improvement(search): return a longer live read window when an MCP client asks for more chunks

Author: waleedlatif1

apps/sim/lib/knowledge/mcp/server.protocol.test.ts

@@ -93,8 +93,15 @@ describe('Search MCP protocol', () => {
       expect(JSON.parse(searchContent.text)).toMatchObject({
         results: [{ documentId, queryIndexes: [0] }],
       })
-      const read = await client.callTool({ name: 'read_document', arguments: { documentId } })
+      const read = await client.callTool({
+        name: 'read_document',
+        arguments: { documentId, limit: 8 },
+      })
       expect(read.isError).not.toBe(true)
+      expect(mocks.liveRead.mock.lastCall?.[0].input).toMatchObject({
+        limit: 8,
+        scaleWindowWithLimit: true,
+      })
       expect(read.content).toEqual([
         { type: 'text', text: expect.stringContaining('Read evidence') },
       ])

apps/sim/lib/knowledge/mcp/server.ts

@@ -298,6 +298,7 @@ export function createKnowledgeMcpServer(context: KnowledgeMcpContext): McpServe
                 ...options,
                 documentId: id,
                 organizationId,
+                scaleWindowWithLimit: true,
                 resultSecretRegistry: registry,
                 signal,
               },

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

@@ -1189,6 +1189,51 @@ describe('authorized live retrieval', () => {
       })
     ).rejects.toBe('user_stop:test')
   })
+  describe('read window', () => {
+    const content = 'x'.repeat(30_000)
+    async function readWindow(extra: Record<string, unknown>) {
+      const [result] = (await searchLiveKnowledge.execute({ principal, input })).results
+      if (!result) throw new Error('Expected the search fixture to return a result')
+      mocks.read.mockResolvedValue({ ...document, content })
+      const read = await readLiveDocument.execute({
+        principal,
+        input: {
+          workspaceId: 'workspace',
+          documentId: result.documentId,
+          limit: 3,
+          resultSecretRegistry: new ResolvedSecretTraceRegistry([]),
+          ...extra,
+        },
+      })
+      const [chunk] = read.chunks
+      if (!chunk) throw new Error('Expected the read to return a window')
+      return { length: chunk.content.length, next: read.next }
+    }
+
+    it('returns a proportionally longer window when a scaling caller asks for more chunks', async () => {
+      const { length, next } = await readWindow({ limit: 8, scaleWindowWithLimit: true })
+      expect(length).toBe(21_333)
+      expect(next).toEqual({ startChunkIndex: 0, startOffset: 21_333 })
+    })
+
+    it('continues a longer window from where the previous one ended', async () => {
+      const { length, next } = await readWindow({
+        limit: 8,
+        scaleWindowWithLimit: true,
+        startOffset: 21_333,
+      })
+      expect(length).toBe(30_000 - 21_333)
+      expect(next).toBeNull()
+    })
+
+    it.each([
+      ['at the default limit', { limit: 3, scaleWindowWithLimit: true }],
+      ['below the default limit', { limit: 1, scaleWindowWithLimit: true }],
+      ['for a caller that does not scale', { limit: 8 }],
+    ])('keeps the standard window %s', async (_, extra) => {
+      expect((await readWindow(extra)).length).toBe(8_000)
+    })
+  })
   it('refuses a chunk index other than zero before reading the document from its provider', async () => {
     const [result] = (await searchLiveKnowledge.execute({ principal, input })).results
     if (!result) throw new Error('Expected the search fixture to return a result')

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

@@ -287,8 +287,20 @@ export const LIVE_NOT_STARTED_MESSAGE =
 export const LIVE_OVER_ACCOUNT_LIMIT_MESSAGE = `Not searched: one search covers at most ${MAX_ACCOUNTS} accounts. Search this source or account alone.`
 /** Previews center on the query's longest matching term, like indexed passages. */
 const PREVIEW_CHARACTERS = 1800
-/** Characters one read returns, the same page budget indexed document reads use. */
+/** Characters one read returns at the default limit, the page budget indexed document reads use. */
 const READ_WINDOW_CHARACTERS = 8000
+/** The read limit {@link READ_WINDOW_CHARACTERS} is sized for, which is the default limit. */
+const READ_WINDOW_LIMIT = 3
+
+/**
+ * Characters one read returns. A live document is one chunk, so a caller that scales the window
+ * gets a longer one when it asks for more chunks than the default; at or below the default, and
+ * for every other caller, the window stays the standard page.
+ */
+function readWindowCharacters(input: Pick<LiveReadInput, 'limit' | 'scaleWindowWithLimit'>) {
+  if (!input.scaleWindowWithLimit || input.limit <= READ_WINDOW_LIMIT) return READ_WINDOW_CHARACTERS
+  return Math.floor((READ_WINDOW_CHARACTERS * input.limit) / READ_WINDOW_LIMIT)
+}
 
 /**
  * Verifies candidates in rank order. A revoked grant fails the whole account; any other
@@ -980,6 +992,8 @@ export type LiveReadInput = ResourceOwner & {
   documentId: string
   filters?: WorkspaceSearchFilters
   limit: number
+  /** Lengthens the window for a limit above the default, as {@link readWindowCharacters} sizes it. */
+  scaleWindowWithLimit?: boolean
   startChunkIndex?: number
   startOffset?: number
   resultSecretRegistry: ResolvedSecretTraceRegistry
@@ -1125,7 +1139,7 @@ export const readLiveDocument = defineAuthorizedKnowledgeUseCase({
         'validation',
         'This document changed. Read again from the beginning'
       )
-    const end = Math.min(content.length, start + READ_WINDOW_CHARACTERS)
+    const end = Math.min(content.length, start + readWindowCharacters(input))
     return {
       documentId: input.documentId,
       knowledgeBaseId: '',