@@ -46,6 +46,20 @@ These reads/mutations are **not** anti-patterns and stay as-is:
 - **Route navigations** — `router.push('/path/[id]?folderId=x')` that changes the route *path*, not just the current query. A nuqs setter only mutates the query on the current path; cross-path navigation stays on `router`.
 - **Read-once auth / redirect signals** — `token`, `callbackUrl`, `redirect`, `error`, `invite_flow`, `new` (invite signup flow), `upgraded`, `redirect_workflow`, etc. These are navigation signals consumed once (often read-then-strip), not synced view-state. Leave them on `useSearchParams`. Key names are per-surface: files' `new` is a genuine nuqs param (`files/search-params.ts`), while invite's `new` is a one-shot signup signal.
 
+### Remembered list-preference exception
+
+Files, Tables, and Knowledge may persist their last-used filter/sort snapshot through
+`useResourceListPreferences`. This is a fallback preference, not a second live source of truth:
+
+- nuqs remains authoritative while the module is open.
+- Zustand is consulted once on a clean module entry, after persisted state hydrates.
+- An explicit URL filter/sort parameter wins even when it resolves to the module default. The
+  complete resolved URL snapshot becomes the remembered value; omitted fields use URL defaults
+  rather than merging with storage.
+- Explicit filter/sort gestures commit the same complete snapshot to nuqs and Zustand together.
+- Never mirror subsequent URL changes with a synchronization effect or `popstate` listener.
+- Search and folder navigation remain URL-only and are excluded from the persisted snapshot.
+
 ## Per-feature `search-params.ts` — single source of truth
 
 Co-locate a `search-params.ts` next to the feature. Export the parser map (and shared options). Both the client (`useQueryStates`/`useQueryState`) and any server component (`createSearchParamsCache` from `nuqs/server`) import from this one file. Import parsers from `nuqs/server` so the module is safe to import in both client and server contexts.
 
@@ -182,17 +182,17 @@ jobs:
       # without `--coverage`. See the Codecov note below.
       #
       # apps/sim runs only its first shard here; `test-shard` below runs the
-      # other. That suite is bound by the single Vite server thread that feeds
+      # others. That suite is bound by the single Vite server thread that feeds
       # every worker — wall time is flat from 4 to 13 workers — so a bigger
-      # runner buys nothing and a second runner halves it.
+      # runner buys nothing and each extra runner takes a proportional slice.
       - name: Run tests
         env:
           NODE_OPTIONS: '--no-warnings --max-old-space-size=8192'
           NEXT_PUBLIC_APP_URL: 'https://www.sim.ai'
           DATABASE_URL: 'postgresql://postgres:postgres@localhost:5432/simstudio'
           ENCRYPTION_KEY: '0000000000000000000000000000000000000000000000000000000000000000' # dummy key for CI only
           TURBO_CACHE_DIR: .turbo
-          SIM_TEST_SHARD: 1/2
+          SIM_TEST_SHARD: 1/3
         run: bun run test
 
       - name: Check schema and migrations are in sync
@@ -208,15 +208,20 @@ jobs:
           fi
           echo "✅ Schema and migrations are in sync"
 
-  # The second half of apps/sim's Vitest suite. Everything else — lint, the
-  # audits, type-check, the other workspaces' suites — lives in `test-build`
-  # with shard 1; this job exists only because that suite cannot go faster on
-  # one machine (see the "Run tests" note there). The Turbo cache disk gets
-  # its own key so the two shards' entries do not evict each other.
+  # The remaining shards of apps/sim's Vitest suite. Everything else — lint,
+  # the audits, type-check, the other workspaces' suites — lives in
+  # `test-build` with shard 1; these jobs exist only because that suite cannot
+  # go faster on one machine (see the "Run tests" note there). Three shards
+  # put each runner at roughly the fixed cost of checkout + install. The Turbo
+  # cache disk gets its own key so the shards' entries do not evict each other.
   test-shard:
-    name: Test (shard 2)
+    name: Test (shard ${{ matrix.shard }})
     runs-on: ${{ (vars.CI_PROVIDER == '' || vars.CI_PROVIDER == 'blacksmith') && 'blacksmith-8vcpu-ubuntu-2404' || 'ubuntu-latest' }}
     timeout-minutes: 15
+    strategy:
+      fail-fast: false
+      matrix:
+        shard: [2, 3]
 
     steps:
       - name: Checkout code
@@ -250,7 +255,7 @@ jobs:
         uses: ./.github/actions/cache-mount
         with:
           provider: ${{ vars.CI_PROVIDER }}
-          key: ${{ github.repository }}-turbo-cache-shard-${{ github.event_name }}${{ github.event.pull_request.head.repo.fork && '-fork' || '' }}
+          key: ${{ github.repository }}-turbo-cache-shard-${{ matrix.shard }}-${{ github.event_name }}${{ github.event.pull_request.head.repo.fork && '-fork' || '' }}
           path: ./.turbo
 
       - name: Install dependencies
@@ -259,14 +264,14 @@ jobs:
       - name: Install ripgrep
         run: command -v rg || (sudo apt-get update && sudo apt-get install -y ripgrep)
 
-      - name: Run tests (apps/sim shard 2/2)
+      - name: Run tests (apps/sim shard ${{ matrix.shard }}/3)
         env:
           NODE_OPTIONS: '--no-warnings --max-old-space-size=8192'
           NEXT_PUBLIC_APP_URL: 'https://www.sim.ai'
           DATABASE_URL: 'postgresql://postgres:postgres@localhost:5432/simstudio'
           ENCRYPTION_KEY: '0000000000000000000000000000000000000000000000000000000000000000' # dummy key for CI only
           TURBO_CACHE_DIR: .turbo
-          SIM_TEST_SHARD: 2/2
+          SIM_TEST_SHARD: ${{ matrix.shard }}/3
         run: bunx turbo run test --filter=@sim/app
 
   # Next.js production build, in parallel with lint + tests. Sticky disks are
 
@@ -16,7 +16,7 @@ The File block is a built-in Sim block for working with files stored in the work
 With the File block, you can:
 
 - **Read and extract content**: Load workspace file objects and extract their text content
-- **Search workspace content**: Find literal text across indexed active workspace files with bounded line-level results
+- **Search workspace content**: Match a regular expression, or an exact piece of text, against the indexed lines of active workspace files with bounded line-level results
 - **Fetch from URLs**: Retrieve and parse files from external URLs with custom headers
 - **Write and append**: Create new workspace files or append content to existing ones
 - **Compress and decompress**: Bundle files into a .zip archive or extract an archive into the workspace
@@ -70,13 +70,14 @@ Extract the text content of one or more workspace files from selected file objec
 
 ### File Search
 
-Search indexed text across active workspace files using literal smart-case substring matching.
+Search the indexed text of active workspace files for lines matching a regular expression, and return each matching line once with its file ID and line number. Coverage is what the index currently holds, so check "complete" and "indexStatus" before concluding that something is absent.
 
 #### Input
 
 | Parameter | Type | Required | Description |
 | --------- | ---- | -------- | ----------- |
-| `query` | string | Yes | Literal text to find \(3-512 characters\). Uppercase Unicode letters make matching case-sensitive. |
+| `query` | string | Yes | A regular expression matched against each line, 3-512 characters. Supports "." "*" "+" "?" "\{n,m\}" and their lazy forms, character classes such as "\[a-z\]" and "\[^0-9\]", the classes \d \w \s and \D \W \S, alternation "\|", groups "\(...\)" and "\(?:...\)", the anchors "^" and "$", and the word boundary \b. Lookahead, lookbehind, backreferences, named groups, inline flags such as "\(?i\)", \p\{...\} and POSIX "\[\[:alpha:\]\]" classes are not supported, and a pattern cannot span a line break. The pattern must contain at least 3 consecutive literal characters that every match will include — write "error \d+" rather than "\w+ \d+". Escape any metacharacter you mean literally. Matching is case-insensitive until the pattern contains an uppercase letter you are searching for; uppercase inside an escape or a character class, such as \D or \[A-Z\], does not make it case-sensitive. When the workflow builder sets Match to exact instead, the query is matched verbatim and no metacharacter needs escaping. |
+| `mode` | string | No | How the query is read, chosen by the workflow builder: "regex" \(default\) as a regular expression, or "exact" as verbatim text. |
 | `maxResults` | number | No | Hard result cap configured by the workflow builder \(1-200, default 50\). |
 
 #### Output
 
@@ -109,7 +109,7 @@ Live tool-call chips stream for **OpenAI, Anthropic, Azure Anthropic, Google, Ve
 | Provider | Streamed thinking | Models |
 |----------|-------------------|--------|
 | OpenAI | Summaries only — Requires OpenAI organization verification; falls back to no summaries. | `gpt-5.6-sol`, `gpt-5.6-terra`, `gpt-5.6-luna`, `gpt-5.5-pro`, `gpt-5.5`, `gpt-5.4-pro`, `gpt-5.4`, `gpt-5.4-mini`, `gpt-5.4-nano`, `gpt-5.2-pro`, `gpt-5.2`, `gpt-5.1`, `gpt-5-pro`, `gpt-5`, `gpt-5-mini`, `gpt-5-nano`, `o4-mini`, `o3`, `o3-mini`, `o1` |
-| Anthropic | Summaries only — These generations omit full thinking; Sim requests summarized thinking on streaming runs. | `claude-fable-5`, `claude-sonnet-5`, `claude-opus-5`, `claude-opus-4-8`, `claude-opus-4-7`, `claude-opus-4-6`, `claude-sonnet-4-6`, `claude-opus-4-5`, `claude-opus-4-1`, `claude-sonnet-4-5`, `claude-haiku-4-5` |
+| Anthropic | Summaries only — These generations omit full thinking; Sim requests summarized thinking on streaming runs. | `claude-fable-5-1`, `claude-fable-5`, `claude-sonnet-5`, `claude-opus-5`, `claude-opus-4-8`, `claude-opus-4-7`, `claude-opus-4-6`, `claude-sonnet-4-6`, `claude-opus-4-5`, `claude-opus-4-1`, `claude-sonnet-4-5`, `claude-haiku-4-5` |
 | Azure OpenAI | Summaries only — Requires OpenAI organization verification; falls back to no summaries. | `azure/gpt-5.4`, `azure/gpt-5.4-mini`, `azure/gpt-5.4-nano`, `azure/gpt-5.2`, `azure/gpt-5.1`, `azure/gpt-5.1-codex`, `azure/gpt-5`, `azure/gpt-5-mini`, `azure/gpt-5-nano`, `azure/o3`, `azure/o4-mini` |
 | Azure Anthropic | Summaries only — These generations omit full thinking; Sim requests summarized thinking on streaming runs. | `azure-anthropic/claude-opus-4-6`, `azure-anthropic/claude-opus-4-5`, `azure-anthropic/claude-sonnet-4-5`, `azure-anthropic/claude-opus-4-1`, `azure-anthropic/claude-haiku-4-5` |
 | Google | Summaries only | `gemini-3.6-flash`, `gemini-3.5-flash-lite`, `gemini-3.5-flash`, `gemini-3.1-pro-preview`, `gemini-3.1-flash-lite`, `gemini-3-flash-preview`, `gemini-2.5-pro`, `gemini-2.5-flash`, `gemini-2.5-flash-lite` |
 
@@ -42,7 +42,6 @@
   "devDependencies": {
     "@sim/tsconfig": "workspace:*",
     "@tailwindcss/postcss": "^4.0.12",
-    "@types/mdx": "^2.0.13",
     "@types/node": "24.2.1",
     "@types/react": "^19.1.2",
     "@types/react-dom": "^19.0.4",
 
@@ -1,5 +1,5 @@
 /**
- * @vitest-environment jsdom
+ * @vitest-environment node
  *
  * Guards against drift between the code-block language picker and the Prism grammars actually
  * registered by CodeBlockHighlight: every selectable language must have a registered grammar, or it
 
@@ -157,41 +157,6 @@ describe('agent-stream applier', () => {
     expect(freshText).toContain('Gamma paragraph')
   })
 
-  it('preserves a concurrent peer edit to a region the agent snapshot does not include', () => {
-    // This is the core "AI as a CRDT peer" guarantee: the agent relays only its OWN delta (computed
-    // against a private shadow), never a whole-document reconcile that would revert a peer's edit.
-    const { editor, doc } = track(makeCollabEditor())
-
-    const session = beginAgentStream(editor)!
-    applyAgentStreamFrame(editor, session, 'Alpha paragraph.\n\nBeta paragraph.')
-
-    // A peer edits the FIRST paragraph directly on the shared doc — the agent's later snapshot still
-    // carries the ORIGINAL first paragraph (it was built from the base, before this edit).
-    const peer = new Y.Doc()
-    Y.applyUpdate(peer, Y.encodeStateAsUpdate(doc))
-    const peerFrag = peer.getXmlFragment('default')
-    peer.transact(() => {
-      const firstPara = peerFrag.get(0) as Y.XmlElement
-      const textNode = firstPara.get(0) as Y.XmlText
-      textNode.insert(textNode.toString().length, ' EDITED')
-    })
-    Y.applyUpdate(doc, Y.encodeStateAsUpdate(peer, Y.encodeStateVector(doc)))
-    peer.destroy()
-
-    // The agent appends a third paragraph. Its snapshot's first paragraph is the stale original, but the
-    // shadow-relayed delta only inserts the new paragraph — so the peer's " EDITED" must survive.
-    applyAgentStreamFrame(
-      editor,
-      session,
-      'Alpha paragraph.\n\nBeta paragraph.\n\nGamma paragraph.'
-    )
-    endAgentStream(session)
-
-    const live = doc.getXmlFragment('default').toString()
-    expect(live).toContain('EDITED')
-    expect(live).toContain('Gamma paragraph')
-  })
-
   it('reuses cached binding metadata across frames, still emitting minimal per-frame deltas', () => {
     // The binding `meta` is built ONCE (first frame) and reused — `updateYFragment` maintains it in place,
     // so we skip an O(doc) `initProseMirrorDoc` rebuild per frame. This guards that caching preserves the
 
@@ -1,5 +1,5 @@
 /**
- * @vitest-environment jsdom
+ * @vitest-environment node
  *
  * Dragging an image to reposition it inside a document must MOVE it, not import it again.
  *
 
@@ -1,5 +1,5 @@
 /**
- * @vitest-environment jsdom
+ * @vitest-environment node
  */
 import { describe, expect, it } from 'vitest'
 import {
 
@@ -2,7 +2,7 @@
  * @vitest-environment jsdom
  */
 import { Editor } from '@tiptap/core'
-import { afterEach, describe, expect, it } from 'vitest'
+import { afterAll, describe, expect, it } from 'vitest'
 import { createMarkdownContentExtensions } from './extensions'
 import { parseMarkdownToDoc, serializeMarkdownBody, splitMarkdownBlocks } from './markdown-parse'
 import { isRoundTripSafe } from './round-trip-safety'
@@ -12,19 +12,20 @@ const isEmptyPara = (n: { type?: string; content?: unknown[] }): boolean =>
   n.type === 'paragraph' && !n.content?.length
 
 let editor: Editor | null = null
-afterEach(() => {
+afterAll(() => {
   editor?.destroy()
   editor = null
 })
 
-/** The current whole-document path: parse markdown in one shot, serialize back. */
+/**
+ * The current whole-document path: parse markdown in one shot, serialize back. One editor serves
+ * every call — `setContent` replaces the document wholesale, so a fresh instance per call only adds
+ * the cost of building the view, which the property tests below paid hundreds of times over.
+ */
 function oneShot(body: string): string {
-  editor = new Editor({ extensions: createMarkdownContentExtensions() })
+  editor ??= new Editor({ extensions: createMarkdownContentExtensions() })
   editor.commands.setContent(body, { contentType: 'markdown' })
-  const out = editor.getMarkdown()
-  editor.destroy()
-  editor = null
-  return out
+  return editor.getMarkdown()
 }
 
 /**
Original file line number	Diff line number	Diff line change
`@@ -1,5 +1,5 @@`
`1`	`1`	`/**`
`2`		`- * @vitest-environment jsdom`
	`2`	`+ * @vitest-environment node`
`3`	`3`	`*`
`4`	`4`	`* Guards against drift between the code-block language picker and the Prism grammars actually`
`5`	`5`	`* registered by CodeBlockHighlight: every selectable language must have a registered grammar, or it`