fix(slack): fold the channel header into the content hash so a rename or topic edit re-indexes

Author: waleedlatif1

apps/sim/connectors/slack/slack.test.ts

@@ -38,13 +38,16 @@ let listNextCursor = ''
 let history: Record<string, unknown>[] = MESSAGES
 /** Whether `conversations.info` reports the channel as missing; per-test overridable. */
 let channelMissing = false
+/** The channel `conversations.info` returns; per-test overridable. */
+let infoChannel: Record<string, unknown> = GENERAL
 
 beforeEach(() => {
   requestedUrls.length = 0
   listedChannels = [GENERAL, PLATFORM]
   listNextCursor = ''
   history = MESSAGES
   channelMissing = false
+  infoChannel = GENERAL
   fetchMock.mockReset()
   fetchMock.mockImplementation(async (input) => {
     const url = new URL(String(input))
@@ -61,7 +64,7 @@ beforeEach(() => {
       case '/api/conversations.info':
         return channelMissing
           ? jsonResponse({ ok: false, error: 'channel_not_found' })
-          : jsonResponse({ ok: true, channel: GENERAL })
+          : jsonResponse({ ok: true, channel: infoChannel })
       case '/api/conversations.history':
         return jsonResponse({ ok: true, messages: history, response_metadata: {} })
       case '/api/users.info': {
@@ -163,7 +166,9 @@ describe('getDocument', () => {
     expect(doc).toMatchObject({
       externalId: 'C0GENERAL',
       title: '#general',
-      contentHash: 'slack-v3:C0GENERAL:1700000000.000100:1700000200.000100:3:noedit:noreply:0',
+      contentHash: expect.stringMatching(
+        /^slack-v3:C0GENERAL:[0-9a-f]{16}:1700000000\.000100:1700000200\.000100:3:noedit:noreply:0$/
+      ),
       metadata: expect.objectContaining({ channelName: 'general', messageCount: 2 }),
     })
     expect(doc?.content).toBe(
@@ -177,6 +182,22 @@ describe('getDocument', () => {
     )
   })
 
+  it('moves the hash when the header changes without any message changing', async () => {
+    const before = await slackConnector.getDocument('token', {}, 'C0GENERAL', {})
+
+    infoChannel = { ...GENERAL, name: 'general-renamed' }
+    const renamed = await slackConnector.getDocument('token', {}, 'C0GENERAL', {})
+    expect(renamed?.contentHash).not.toBe(before?.contentHash)
+
+    infoChannel = { ...GENERAL, topic: { value: 'A new topic' } }
+    const retopiced = await slackConnector.getDocument('token', {}, 'C0GENERAL', {})
+    expect(retopiced?.contentHash).not.toBe(before?.contentHash)
+
+    infoChannel = GENERAL
+    const again = await slackConnector.getDocument('token', {}, 'C0GENERAL', {})
+    expect(again?.contentHash).toBe(before?.contentHash)
+  })
+
   it('keeps a channel with no messages as a live document', async () => {
     history = []
     const doc = await slackConnector.getDocument('token', {}, 'C0GENERAL', {})

apps/sim/connectors/slack/slack.ts

@@ -1,3 +1,4 @@
+import { createHash } from 'node:crypto'
 import { createLogger } from '@sim/logger'
 import { toError } from '@sim/utils/errors'
 import { generateId } from '@sim/utils/id'
@@ -602,13 +603,15 @@ async function buildSlackChannelDocument(
     maxMessages
   )
 
-  const lines = new BoundedLines(CONNECTOR_TEXT_DOCUMENT_MAX_BYTES, 'last')
-  lines.pin(`Channel: #${channel.name}`)
+  const header = [`Channel: #${channel.name}`]
   const topic = channel.topic?.value?.trim()
-  if (topic) lines.pin(`Topic: ${topic}`)
+  if (topic) header.push(`Topic: ${topic}`)
   const purpose = channel.purpose?.value?.trim()
-  if (purpose) lines.pin(`Purpose: ${purpose}`)
-  lines.pin('')
+  if (purpose) header.push(`Purpose: ${purpose}`)
+
+  /** The newest messages survive when the window does not fit; the header always does. */
+  const lines = new BoundedLines(CONNECTOR_TEXT_DOCUMENT_MAX_BYTES, 'last')
+  lines.pin(...header, '')
   await appendMessages(accessToken, lines, messages, syncContext)
   const messageCount = lines.count
 
@@ -633,13 +636,20 @@ async function buildSlackChannelDocument(
    * in catches deletes (count drops) but still cannot detect reply edits
    * without fetching `conversations.replies` for each parent.
    *
+   * The header is digested into the hash because it is part of the document:
+   * renaming a channel or editing its topic changes the indexed text without
+   * touching a single message, and the sync engine drops a refresh whose hash
+   * matches the stored one. A digest keeps the hash bounded and free of the
+   * delimiter collisions raw topic text would bring.
+   *
    * The `slack-v3` prefix forces a one-time re-index of channels indexed
    * before the document gained its header and size ceiling; `slack-v2` did the
    * same when attachment and Block Kit content started being extracted.
    * Per-message `ts` and the window are unchanged by either, so without the
    * bump the hash would match and the richer content would never be embedded.
    */
-  const contentHash = `slack-v3:${channel.id}:${oldestTs ?? 'empty'}:${lastActivityTs ?? 'empty'}:${messages.length}:${maxEditTs || 'noedit'}:${maxReplyTs || 'noreply'}:${totalReplies}`
+  const headerDigest = createHash('sha256').update(header.join('\n')).digest('hex').slice(0, 16)
+  const contentHash = `slack-v3:${channel.id}:${headerDigest}:${oldestTs ?? 'empty'}:${lastActivityTs ?? 'empty'}:${messages.length}:${maxEditTs || 'noedit'}:${maxReplyTs || 'noreply'}:${totalReplies}`
 
   return { content: lines.join(), contentHash, messageCount, lastActivityTs }
 }