts/knowledge'
 import { useSession } from '@/lib/auth/client'
 import { type ResourceScope, resourceScopeKey } from '@/lib/core/resource-scope'
+import { organizationRoutes } from '@/lib/navigation/paths'
 import { SearchFilters } from '@/app/workspace/[workspaceId]/home/components/knowledge-search-results/search-filters'
 import {
   groupResultsByDocument,
@@ -150,6 +151,25 @@ function LiveSearchResults({
             Search found no results.
           </p>
         )}
+      {!isFetching && data?.live && accounts.length === 0 && (
+        <div className='flex items-center gap-2 px-2 py-2'>
+          <p role='status' className='text-[var(--text-muted)] text-caption'>
+            {filters.source
+              ? 'You haven’t connected this app.'
+              : 'Connect your apps to search them.'}
+          </p>
+          <ChipLink
+            variant='border'
+            href={
+              scope.kind === 'organization'
+                ? organizationRoutes(scope.organizationId).integrations
+                : `/workspace/${scope.workspaceId}/integrations`
+            }
+          >
+            Connect apps
+          </ChipLink>
+        </div>
+      )}
       {accounts.some((account) =>
         ['unavailable', 'timeout', 'rate_limited', 'reconnect'].includes(account.status)
       ) && (

apps/sim/lib/mothership/tools/server/knowledge/workspace-search.test.ts

@@ -246,6 +246,19 @@ describe('Assistant retrieval tools', () => {
     )
     expect(JSON.stringify(result)).not.toContain(secret)
   })
+  it('tells the model a search no account served searched nothing, not to narrow and retry', async () => {
+    mocks.search.mockResolvedValueOnce({
+      query: 'roadmap',
+      retrieval: { status: 'partial', timedOutLegs: [] },
+      results: [],
+      live: { backend: 'live', accounts: [], guidance: 'Nothing was searched: connect one.' },
+    })
+    const result = await searchWorkspaceServerTool.execute({ query: 'roadmap' }, context)
+    expect(result).toMatchObject({
+      success: true,
+      message: expect.stringMatching(/^Nothing was searched: (?!.*more specific query)/),
+    })
+  })
   it('rejects untrusted contexts, incompatible sources and out-of-scope document reads', async () => {
     expect(
       await searchWorkspaceServerTool.execute(

apps/sim/lib/mothership/tools/server/knowledge/workspace-search.ts

@@ -85,9 +85,13 @@ export const searchWorkspaceServerTool: BaseServerTool = {
                   ...liveInput,
                   workspaceId: scope.workspaceId,
                 })
+          /** Every searched account reports a status, so none means no account could serve it. */
+          const nothingSearched = data.live?.accounts.length === 0
           return {
             success: true,
-            message: `${data.retrieval.status === 'partial' ? 'Search coverage is incomplete. Continue with a more specific query or source filter; these results cannot establish absence or completeness. ' : ''}Found ${data.results.length} live results. Read a documentId when its passage does not answer the question or more context is needed. ${CITATION_INSTRUCTION}`,
+            message: nothingSearched
+              ? 'Nothing was searched: no connected account can serve this search, so it says nothing about what exists. Follow the first line of live.guidance instead of retrying the same search.'
+              : `${data.retrieval.status === 'partial' ? 'Search coverage is incomplete. Continue with a more specific query or source filter; these results cannot establish absence or completeness. ' : ''}Found ${data.results.length} live results. Read a documentId when its passage does not answer the question or more context is needed. ${CITATION_INSTRUCTION}`,
             data: {
               ...data,
               results: data.results.map((item) => ({

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

@@ -666,6 +666,28 @@ describe('authorized live retrieval', () => {
     expect(result.live?.accounts.map((status) => status.queryIndex)).toEqual([0, 1])
     expect(result.live?.guidance).toBe('Live coverage: slack')
   })
+  it('reports a search that reached no account as nothing searched, not as no results', async () => {
+    mocks.accounts.mockResolvedValue([])
+    const result = await searchLiveKnowledge.execute({ principal, input })
+    expect(mocks.search).not.toHaveBeenCalled()
+    expect(result.results).toEqual([])
+    expect(result.retrieval.status).toBe('partial')
+    expect(result.live?.accounts).toEqual([])
+    expect(result.live?.guidance).toMatch(
+      /^Nothing was searched: you have no connected sources here\. Connect your accounts at \S*\/workspace\/workspace\/integrations,/
+    )
+  })
+  it('names the filtered source when only other sources are connected', async () => {
+    const result = await searchLiveKnowledge.execute({
+      principal,
+      input: { ...input, filters: { source: 'slack' } },
+    })
+    expect(mocks.search).not.toHaveBeenCalled()
+    expect(result.retrieval.status).toBe('partial')
+    expect(result.live?.guidance).toMatch(
+      /^Nothing was searched: you have no connected slack account here\. .* or search without the source filter\./
+    )
+  })
   describe('result previews', () => {
     const excerpts = ['a', 'b', 'c', 'd'].map((id) => ({
       ...document,

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

@@ -32,6 +32,7 @@ import {
   mapWithConcurrency,
   waitWithAbort,
 } from '@/lib/core/utils/concurrency'
+import { getBaseUrl } from '@/lib/core/utils/urls'
 import { MANAGED_MCP_CONNECTORS } from '@/lib/credential-groups/managed-mcp-connectors'
 import { requireOrganizationSearchAvailable } from '@/lib/knowledge/access/availability'
 import { defineAuthorizedKnowledgeUseCase } from '@/lib/knowledge/application/authorized-knowledge-use-case'
@@ -49,6 +50,7 @@ import {
 import { RRF_K } from '@/lib/knowledge/search/recency'
 import { matchPassage } from '@/lib/knowledge/search/snippet'
 import { isKnowledgeSourceUrl } from '@/lib/knowledge/search/source-url'
+import { organizationRoutes } from '@/lib/navigation/paths'
 import { connectorDisplayName } from '@/lib/sim-search/connectors'
 import {
   type LiveAccountSession,
@@ -440,6 +442,24 @@ function compareAccounts(left: LiveAccount, right: LiveAccount): number {
   )
 }
 
+/**
+ * Why a plain search that reached no account returned nothing, and where the caller connects one;
+ * an empty result would otherwise read as a search that found nothing.
+ */
+function nothingSearchedGuidance(
+  owner: ResourceOwner,
+  source: string | undefined,
+  hasOtherAccounts: boolean
+): string {
+  const path = owner.organizationId
+    ? organizationRoutes(owner.organizationId).integrations
+    : `/workspace/${owner.workspaceId}/integrations`
+  const connect = new URL(path, getBaseUrl()).href
+  return source && hasOtherAccounts
+    ? `Nothing was searched: you have no connected ${source} account here. Connect one at ${connect}, or search without the source filter.`
+    : `Nothing was searched: you have no connected sources here. Connect your accounts at ${connect}, then search again.`
+}
+
 const liveKnowledgeSearch = defineAuthorizedKnowledgeUseCase({
   operation: knowledgeOperations.search,
   resolveContext: ({ input }: { input: LiveSearchInput }) => resolveKnowledgeOwnerContext(input),
@@ -503,6 +523,7 @@ const liveKnowledgeSearch = defineAuthorizedKnowledgeUseCase({
           nativesFor(account).length > 0
       )
       .sort(compareAccounts)
+    const nothingSearched = !queries && eligible.length === 0
     const selected = eligible.slice(0, MAX_ACCOUNTS)
     const openingAccount = createConcurrencyLimit(ACCOUNT_OPEN_CONCURRENCY)
     annotateSearchDiagnostics({
@@ -909,9 +930,14 @@ const liveKnowledgeSearch = defineAuthorizedKnowledgeUseCase({
       }
     })
     /** Live coverage guides the providers this call searched or tried to search. */
-    const guidance = liveSearchGuidance(
-      accounts.filter((account) => account.status !== 'reconnect').map(({ provider }) => provider)
-    )
+    const guidance = [
+      ...(nothingSearched
+        ? [nothingSearchedGuidance(input, filters?.source, allAccounts.length > 0)]
+        : []),
+      liveSearchGuidance(
+        accounts.filter((account) => account.status !== 'reconnect').map(({ provider }) => provider)
+      ),
+    ].join('\n')
     for (const account of eligible.slice(MAX_ACCOUNTS)) {
       for (const { queryIndex } of nativesFor(account))
         accounts.push({
@@ -934,9 +960,10 @@ const liveKnowledgeSearch = defineAuthorizedKnowledgeUseCase({
           message: 'No connection with this provider is configured and approved in this scope.',
         })
     }
-    const retrievalStatus = accounts.some((account) => account.status !== 'ok')
-      ? 'partial'
-      : 'complete'
+    const retrievalStatus =
+      nothingSearched || accounts.some((account) => account.status !== 'ok')
+        ? 'partial'
+        : 'complete'
     annotateSearchDiagnostics({
       resultCount: Math.min(ranked.length, input.topK),
       retrievalStatus,

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

@@ -244,6 +244,13 @@ describe('Jira any-term search', () => {
     const page = await searchAtlassian(api, 'confluence', { ...query, textMatch: 'any' })
     expect(page.documents.map((document) => document.id)).toEqual(['5'])
   })
+
+  it('searches nothing for a Confluence query with no word left, which would list every page', async () => {
+    const api = client({})
+    const page = await searchAtlassian(api, 'confluence', { ...query, query: String.raw` :\ ` })
+    expect(page.documents).toEqual([])
+    expect(api.json).not.toHaveBeenCalled()
+  })
 })
 
 describe('Confluence plain search ranking', () => {

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

@@ -132,6 +132,12 @@ export async function searchAtlassian(
   provider: 'jira' | 'confluence',
   input: NativeSearchInput
 ): Promise<NativePage> {
+  const text = nativeText(input)
+  const confluenceClause =
+    provider === 'confluence' && text && !input.native?.query
+      ? confluenceTextClause(text, input.textMatch)
+      : ''
+  if (confluenceClause === undefined) return { documents: [] }
   const allSites = (await sites(client)).filter(
     (site) =>
       !input.policy?.sites.length ||
@@ -166,7 +172,6 @@ export async function searchAtlassian(
       .join(' AND ') || undefined
   const direction = dateSortDirection(input.filters)
   const order = direction ? `${field} ${direction.toUpperCase()}` : undefined
-  const text = nativeText(input)
   // Jira's any-term JQL matches issues sharing two common words; like minimum-should-match,
   // keep only issues holding at least half the terms.
   const relevant = input.textMatch === 'any' && text ? holdsHalfTheTerms(text) : undefined
@@ -234,13 +239,15 @@ export async function searchAtlassian(
         }
       }
       const exact = await searchPages(
-        text ? confluenceTextClause(text, input.textMatch) : '',
+        confluenceClause,
         input.native?.cursor && single ? input.native.cursor : undefined
       )
+      const relevance = confluenceTextClause(text, 'any')
       // An all-terms match larger than a page comes back by raw term matches; rank it by relevance.
-      if (!text || input.native || order || input.textMatch === 'any' || !exact.next) return exact
+      if (!relevance || input.native || order || input.textMatch === 'any' || !exact.next)
+        return exact
       try {
-        const ranked = await searchPages(confluenceTextClause(text, 'any'))
+        const ranked = await searchPages(relevance)
         return { ...ranked, next: undefined, ranking: 'relevance' }
       } catch {
         // A cancelled search stops here; any other failure keeps the all-terms page.

apps/sim/lib/sim-search/live/service-session.ts

@@ -146,6 +146,7 @@ export async function createLiveServiceSession(input: {
         const query =
           search.native?.query ??
           (search.query ? confluenceTextClause(search.query, search.textMatch) : '')
+        if (query === undefined) return null
         return {
           ...search,
           native: { provider, ...search.native, query: scopeAtlassianQuery(query, scope) },

apps/sim/lib/sim-search/live/text-match.test.ts

@@ -59,6 +59,16 @@ describe('text match clauses', () => {
     )
   })
 
+  it('sends a colon or backslash to Confluence as a word break, which it otherwise rejects', () => {
+    expect(confluenceTextClause('Security: Weekly Report', 'all')).toBe(
+      'text ~ "Security  Weekly Report"'
+    )
+    expect(confluenceTextClause(String.raw`C:\builds\out`, 'any')).toBe(
+      'siteSearch ~ "C  builds out"'
+    )
+    expect(confluenceTextClause(String.raw` :\ `, 'all')).toBeUndefined()
+  })
+
   it('falls back to the all-terms clause when fewer than two terms are distinctive', () => {
     expect(jiraTextClause('how does it work', 'any')).toBe('text ~ "how does it work"')
   })

apps/sim/lib/sim-search/live/text-match.ts

@@ -150,9 +150,16 @@ export function jiraTextClause(text: string, match: TextMatch = 'all'): string {
  * The text clause for a plain query in Confluence CQL. `any` uses `siteSearch`, Confluence's own
  * relevance search, which ranks a page with the query's rarer words above the many pages sharing
  * one common word, where a `text ~` OR query ranks them the other way round.
+ *
+ * Confluence answers HTTP 400 to the escaped colon or backslash this phrase would carry, and its
+ * analyzer splits words on both, so they go as spaces. `undefined` means nothing is left to match:
+ * Confluence would list every page for such a phrase, so callers search nothing instead.
  */
-export function confluenceTextClause(text: string, match: TextMatch = 'all'): string {
-  return `${match === 'any' ? 'siteSearch' : 'text'} ~ ${escapeSearchPhrase(text)}`
+export function confluenceTextClause(text: string, match: TextMatch = 'all'): string | undefined {
+  const searchable = text.replace(/[:\\]/g, ' ').trim()
+  return searchable
+    ? `${match === 'any' ? 'siteSearch' : 'text'} ~ ${escapeSearchPhrase(searchable)}`
+    : undefined
 }
 
 /** How many distinct terms occur as whole words in the text. */

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

@@ -445,7 +445,15 @@ try {
       const body = toRecord(response.request().postDataJSON())
       assert.equal(body.organizationId, organizationId)
       assert.equal(body.query, 'Launch notes')
-      await expect(page.getByText('Search found no results.', { exact: true })).toBeVisible()
+      // The seeded member has connected no account, so nothing was searched.
+      await expect(
+        page.getByText('Connect your apps to search them.', { exact: true })
+      ).toBeVisible()
+      await expect(page.getByText('Search found no results.', { exact: true })).toHaveCount(0)
+      await expect(page.getByRole('link', { name: 'Connect apps' })).toHaveAttribute(
+        'href',
+        `${root}/integrations`
+      )
       await expect(page).toHaveURL(
         (url) => url.pathname === `${root}/search` && url.searchParams.get('q') === 'Launch notes'
       )