@@ -37,10 +37,10 @@ When the user runs `/ship`:
   - `bun run check:migrations origin/staging` must pass (staging is the PR base). Do not silence a flagged statement with a `-- migration-safe:` annotation unless `/db-migrate` confirmed the old code no longer depends on it; otherwise split the destructive change into a later deploy.
 6. **Run pre-ship checks** from the repo root before staging. This has two phases: first **regenerate** every committed artifact so generated files never drift into a CI failure (this is what catches things like `agent-stream-docs` going stale after a `models.ts` edit), then run the **full audit suite** CI's `Lint and Test` job enforces. Both phases parallelize — but only across commands that write **disjoint** outputs — and a bare `wait` swallows child exit codes, so both phases below explicitly collect each job's status and abort ship if any failed.
 
-  **Phase A — regenerate the always-in-repo committed artifacts (parallel), then let step 7 stage whatever changed.** Regenerate only the generators whose inputs live entirely in this repo and that any ordinary code change can drift — `agent-stream-docs:generate` (derives from the provider model registry) and `skills:sync` (derives from `.agents/skills/**`). They write disjoint trees (`apps/docs/…/agent.mdx` vs the `.claude`/`.cursor` command projections), so they parallelize safely, and each is idempotent (a no-op when already in sync):
+  **Phase A — regenerate the always-in-repo committed artifacts (parallel), then let step 7 stage whatever changed.** Regenerate only the generators whose inputs live entirely in this repo and that any ordinary code change can drift — `agent-stream-docs:generate` (derives from the provider model registry), `skills:sync` (derives from `.agents/skills/**`), and `docs-manifest:generate` (derives from the `apps/docs/content/docs` page tree, so adding, removing, or renaming a page drifts it). They write disjoint files (`apps/docs/…/agent.mdx`, the `.claude`/`.cursor` command projections, and `apps/sim/lib/copilot/generated/docs-manifest.ts`), so they parallelize safely, and each is idempotent (a no-op when already in sync):
   ```bash
   rm -f /tmp/ship-gen-results
-  for g in agent-stream-docs:generate skills:sync; do
+  for g in agent-stream-docs:generate skills:sync docs-manifest:generate; do
     ( bun run "$g" >"/tmp/ship-gen-${g//:/-}.log" 2>&1; echo "$? $g" >>/tmp/ship-gen-results ) &
   done
   wait
 
@@ -100,7 +100,7 @@ sim --version
 Then upgrade with the package manager you installed it with — using a different
 one installs a second copy instead of replacing the executable on your `PATH`:
 
-<Tabs items={['npm', 'pnpm', 'bun']}>
+<Tabs items={['npm', 'pnpm', 'bun', 'yarn']}>
   <Tab value="npm">
     ```bash
     npm install -g sim@latest
@@ -116,6 +116,11 @@ one installs a second copy instead of replacing the executable on your `PATH`:
     bun add -g sim@latest
     ```
   </Tab>
+  <Tab value="yarn">
+    ```bash
+    yarn global add sim@latest
+    ```
+  </Tab>
 </Tabs>
 
 The CLI can also tell you this through a cached daily check on eligible
 
@@ -24,7 +24,9 @@ Sim uses Redis as a message bus and shared cache. Both deployments ship it by de
   With more than one app or realtime replica and no `REDIS_URL`, users on different pods stop seeing each other's edits and live status updates. Beyond one startup log line noting single-pod mode, nothing is logged — the app looks healthy and quietly loses events. Treat Redis as mandatory the moment `replicaCount` exceeds 1.
 </Callout>
 
-Everything Sim keeps in Redis is cache, coordination state, or an in-flight event — never committed data, which lives in PostgreSQL and object storage. Persistence is therefore not required. Losing or restarting the instance is not free, though: cancellation markers and the cross-pod half of execution streaming live here, so active runs stop streaming and a cancellation issued across the gap may not land. Completed work is unaffected.
+Everything Sim keeps in Redis is cache, coordination state, or an in-flight event — never committed data, which lives in PostgreSQL and object storage. Persistence is therefore not required. Losing or restarting the instance is not free, though: cancellation markers and the cross-pod half of execution streaming live here, so active runs stop streaming and a cancellation issued across the gap may not land.
+
+One exception is worth planning for. High-volume webhook triggers claim their idempotency keys in Redis by default, trading a narrow durability window for the throughput. A restart drops those claims, so a provider that retries a webhook Sim already finished can have it execute a second time. Flows where a repeat is not acceptable — anything touching money, billing, or compliance — should claim in PostgreSQL instead, which fate-shares the claim with the transaction; see `forceStorage: 'database'` on the idempotency service. Enabling Redis persistence narrows the window but does not close it.
 
 ## Configuration
 
 
@@ -13,7 +13,7 @@ export const GET = defineInternalJsonRoute({
   auth: internalSessionAuth,
   operation: knowledgeOperations.listWorkspaceMemberConnectors,
   rateLimit: internalRateLimits.none({ reason: 'One small read per visit to the Search tab' }),
-  errorPolicy: internalKnowledgeErrorPolicies.connectors,
+  errorPolicy: internalKnowledgeErrorPolicies.memberConnectors,
   mapInput: ({ query }) => ({ workspaceId: query.workspaceId }),
   useCase: listWorkspaceMemberConnectors,
   present: ({ connectors }) => ({ success: true as const, data: connectors }),
 
@@ -543,23 +543,26 @@ export function Home({ chatId, userName, userId }: HomeProps) {
     ]
   )
 
+  /** An emptied search box returns to the sources; a send in any other mode has no search to clear. */
+  const clearSearch = useCallback(() => {
+    if (searchQueryValue !== null) setSearchQuery('')
+  }, [searchQueryValue, setSearchQuery])
+
   /**
    * A queued message re-enters the composer in the mode it was written in: an
    * Assistant question edits as an Assistant question, and never as a Search,
-   * which submits nothing and would leave the edit stranded.
+   * which submits nothing and would leave the edit stranded. Leaving Search
+   * drops the query as it does everywhere else; a live one would put the
+   * composer straight back into Search and strand the edit anyway.
    */
   const restoreQueuedMode = useCallback(
     (requestMode: QueuedMessage['requestMode']) => {
+      clearSearch()
       void setComposerMode(requestMode === 'ask' ? 'assistant' : 'build')
     },
-    [setComposerMode]
+    [clearSearch, setComposerMode]
   )
 
-  /** An emptied search box returns to the sources; a send in any other mode has no search to clear. */
-  const clearSearch = useCallback(() => {
-    if (searchQueryValue !== null) setSearchQuery('')
-  }, [searchQueryValue, setSearchQuery])
-
   /**
    * Summarize or Answer on a result: switch to Assistant and hand the question
    * to it. The submit reads the mode from this render, so it is sent as an
 
@@ -12,7 +12,10 @@ import {
   WorkspaceApiKeyScopeAuthorizationError,
 } from '@/lib/core/application'
 import { OrchestrationError } from '@/lib/core/orchestration/types'
-import { v2KnowledgeErrorPolicies } from '@/lib/knowledge/api/route-policies'
+import {
+  internalKnowledgeErrorPolicies,
+  v2KnowledgeErrorPolicies,
+} from '@/lib/knowledge/api/route-policies'
 
 describe('v2 knowledge error policies', () => {
   it.each([
@@ -70,3 +73,21 @@ describe('v2 knowledge error policies', () => {
     })
   })
 })
+
+describe('internal knowledge error policies', () => {
+  it('conceals a knowledge-base-scoped connector authorization failure', () => {
+    const projected = internalKnowledgeErrorPolicies.connectors.project(
+      new NoWorkspaceAccessError()
+    )
+    expect(projected?.status).toBe(404)
+    expect(projected?.body).toMatchObject({ error: 'Knowledge base not found' })
+  })
+
+  it('does not conceal the workspace-scoped member-connector listing', () => {
+    const projected = internalKnowledgeErrorPolicies.memberConnectors.project(
+      new NoWorkspaceAccessError()
+    )
+    expect(projected?.status).not.toBe(404)
+    expect(JSON.stringify(projected?.body)).not.toContain('Knowledge base not found')
+  })
+})
@@ -110,6 +110,8 @@ export const internalKnowledgeErrorPolicies = {
     internalKnowledgeErrorPolicy('Failed to process knowledge tag request')
   ),
   connectors: concealKnowledgeBase(internalKnowledgeErrorPolicy('Internal server error')),
+  /** Workspace-scoped like the bulk routes above, so likewise not concealed. */
+  memberConnectors: internalKnowledgeErrorPolicy('Failed to fetch member connectors'),
   uploads: concealKnowledgeBase(internalKnowledgeUploadErrorPolicy),
 } as const