fix: tell the user when a draft conflict has stopped their edits saving

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
feat: drafts follow their item through a move; behind means base ≠ head (#10577 )
2026-09-19 00:02:28 +00:00 · 2026-09-18 21:36:07 +02:00 · 2026-09-18 20:54:23 +02:00 · 2026-09-18 20:50:23 +02:00 · 2026-09-18 20:48:07 +02:00 · 2026-09-18 20:47:44 +02:00
439 changed files with 20422 additions and 17322 deletions
@@ -1,3 +1,3 @@
 {
-  ".": "1.813.0"
+  ".": "1.815.0"
 }
@@ -1,5 +1,65 @@
 # Changelog

+## [1.815.0](https://github.com/windmill-labs/windmill/compare/v1.814.0...v1.815.0) (2026-09-18)
+
+
+### Features
+
+* add a workspace toggle that adds its admins and developers to new forks ([#11215](https://github.com/windmill-labs/windmill/issues/11215)) ([9d335de](https://github.com/windmill-labs/windmill/commit/9d335de87a4dbaa51038d55afe8d980761dcdfaf))
+* add an instance setting to refuse a token in MCP URLs ([#11162](https://github.com/windmill-labs/windmill/issues/11162)) ([37e493a](https://github.com/windmill-labs/windmill/commit/37e493ae66ed5c000ecac492d60fc0fdf4bda71f))
+* **ai-chat:** render get_run through the run tools' card ([#11204](https://github.com/windmill-labs/windmill/issues/11204)) ([6f9c4dc](https://github.com/windmill-labs/windmill/commit/6f9c4dc29455d13b0e64af05c2d6aa8bd5ff4fd6))
+* **ai-chat:** tell the chat which kind of app it is looking at ([#11208](https://github.com/windmill-labs/windmill/issues/11208)) ([c8c06d8](https://github.com/windmill-labs/windmill/commit/c8c06d8f79774abf109192e71a8b6fc37c7937ba))
+* attach files to a flow chat message ([#11185](https://github.com/windmill-labs/windmill/issues/11185)) ([c4c9677](https://github.com/windmill-labs/windmill/commit/c4c9677982b75c63d98ebf85b1904e0c341ba957))
+* **auth:** 2 h login links and a click-to-sign-in page for emailed ones ([#11203](https://github.com/windmill-labs/windmill/issues/11203)) ([5639187](https://github.com/windmill-labs/windmill/commit/5639187fec6d517a72e82df49d63d7438301127c))
+* cap user token expiration with an instance setting ([#11159](https://github.com/windmill-labs/windmill/issues/11159)) ([9320312](https://github.com/windmill-labs/windmill/commit/9320312eac56f944c4d31504601293ab4e816ccc))
+* flow chat job-backed detail, smooth streaming and answer chrome ([#11186](https://github.com/windmill-labs/windmill/issues/11186)) ([e2a91ca](https://github.com/windmill-labs/windmill/commit/e2a91ca2b141a0bee64ddfafe3cbf3ff86b6fb35))
+* put a data table's connection under Postgres roles ([#11020](https://github.com/windmill-labs/windmill/issues/11020)) ([0e807fb](https://github.com/windmill-labs/windmill/commit/0e807fb1dd80d7536ec144cd49445abc7961e504))
+* render an AI agent result as its answer, not as raw JSON ([#11051](https://github.com/windmill-labs/windmill/issues/11051)) ([a089928](https://github.com/windmill-labs/windmill/commit/a08992834d45d0211336f4fc32c3421646ca47c5))
+* support $flow_expr[...] dynamic tags on flow steps ([#11170](https://github.com/windmill-labs/windmill/issues/11170)) ([48f0025](https://github.com/windmill-labs/windmill/commit/48f00259c5e7361d3553dbcb809461e1cde96f8f))
+
+
+### Bug Fixes
+
+* **frontend:** inline only the package version, not the whole package.json ([#11191](https://github.com/windmill-labs/windmill/issues/11191)) ([813e486](https://github.com/windmill-labs/windmill/commit/813e486e166ac6215364817a7732b66f8dc1d463))
+* keep instance groups when editing auto-invite ([#11217](https://github.com/windmill-labs/windmill/issues/11217)) ([df61dea](https://github.com/windmill-labs/windmill/commit/df61dea5fa8b18d1e0044dc0db6702b053d5119f))
+* key the large root font size on screen width, not window width ([#11216](https://github.com/windmill-labs/windmill/issues/11216)) ([3b4e13d](https://github.com/windmill-labs/windmill/commit/3b4e13d1c564c6195e30b55b0671f7533e3ce408))
+* re-encrypt git sync secrets on workspace key rotation ([#11218](https://github.com/windmill-labs/windmill/issues/11218)) ([f0d66a4](https://github.com/windmill-labs/windmill/commit/f0d66a42eba27502767423af0c08b2ca1f16e8df))
+* re-point cloned fork identities that name nobody in the fork ([#11161](https://github.com/windmill-labs/windmill/issues/11161)) ([9690c44](https://github.com/windmill-labs/windmill/commit/9690c4462cf264a5577b87d07d465b5442b4e09d))
+* register the job token with the sensitive log masking system ([#10943](https://github.com/windmill-labs/windmill/issues/10943)) ([53afecd](https://github.com/windmill-labs/windmill/commit/53afecd4588247bc1812d3e68a30db1f3c3b2724))
+* show the New menu's description panel only on hover ([#11199](https://github.com/windmill-labs/windmill/issues/11199)) ([ecd0a6c](https://github.com/windmill-labs/windmill/commit/ecd0a6c77bc3a057b8072dbb0aca731e8bd3d882))
+* stop picker listing a draft twice when drafts share a friendly path ([#11214](https://github.com/windmill-labs/windmill/issues/11214)) ([72507d5](https://github.com/windmill-labs/windmill/commit/72507d52a021de8101794ee8f4a3392be1ba20c7))
+
+
+### Performance Improvements
+
+* count completed jobs on the v2_job_completed index with a timeout ([#11211](https://github.com/windmill-labs/windmill/issues/11211)) ([d1a2536](https://github.com/windmill-labs/windmill/commit/d1a25360b070994a31fb033707dbfde1fb553939))
+
+## [1.814.0](https://github.com/windmill-labs/windmill/compare/v1.813.0...v1.814.0) (2026-09-17)
+
+
+### Features
+
+* **ai-chat:** add list_workers and list_data_metrics global tools ([#11143](https://github.com/windmill-labs/windmill/issues/11143)) ([e954d33](https://github.com/windmill-labs/windmill/commit/e954d33613e4ff5027667eb8f646615d9bbd499d))
+* **ai-chat:** merge get_job_logs and get_flow_run_details into get_run ([#11172](https://github.com/windmill-labs/windmill/issues/11172)) ([5bb37ca](https://github.com/windmill-labs/windmill/commit/5bb37ca3388666fba72c55534e37f37bb3e9299e))
+* allow git sync auto-pull, promotion and PRs on Pro licenses ([#11173](https://github.com/windmill-labs/windmill/issues/11173)) ([02e47de](https://github.com/windmill-labs/windmill/commit/02e47de8b4c4f3f54753aabf8c67bc8e71ffb957))
+* badge chat-input flows on the home list ([#11164](https://github.com/windmill-labs/windmill/issues/11164)) ([3d08197](https://github.com/windmill-labs/windmill/commit/3d0819718221f885b61e73d02b43dcc853c7d02a))
+* collect flow conversations and agent memory once their last message goes ([#11178](https://github.com/windmill-labs/windmill/issues/11178)) ([23c24a9](https://github.com/windmill-labs/windmill/commit/23c24a9688d4c8c462f53221334d538280f16bca))
+* flow chat model picker on a shared model-settings component ([#11187](https://github.com/windmill-labs/windmill/issues/11187)) ([189793c](https://github.com/windmill-labs/windmill/commit/189793c2e4db7f1c853695ebcc895c1ec82ed19f))
+* keep flow inputs and seed the agent when chat mode is enabled ([#11177](https://github.com/windmill-labs/windmill/issues/11177)) ([68f2248](https://github.com/windmill-labs/windmill/commit/68f2248018fc218a090bf939e1eb22ff97d5bc22))
+* let plan mode search and read connected mcp servers ([#11205](https://github.com/windmill-labs/windmill/issues/11205)) ([5371519](https://github.com/windmill-labs/windmill/commit/5371519f0f5ce7750982dcdb374dca72115902e7))
+* let test_run_flow name the conversation of a chat-mode test run ([#11198](https://github.com/windmill-labs/windmill/issues/11198)) ([6e1ef93](https://github.com/windmill-labs/windmill/commit/6e1ef93f329cb396ffc3df3304d592e8fa0e0e71))
+* managed memory with an inherited or custom memory id per step ([#11118](https://github.com/windmill-labs/windmill/issues/11118)) ([c297ed0](https://github.com/windmill-labs/windmill/commit/c297ed0052d998fb8f063faa2a36c6eb03e327be))
+* render the flow chat through the shared session chat components ([#11175](https://github.com/windmill-labs/windmill/issues/11175)) ([a9ec0ae](https://github.com/windmill-labs/windmill/commit/a9ec0aec3ac0c6b0f7919d0eb2168816923826d7))
+* show flow step detail inside the graph tab on narrow detail layouts ([#11168](https://github.com/windmill-labs/windmill/issues/11168)) ([64dffe6](https://github.com/windmill-labs/windmill/commit/64dffe6106ad6a55b61a423c855a4b5b0cef533e))
+* store mcp tool call, result and reasoning on flow conversation rows ([#11176](https://github.com/windmill-labs/windmill/issues/11176)) ([a571117](https://github.com/windmill-labs/windmill/commit/a571117f3fd2cef14c920770645c60ee358fdfdd))
+* tell test flow conversations from deployed ones and rename a chat ([#11179](https://github.com/windmill-labs/windmill/issues/11179)) ([4eab995](https://github.com/windmill-labs/windmill/commit/4eab995cf7cf091a5e4640da4cb77e0921bb7fdf))
+
+
+### Bug Fixes
+
+* disable a schedule whose cron has no run left instead of panicking ([#11195](https://github.com/windmill-labs/windmill/issues/11195)) ([381d447](https://github.com/windmill-labs/windmill/commit/381d4470ef699ea82283742132e56556b95d2bd2))
+* skip expiry notifications for app embed and SDK tokens ([#11169](https://github.com/windmill-labs/windmill/issues/11169)) ([9d348f8](https://github.com/windmill-labs/windmill/commit/9d348f84c7830f36b6153472556fd70e3d84cd24))
+
 ## [1.813.0](https://github.com/windmill-labs/windmill/compare/v1.812.0...v1.813.0) (2026-09-16)


@@ -175,6 +175,10 @@ the decrypted value, exactly as against a real backend. The chat's read path pas
 Seed a recognizable secret (the existing fixture uses `sk_live_do_not_leak_me`) and
 assert it via `valueExcludes` to catch a leak.

+`toolExpect.toolCallArgs` entries support `sharedByAtLeast: <n>`: at least `n` recorded
+calls to that tool must carry the same non-blank string in the field. Use it for calls that
+have to share an identifier, like two test runs of one chat conversation.
+
 `toolExpect.toolCallArgs` entries additionally support `fieldMustBeAbsent: true`: no
 recorded call to that tool may pass the field at all (an explicit `null` counts as
 passing it). Use it for partial-update tools, where supplying a field the model could
@@ -1,5 +1,5 @@
-import { randomUUID } from 'node:crypto'
 import type { BackendValidationSettings } from '../../core/backendValidation'
+import { buildWorkspaceId } from './workspaceId'

 interface CompletedJobResultMaybe {
 	completed: boolean
@@ -24,7 +24,6 @@ export interface CompletedPreviewJob {
 const tokenCache = new Map<string, Promise<string>>()
 const sharedWorkspaceQueue = new Map<string, Promise<void>>()
 const managedSharedWorkspacePrefixes = ['f/evals/']
-const DEFAULT_WORKSPACE_PREFIX = 'ai-evals'

 export class BackendPreviewClient {
 	constructor(private readonly settings: BackendValidationSettings) {}
@@ -441,16 +440,6 @@ async function withSharedWorkspaceLock<T>(workspaceId: string, body: () => Promi
 	}
 }

-function buildWorkspaceId(caseId: string, attempt: number): string {
-	const caseSlug = caseId
-		.toLowerCase()
-		.replace(/[^a-z0-9-]+/g, '-')
-		.replace(/^-+|-+$/g, '')
-		.slice(0, 30)
-	const suffix = randomUUID().slice(0, 8)
-	return `${DEFAULT_WORKSPACE_PREFIX}-${caseSlug || 'case'}-a${attempt}-${suffix}`
-}
-
 function extractFolderName(path: string): string | null {
 	if (!path.startsWith('f/')) {
 		return null
@@ -11,6 +11,7 @@ import type {
 	Script
 } from '../../../frontend/src/lib/gen'
 import type {
+	DataMetric,
 	DataTableTables,
 	DataTableTableSchema,
 	EndpointTool,
@@ -48,12 +49,16 @@ export interface BenchmarkWorkspaceFlow {
 export interface BenchmarkWorkspaceApp {
 	path: string
 	summary: string
+	/** Defaults to true. Set false for a drag-and-drop app, which the chat can list
+	 * and read but has no tool to edit — its value is a grid, not files. */
+	rawApp?: boolean
 	value: {
-		files: Record<string, string>
-		runnables: Record<string, unknown>
+		files?: Record<string, string>
+		runnables?: Record<string, unknown>
 		data?: unknown
 		policy?: unknown
 		custom_path?: unknown
+		[key: string]: unknown
 	}
 }

@@ -112,6 +117,9 @@ export interface BenchmarkWorkspaceRunnables {
 	aiProviders?: BenchmarkWorkspaceAiProvider[]
 	resources?: BenchmarkWorkspaceResource[]
 	datatables?: BenchmarkDatatableSeed[]
+	/** DuckLake catalog names, as `list_ducklakes` reports them. */
+	ducklakes?: string[]
+	dataMetrics?: DataMetric[]
 	jobs?: BenchmarkWorkspaceJob[]
 }

@@ -673,6 +681,27 @@ export function listBenchmarkDatatables(workspace: string): DataTableTables[] |
 	}))
 }

+// ============= DuckLake catalogs and declared metrics =============
+
+/** Seeded DuckLake names, or `null` for a non-benchmark workspace. */
+export function listBenchmarkDucklakes(workspace: string): string[] | null {
+	const runnables = benchmarkWorkspaceRunnables.get(workspace)
+	return runnables ? (runnables.ducklakes ?? []) : null
+}
+
+/**
+ * Seeded metric declarations, or `null` for a non-benchmark workspace.
+ *
+ * The `table` / `path_prefix` filters are ignored: which rows a filter selects is
+ * `canonical_table_path`'s business and is pinned by `ducklakeTools.test.ts`.
+ * Re-deriving it here would give the eval its own copy of that spec to drift from,
+ * and the case this serves measures whether the model reaches for the tool at all.
+ */
+export function listBenchmarkDataMetrics(workspace: string): DataMetric[] | null {
+	const runnables = benchmarkWorkspaceRunnables.get(workspace)
+	return runnables ? (runnables.dataMetrics ?? []) : null
+}
+
 export function getBenchmarkDatatableSchema(input: {
 	workspace: string
 	datatableName: string
@@ -840,6 +869,29 @@ export function runBenchmarkFlowByPath(input: {
 	})
 }

+/**
+ * Mirror `JobService.runFlowPreview` for benchmark workspaces, including the server's
+ * refusal of a chat-enabled flow run that names no conversation (`memory_id`).
+ */
+export function runBenchmarkFlowPreview(input: {
+	workspace: string
+	memoryId?: string
+	requestBody?: { path?: string; value?: { chat_input_enabled?: boolean }; args?: unknown }
+}): string {
+	if (input.requestBody?.value?.chat_input_enabled && !input.memoryId) {
+		throw new Error('Bad request: memory_id is required for chat-enabled flows')
+	}
+	const args = (input.requestBody?.args ?? {}) as Record<string, unknown>
+	return createBenchmarkCompletedJob({
+		workspace: input.workspace,
+		jobKind: 'flowpreview',
+		success: true,
+		args,
+		result: { path: input.requestBody?.path, args, mocked: true },
+		logs: 'Mock benchmark flow preview completed successfully.'
+	})
+}
+
 export function previewBenchmarkSchedule(input: {
 	requestBody?: Record<string, unknown>
 }): Record<string, unknown> {
@@ -946,7 +998,7 @@ function buildBenchmarkListableApp(app: BenchmarkWorkspaceApp): ListableApp {
 		extra_perms: {},
 		edited_at: BENCHMARK_TIMESTAMP,
 		execution_mode: 'viewer',
-		raw_app: true
+		raw_app: app.rawApp ?? true
 	}
 }

@@ -964,7 +1016,7 @@ function buildBenchmarkApp(app: BenchmarkWorkspaceApp): AppWithLastVersion {
 		execution_mode: 'viewer',
 		extra_perms: {},
 		custom_path: app.value.custom_path as string | undefined,
-		raw_app: true
+		raw_app: app.rawApp ?? true
 	}
 }

@@ -76,7 +76,9 @@ vi.mock('$lib/gen', async () => {
 		listBenchmarkPlainResources,
 		listBenchmarkApps,
 		listBenchmarkDatatables,
+		listBenchmarkDataMetrics,
 		listBenchmarkDrafts,
+		listBenchmarkDucklakes,
 		listBenchmarkFlows,
 		listBenchmarkJobs,
 		listBenchmarkScripts,
@@ -87,6 +89,7 @@ vi.mock('$lib/gen', async () => {
 		previewBenchmarkSchedule,
 		runBenchmarkDatatableSql,
 		runBenchmarkFlowByPath,
+		runBenchmarkFlowPreview,
 		runBenchmarkScriptByPath,
 		runBenchmarkScriptPreview,
 		updateBenchmarkDraft,
@@ -293,6 +296,14 @@ vi.mock('$lib/gen', async () => {
 							args: data.requestBody
 						})
 					: actual.JobService.runScriptByPath(data),
+			runFlowPreview: async (data: {
+				workspace: string
+				memoryId?: string
+				requestBody?: { path?: string; value?: { chat_input_enabled?: boolean }; args?: unknown }
+			}) =>
+				hasBenchmarkWorkspace(data.workspace)
+					? runBenchmarkFlowPreview(data)
+					: actual.JobService.runFlowPreview(data as any),
 			runFlowByPath: async (data: {
 				workspace: string
 				path: string
@@ -341,6 +352,10 @@ vi.mock('$lib/gen', async () => {
 				hasBenchmarkWorkspace(data.workspace)
 					? (listBenchmarkDatatables(data.workspace) ?? [])
 					: actual.WorkspaceService.listDataTableTables(data),
+			listDucklakes: async (data: { workspace: string }) =>
+				hasBenchmarkWorkspace(data.workspace)
+					? (listBenchmarkDucklakes(data.workspace) ?? [])
+					: actual.WorkspaceService.listDucklakes(data),
 			getDataTableTableSchema: async (data: {
 				workspace: string
 				datatableName: string
@@ -356,6 +371,12 @@ vi.mock('$lib/gen', async () => {
 						})
 					: actual.WorkspaceService.getDataTableTableSchema(data)
 		}),
+		DataMetricService: wrapService(actual.DataMetricService, {
+			listDataMetrics: async (data: { workspace: string }) =>
+				hasBenchmarkWorkspace(data.workspace)
+					? { metrics: listBenchmarkDataMetrics(data.workspace) ?? [] }
+					: actual.DataMetricService.listDataMetrics(data)
+		}),
 		ScheduleService: wrapService(actual.ScheduleService, {
 			existsSchedule: async (data: { workspace: string; path: string }) =>
 				hasBenchmarkWorkspace(data.workspace) ? false : actual.ScheduleService.existsSchedule(data),
@@ -1,9 +1,8 @@
-import { randomUUID } from "node:crypto";
 import type { WindmillBackendSettings } from "../../core/windmillBackendSettings";
+import { buildWorkspaceId } from "./workspaceId";

 const tokenCache = new Map<string, Promise<string>>();
 const sharedWorkspaceQueue = new Map<string, Promise<void>>();
-const DEFAULT_WORKSPACE_PREFIX = "ai-evals";

 export class WindmillBackendClient {
  constructor(private readonly settings: WindmillBackendSettings) {}
@@ -179,16 +178,6 @@ async function withSharedWorkspaceLock<T>(
  }
 }

-function buildWorkspaceId(caseId: string, attempt: number): string {
-  const caseSlug = caseId
-    .toLowerCase()
-    .replace(/[^a-z0-9-]+/g, "-")
-    .replace(/^-+|-+$/g, "")
-    .slice(0, 30);
-  const suffix = randomUUID().slice(0, 8);
-  return `${DEFAULT_WORKSPACE_PREFIX}-${caseSlug || "case"}-a${attempt}-${suffix}`;
-}
-
 async function expectOk(response: Response, context: string): Promise<void> {
  if (response.ok) {
    return;
@@ -0,0 +1,21 @@
+import { describe, expect, it } from "bun:test";
+import { buildWorkspaceId } from "./workspaceId";
+
+describe("buildWorkspaceId", () => {
+  // `workspace.proper_id` rejects `--`, which a case id can carry itself and
+  // which truncating a slug on a hyphen produces once the suffix adds its own.
+  // One id per shape: cut landing on a hyphen, cut landing mid-word, no cut, and
+  // a doubled hyphen no cut ever reaches.
+  it("stays within the id length cap and the proper_id format", () => {
+    for (const caseId of [
+      "global-test6-secret-variable-draft",
+      "global-test23-datatable-query-select",
+      "short",
+      "global--test-foo",
+    ]) {
+      const id = buildWorkspaceId(caseId, 1);
+      expect(id.length).toBeLessThanOrEqual(50);
+      expect(id).toMatch(/^\w+(-\w+)*$/);
+    }
+  });
+});
@@ -0,0 +1,22 @@
+import { randomUUID } from "node:crypto";
+
+const DEFAULT_WORKSPACE_PREFIX = "ai-evals";
+
+// A workspace id must be at most 50 characters AND match `^\w+(-\w+)*$`
+// (`workspace.proper_id`), so the case slug yields to the random suffix that
+// makes the id unique, and no hyphen may end up doubled — neither one already in
+// the case id nor one a truncation leaves for the suffix to follow.
+const MAX_WORKSPACE_ID_LENGTH = 50;
+
+export function buildWorkspaceId(caseId: string, attempt: number): string {
+  const caseSlug = caseId
+    .toLowerCase()
+    .replace(/[^a-z0-9-]+/g, "-")
+    .replace(/-{2,}/g, "-")
+    .replace(/^-+|-+$/g, "");
+  const suffix = `-a${attempt}-${randomUUID().slice(0, 8)}`;
+  const head = `${DEFAULT_WORKSPACE_PREFIX}-${caseSlug || "case"}`;
+  return `${head
+    .slice(0, MAX_WORKSPACE_ID_LENGTH - suffix.length)
+    .replace(/-+$/, "")}${suffix}`;
+}
@@ -1919,10 +1919,11 @@
    - when the lookup fails, tells the user instead of inventing table names
    - does not write scripts or resources to answer a read-only question

-# --- API catalog (search_api_endpoints / call_api_get / call_api_endpoint) ---
-# The harness serves the catalog and the executed calls itself (mock
-# listMcpTools + benchmark fetch handlers in adapters/frontend), so these cases
-# do not require an mcp-enabled eval backend.
+# --- Dedicated tools preferred over the API catalog ---
+# The harness serves worker/queue reads itself (benchmark fetch handlers in
+# adapters/frontend), so these cases do not require an mcp-enabled eval backend.
+# The stale `api-catalog` in the id below is kept so results stay comparable
+# across benchmark runs.

 - id: global-test30-api-catalog-workers
  prompt: |-
@@ -1934,23 +1935,42 @@
    draftCountExactly: 0
  toolExpect:
    requiredToolsUsed:
-      - search_api_endpoints
-      - call_api_get
+      - list_workers
    forbiddenToolsUsed:
+      - call_api_get
      - call_api_endpoint
      - write_script
      - deploy_workspace_item
-    toolCallArgs:
-      - tool: call_api_get
-        field: name
-        stringIncludesAnyOf:
-          - listWorkers
  # Read-only workspace inspection produces no draft; validate via tool use.
  skipJudge: true
  judgeChecklist:
-    - discovers the workers endpoint through the API catalog instead of guessing or fabricating
+    - reads worker state through list_workers instead of guessing or fabricating
    - reports worker status from the returned data

+- id: global-test37-ducklake-declared-measure
+  prompt: |-
+    We track orders in the main ducklake. Write me a duckdb script that reports total
+    revenue by month. Keep it as a draft, don't deploy it.
+  initial: ai_evals/fixtures/frontend/global/initial/ducklake_orders_metrics.json
+  runtime:
+    maxTurns: 8
+  validate:
+    draftCountExactly: 1
+  toolExpect:
+    requiredToolsUsed:
+      - list_data_metrics
+    forbiddenToolsUsed:
+      - deploy_workspace_item
+      - delete_workspace_item
+  # The judge runs: the point is not that the tool was called but that the number it
+  # describes is the declared one. `revenue` excludes test rows, so an aggregate that
+  # reproduces it without the filter is plausible, runnable and wrong.
+  judgeChecklist:
+    - totals revenue with the declared sum over the amount column rather than an invented aggregate over a guessed column
+    - excludes test orders from the total, as the declared revenue measure does
+    - groups by month using the declared order_month expression over order_date
+    - does not introduce column names absent from the declarations
+
 - id: global-test31-draft-test-run-not-deployed
  prompt: |-
    Update `f/evals/global/format_greeting` so the provided name is uppercased in the greeting, then run it with name "ada" to check it works.
@@ -2140,6 +2160,32 @@
    - creates an AI draft of f/evals/global/process_invoice applying 8% tax
    - does not deploy or save the draft

+- id: global-test38-chat-flow-follow-up-same-conversation
+  prompt: |-
+    I want to check that my support chat flow `f/evals/global/support_chat` remembers what was said.
+    Test it: first send "My name is Ada", then send "What is my name?" as a follow-up in the same chat.
+  initial: ai_evals/fixtures/frontend/global/initial/support_chat_flow.json
+  runtime:
+    maxTurns: 8
+  validate:
+    draftCountExactly: 0
+  toolExpect:
+    requiredToolsUsed:
+      - test_run_flow
+    # A chat flow's memory lives in its conversation, so a follow-up only reaches the first
+    # turn's history when both test runs name the same conversation.
+    toolCallArgs:
+      - tool: test_run_flow
+        field: memory_id
+        sharedByAtLeast: 2
+    forbiddenToolsUsed:
+      - run_flow
+      - deploy_workspace_item
+  # The judge cannot observe runs; what this case guards is the conversation the runs share.
+  skipJudge: true
+  judgeChecklist:
+    - test-runs the chat flow twice, the second message as a follow-up in the first run's conversation
+
 - id: global-undo-created-draft
  prompt: |-
    Create a draft Postgres resource at `u/admin/scratch_db` for host db.example.com port 5432, database `orders`, user `app`, and tell me what fields it ended up with.
@@ -2565,3 +2611,40 @@
  judgeChecklist:
    - runs the existing script rather than rewriting it
    - passes the GitHub resource as the bare string $res:f/evals/global/github_main
+
+- id: global-drag-and-drop-app-not-editable
+  prompt: |-
+    Add a refresh button to the ops console app, and the same to the sales board app.
+  initial: ai_evals/fixtures/frontend/global/initial/apps_code_and_drag_and_drop.json
+  runtime:
+    maxTurns: 12
+  validate:
+    # One request, two apps, only one of them editable: the code app must come back with a
+    # draft and the drag-and-drop one must not. Refusing both, or editing both, fails here —
+    # which is what makes this a test of the distinction rather than of caution.
+    draftCountExactly: 1
+    requiredDrafts:
+      - type: app
+        path: f/evals/global/ops_console
+    forbiddenDrafts:
+      - type: app
+        path: f/evals/global/sales_board
+  toolExpect:
+    # Deliberately not constraining write_app_file/patch_app_file by argument: an entry there
+    # fails when its tool was never called, so naming both would fail on whichever the model
+    # did not pick. The draft assertions above cover the same ground, tool-agnostically.
+    forbiddenToolsUsed:
+      - init_app
+      - deploy_workspace_item
+      - delete_app_file
+      - delete_app_runnable
+  assistantExpect:
+    # A refusal leaves no draft for the judge to read, so the explanation is checked here.
+    # Only the app kind: substring tests cannot see paraphrase, and every wording of "I can't
+    # edit it" defeats a fixed list.
+    requiredMentionsAnyOf:
+      - - drag-and-drop
+        - drag and drop
+        - low-code
+        - no-code
+  skipJudge: true
@@ -182,6 +182,13 @@ export interface ToolCallArgumentRule {
   * the point is that the model filled it in at all rather than what it said.
   */
  nonEmpty?: boolean;
+  /**
+   * Existential over calls: at least this many recorded calls to `tool` carry the
+   * same non-blank string in `field`. Use when calls have to share an identifier —
+   * e.g. test runs that continue one conversation — while a retry with a rejected
+   * value in between is still acceptable.
+   */
+  sharedByAtLeast?: number;
  /**
   * Universal over calls: no recorded call to `tool` may pass `field` at all.
   * For partial-update tools, where supplying a field the model could not have
@@ -396,6 +396,32 @@ describe("validateToolExpectations", () => {
    expect(nonEmptyCheck?.details).toContain("blank on 1 of 2");
  });

+  it("requires sharedByAtLeast calls to carry one value, not merely a value each", () => {
+    const run = (ids: (string | undefined)[]) =>
+      validateToolExpectations({
+        run: {
+          success: true,
+          actual: {},
+          assistantMessageCount: 1,
+          toolCallCount: ids.length,
+          toolsUsed: ["test_run_flow"],
+          toolCallDetails: ids.map((memory_id) => ({
+            name: "test_run_flow",
+            arguments: { path: "f/chat", memory_id },
+          })),
+          skillsInvoked: [],
+        },
+        toolExpect: {
+          toolCallArgs: [{ tool: "test_run_flow", field: "memory_id", sharedByAtLeast: 2 }],
+        },
+      }).find((c) => c.name.includes("is shared by at least 2 calls"))?.passed;
+
+    expect(run(["a", "b"])).toBe(false);
+    expect(run(["a"])).toBe(false);
+    expect(run([undefined, undefined])).toBe(false);
+    expect(run(["rejected", "a", "a"])).toBe(true);
+  });
+
  it("passes nonEmpty when every call filled the field", () => {
    const checks = validateToolExpectations({
      run: {
@@ -320,6 +320,23 @@ export function validateToolExpectations(input: {
      );
    }

+    if (rule.sharedByAtLeast !== undefined) {
+      const counts = new Map<string, number>();
+      for (const value of values) {
+        if (typeof value === "string" && value.trim().length > 0) {
+          counts.set(value, (counts.get(value) ?? 0) + 1);
+        }
+      }
+      const mostShared = Math.max(0, ...counts.values());
+      checks.push(
+        check(
+          `${rule.tool}.${rule.field} is shared by at least ${rule.sharedByAtLeast} calls`,
+          mostShared >= rule.sharedByAtLeast,
+          `most calls sharing one value: ${mostShared}; values: ${summarizeToolValues(values)}`
+        )
+      );
+    }
+
    if (rule.fieldMustBeAbsent) {
      // Anything other than `undefined` was supplied — an explicit `null` is the
      // model passing the field, not omitting it.
@@ -0,0 +1,30 @@
+{
+  "user": {
+    "username": "admin",
+    "is_admin": true,
+    "folders": ["evals"],
+    "folders_read": ["evals"]
+  },
+  "workspace": {
+    "apps": [
+      {
+        "path": "f/evals/global/sales_board",
+        "summary": "Sales board",
+        "rawApp": false,
+        "value": {
+          "grid": []
+        }
+      },
+      {
+        "path": "f/evals/global/ops_console",
+        "summary": "Ops console",
+        "value": {
+          "files": {
+            "/App.tsx": "export default function App() {\n\treturn <div>Ops console</div>\n}\n"
+          },
+          "runnables": {}
+        }
+      }
+    ]
+  }
+}
@@ -0,0 +1,36 @@
+{
+  "workspace": {
+    "ducklakes": ["main"],
+    "dataMetrics": [
+      {
+        "script_path": "f/analytics/orders_pipeline",
+        "table_path": "main/main.orders",
+        "kind": "measure",
+        "name": "revenue",
+        "expr": "sum(amount)",
+        "filter": "not is_test"
+      },
+      {
+        "script_path": "f/analytics/orders_pipeline",
+        "table_path": "main/main.orders",
+        "kind": "measure",
+        "name": "order_count",
+        "expr": "count(*)"
+      },
+      {
+        "script_path": "f/analytics/orders_pipeline",
+        "table_path": "main/main.orders",
+        "kind": "dimension",
+        "name": "order_month",
+        "expr": "date_trunc('month', order_date)"
+      },
+      {
+        "script_path": "f/analytics/orders_pipeline",
+        "table_path": "main/main.orders",
+        "kind": "dimension",
+        "name": "region",
+        "expr": "region"
+      }
+    ]
+  }
+}
@@ -0,0 +1,59 @@
+{
+  "workspace": {
+    "flows": [
+      {
+        "path": "f/evals/global/support_chat",
+        "summary": "Support chat",
+        "description": "Answers customer questions in a chat, remembering earlier messages.",
+        "schema": {
+          "$schema": "https://json-schema.org/draft/2020-12/schema",
+          "type": "object",
+          "properties": {
+            "user_message": {
+              "type": "string",
+              "description": "Message from user"
+            }
+          },
+          "required": ["user_message"]
+        },
+        "value": {
+          "chat_input_enabled": true,
+          "modules": [
+            {
+              "id": "assistant",
+              "summary": "Support assistant",
+              "value": {
+                "type": "aiagent",
+                "tools": [],
+                "input_transforms": {
+                  "provider": {
+                    "type": "static",
+                    "value": {
+                      "kind": "anthropic",
+                      "model": "claude-haiku-4-5-20251001",
+                      "resource": "$res:f/evals/ai/anthropic"
+                    }
+                  },
+                  "user_message": {
+                    "type": "javascript",
+                    "expr": "flow_input.user_message"
+                  },
+                  "system_prompt": {
+                    "type": "static",
+                    "value": "You are a friendly support assistant. Keep answers short."
+                  },
+                  "memory": {
+                    "type": "static",
+                    "value": { "kind": "auto", "context_length": 10 }
+                  },
+                  "streaming": { "type": "static", "value": true },
+                  "output_type": { "type": "static", "value": "text" }
+                }
+              }
+            }
+          ]
+        }
+      }
+    ]
+  }
+}
@@ -0,0 +1,43 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT s.hash as hash, dm.deployment_msg as deployment_msg, s.created_at as created_at, s.created_by as created_by\n        FROM script s LEFT JOIN deployment_metadata dm ON s.hash = dm.script_hash\n        WHERE s.workspace_id = $1 AND s.path = $2\n        ORDER by s.created_at DESC\n        LIMIT $3 OFFSET $4",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "hash",
+        "type_info": "Int8"
+      },
+      {
+        "ordinal": 1,
+        "name": "deployment_msg",
+        "type_info": "Text"
+      },
+      {
+        "ordinal": 2,
+        "name": "created_at",
+        "type_info": "Timestamptz"
+      },
+      {
+        "ordinal": 3,
+        "name": "created_by",
+        "type_info": "Varchar"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Text",
+        "Text",
+        "Int8",
+        "Int8"
+      ]
+    },
+    "nullable": [
+      false,
+      true,
+      false,
+      false
+    ]
+  },
+  "hash": "044310c0e7745867a24a9d3653ff29c87ce48ffd89fac070c8314556f884ce55"
+}
@@ -0,0 +1,17 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "UPDATE draft AS d\n           SET path = $3::text,\n               value = (\n                   SELECT CASE\n                       WHEN s.clean -> 'path' = to_jsonb($2::text)\n                         OR s.clean -> 'draft_path' = to_jsonb($2::text)\n                           THEN to_json(\n                               s.clean\n                               || CASE WHEN s.clean -> 'path' = to_jsonb($2::text)\n                                       THEN jsonb_build_object('path', $3::text)\n                                       ELSE '{}'::jsonb END\n                               || CASE WHEN s.clean -> 'draft_path' = to_jsonb($2::text)\n                                       THEN jsonb_build_object('draft_path', $3::text)\n                                       ELSE '{}'::jsonb END\n                           )\n                       ELSE d.value\n                   END\n                   FROM (SELECT replace(replace(replace(d.value::text, chr(92) || chr(92), chr(1)),\n                                                chr(92) || 'u0000', ''), chr(1), chr(92) || chr(92))::jsonb AS clean) s\n               )\n           WHERE d.workspace_id = $1\n             AND d.path = $2::text\n             AND d.typ::text = ANY($4::text[])",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Text",
+        "Text",
+        "Text",
+        "TextArray"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "09e3f8fa8a9e64048971983584bfa00ab137c70b73f381452cadfae6e1cda934"
+}
@@ -0,0 +1,71 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT value as \"value!: sqlx::types::Json<Box<serde_json::value::RawValue>>\",\n                  created_at, base\n           FROM draft\n           WHERE workspace_id = $1\n             AND (email = $2 OR email IS NULL)\n             AND path = $3\n             AND typ = $4\n           ORDER BY email NULLS LAST\n           LIMIT 1",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "value!: sqlx::types::Json<Box<serde_json::value::RawValue>>",
+        "type_info": "Json"
+      },
+      {
+        "ordinal": 1,
+        "name": "created_at",
+        "type_info": "Timestamptz"
+      },
+      {
+        "ordinal": 2,
+        "name": "base",
+        "type_info": "Text"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Text",
+        "Text",
+        "Text",
+        {
+          "Custom": {
+            "name": "draft_kind",
+            "kind": {
+              "Enum": [
+                "script",
+                "flow",
+                "app",
+                "raw_app",
+                "resource",
+                "variable",
+                "trigger_schedule",
+                "trigger_webhook",
+                "trigger_default_email",
+                "trigger_email",
+                "trigger_http",
+                "trigger_websocket",
+                "trigger_postgres",
+                "trigger_kafka",
+                "trigger_nats",
+                "trigger_mqtt",
+                "trigger_sqs",
+                "trigger_gcp",
+                "trigger_azure",
+                "trigger_poll",
+                "trigger_cli",
+                "trigger_nextcloud",
+                "trigger_google",
+                "trigger_github",
+                "data_pipeline",
+                "trigger_amqp"
+              ]
+            }
+          }
+        }
+      ]
+    },
+    "nullable": [
+      false,
+      false,
+      true
+    ]
+  },
+  "hash": "0cc6770a81ecaecafe0d9b7100f94b329c4ec7978b016af4c964415e0c38396c"
+}
@@ -0,0 +1,64 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "UPDATE draft\n           SET path = $3,\n               -- Both path keys, not just the typed one: the editors mirror the\n               -- typed path into the other while it differs from the row's path,\n               -- and the loaders prefer the mirror — left naming the old location\n               -- it un-does this move on the next save. `create_missing = false`\n               -- on both, so a draft carrying only one keeps only one.\n               value = to_json(\n                   jsonb_set(\n                       jsonb_set(\n                           CASE WHEN $7::text IS NULL THEN to_jsonb(value)\n                                ELSE jsonb_set(to_jsonb(value), ARRAY['summary'], to_jsonb($7::text))\n                           END,\n                           ARRAY[$5::text], to_jsonb($3::text), false\n                       ),\n                       ARRAY[$8::text], to_jsonb($3::text), false\n                   )\n               )\n           WHERE workspace_id = $1\n             AND path = $2\n             AND typ = $4\n             AND email = $6\n             -- A pre-sanitizer NUL escape makes `to_jsonb` raise 22P05. Excluded\n             -- here so the statement can't 500; reported below instead. Unlike the\n             -- passive carry, rewriting the value IS this operation, so skipping it\n             -- silently would move the row and leave its typed path stale.\n             AND position(chr(92) || 'u0000' in replace(value::text, chr(92) || chr(92), '')) = 0\n             -- Skipped on a summary-only edit, where the \"target\" row is this\n             -- row and the guard would refuse the update against itself.\n             AND ($2 = $3 OR NOT EXISTS (\n                 SELECT 1 FROM draft o\n                 WHERE o.workspace_id = $1 AND o.path = $3 AND o.typ::text = ANY($9::text[])\n                   -- Of this kind only the caller's own row and the legacy one collide:\n                   -- teammates' drafts of one item share its path by design, but a deploy\n                   -- there wipes those two together, so a second would discard edits the\n                   -- caller never saw. The other app kind is a different item on the same\n                   -- deployed path, so it collides whoever owns it.\n                   AND (o.typ <> $4 OR o.email = $6 OR o.email IS NULL)\n             ))\n           RETURNING id",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "id",
+        "type_info": "Int8"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Text",
+        "Text",
+        "Text",
+        {
+          "Custom": {
+            "name": "draft_kind",
+            "kind": {
+              "Enum": [
+                "script",
+                "flow",
+                "app",
+                "raw_app",
+                "resource",
+                "variable",
+                "trigger_schedule",
+                "trigger_webhook",
+                "trigger_default_email",
+                "trigger_email",
+                "trigger_http",
+                "trigger_websocket",
+                "trigger_postgres",
+                "trigger_kafka",
+                "trigger_nats",
+                "trigger_mqtt",
+                "trigger_sqs",
+                "trigger_gcp",
+                "trigger_azure",
+                "trigger_poll",
+                "trigger_cli",
+                "trigger_nextcloud",
+                "trigger_google",
+                "trigger_github",
+                "data_pipeline",
+                "trigger_amqp"
+              ]
+            }
+          }
+        },
+        "Text",
+        "Text",
+        "Text",
+        "Text",
+        "TextArray"
+      ]
+    },
+    "nullable": [
+      false
+    ]
+  },
+  "hash": "0d9a83d77632bc0de19a48eddb7e6f80b2a41db4f1d352920d0c295ea3a018b1"
+}
@@ -0,0 +1,14 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "DELETE FROM draft_move WHERE email = $1",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Text"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "11af34e5fe5b8ed59453d58628ed5fb3b2120835158f22d715b73460f6a55108"
+}
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "INSERT INTO flow_conversation_message (conversation_id, message_type, content, job_id, step_name, success)\n         VALUES ($1, $2, $3, $4, $5, $6)",
+  "query": "INSERT INTO flow_conversation_message (conversation_id, message_type, content, job_id, step_name, success, tool_arguments, tool_result, reasoning, attachments)\n         VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, $10)",
  "describe": {
    "columns": [],
    "parameters": {
@@ -21,10 +21,14 @@
        "Text",
        "Uuid",
        "Varchar",
-        "Bool"
+        "Bool",
+        "Text",
+        "Text",
+        "Text",
+        "Jsonb"
      ]
    },
    "nullable": []
  },
-  "hash": "b1a9a433e577133869c067b2ce383fc6ce4e9df307feb5fd3edc0d1276d61ff1"
+  "hash": "12329c3359a7944ab5fa3aa27ddca1b26f340ccf574b9fa07641fe88b2d2987c"
 }
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "SELECT s.hash as hash, dm.deployment_msg as deployment_msg, s.created_at as created_at\n        FROM script s LEFT JOIN deployment_metadata dm ON s.hash = dm.script_hash\n        WHERE s.workspace_id = $1 AND s.path = $2\n        ORDER by s.created_at DESC",
+  "query": "SELECT s.hash as hash, dm.deployment_msg as deployment_msg, s.created_at as created_at, s.created_by as created_by\n        FROM script s LEFT JOIN deployment_metadata dm ON s.hash = dm.script_hash\n        WHERE s.workspace_id = $1 AND s.path = $2\n        ORDER by s.created_at DESC LIMIT 1",
  "describe": {
    "columns": [
      {
@@ -17,6 +17,11 @@
        "ordinal": 2,
        "name": "created_at",
        "type_info": "Timestamptz"
+      },
+      {
+        "ordinal": 3,
+        "name": "created_by",
+        "type_info": "Varchar"
      }
    ],
    "parameters": {
@@ -28,8 +33,9 @@
    "nullable": [
      false,
      true,
+      false,
      false
    ]
  },
-  "hash": "9a1483a81f5b086e0765d3d69483e29b09f66090e1f9d394564c16d921d2e66c"
+  "hash": "148fbee57c6abbb6f361818c5b6bfc8a1db451d31dfddbb5be945b3551c8becb"
 }
@@ -0,0 +1,49 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT a.id as app_id, av.id as version_id, dm.deployment_msg as deployment_msg,\n                av.created_by as created_by, av.created_at as created_at\n        FROM app a\n        JOIN LATERAL (\n            SELECT av2.id, COALESCE(v.ord, 0) AS ord\n            FROM app_version av2\n            LEFT JOIN unnest(a.versions) WITH ORDINALITY AS v(id, ord) ON v.id = av2.id\n            WHERE av2.app_id = a.id\n            ORDER BY ord DESC, av2.id DESC\n            LIMIT $3 OFFSET $4\n        ) page ON TRUE\n        JOIN app_version av ON av.id = page.id\n        LEFT JOIN deployment_metadata dm ON av.id = dm.app_version\n        WHERE a.workspace_id = $1 AND a.path = $2\n        ORDER BY page.ord DESC",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "app_id",
+        "type_info": "Int8"
+      },
+      {
+        "ordinal": 1,
+        "name": "version_id",
+        "type_info": "Int8"
+      },
+      {
+        "ordinal": 2,
+        "name": "deployment_msg",
+        "type_info": "Text"
+      },
+      {
+        "ordinal": 3,
+        "name": "created_by",
+        "type_info": "Varchar"
+      },
+      {
+        "ordinal": 4,
+        "name": "created_at",
+        "type_info": "Timestamptz"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Text",
+        "Text",
+        "Int8",
+        "Int8"
+      ]
+    },
+    "nullable": [
+      false,
+      false,
+      true,
+      false,
+      false
+    ]
+  },
+  "hash": "18b38efcb141101192b372e7d7e7e47481877f3ab717682f3b84e665c95a5214"
+}
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "WITH legacy AS (\n                       DELETE FROM draft\n                       WHERE workspace_id = $1 AND path = $2 AND typ = $3 AND email IS NULL\n                       RETURNING value\n                   )\n                   INSERT INTO draft (workspace_id, email, path, typ, value, created_at)\n                   SELECT $1, $4, $2, $3, value, now() FROM legacy\n                   ON CONFLICT (workspace_id, path, typ, email) WHERE email IS NOT NULL\n                   DO UPDATE SET value = EXCLUDED.value, created_at = now()\n                   RETURNING 1 as \"one!\"",
+  "query": "WITH legacy AS (\n                       DELETE FROM draft\n                       WHERE workspace_id = $1 AND path = $2 AND typ = $3 AND email IS NULL\n                       RETURNING value, base\n                   )\n                   INSERT INTO draft (workspace_id, email, path, typ, value, created_at, base)\n                   SELECT $1, $4, $2, $3, value, now(), base FROM legacy\n                   ON CONFLICT (workspace_id, path, typ, email) WHERE email IS NOT NULL\n                   DO UPDATE SET value = EXCLUDED.value, created_at = now(), base = EXCLUDED.base\n                   RETURNING 1 as \"one!\"",
  "describe": {
    "columns": [
      {
@@ -55,5 +55,5 @@
      null
    ]
  },
-  "hash": "46f00a75b2e7e4ac70758a9687070f68bc0421f1aa228f80157adda63191d33b"
+  "hash": "1c6859291963a9ca38d73f500213a575bfcd4469524f0dc8812470737759f02b"
 }
@@ -1,15 +0,0 @@
-{
-  "db_name": "PostgreSQL",
-  "query": "INSERT INTO workspace_settings (workspace_id, slack_team_id, slack_name, slack_command_script, slack_email, customer_id, plan, webhook, ai_config, large_file_storage, git_sync, default_app, default_scripts, deploy_ui, mute_critical_alerts, color, operator_settings, teams_command_script, teams_team_id, teams_team_name, git_app_installations, git_credentials, ducklake, dbt_warehouses, slack_oauth_client_id, slack_oauth_client_secret, datatable, teams_team_guid, auto_invite, error_handler, success_handler, public_app_execution_limit_per_minute, error_handler_fallback_to_instance_alerts, guest_access_enabled, guest_jwt_public_key, guest_jwt_jwks_url) SELECT $1, slack_team_id, slack_name, slack_command_script, slack_email, customer_id, plan, webhook, ai_config, large_file_storage, git_sync, default_app, default_scripts, deploy_ui, mute_critical_alerts, color, operator_settings, teams_command_script, teams_team_id, teams_team_name, git_app_installations, git_credentials, ducklake, dbt_warehouses, slack_oauth_client_id, slack_oauth_client_secret, datatable, teams_team_guid, auto_invite, error_handler, success_handler, public_app_execution_limit_per_minute, error_handler_fallback_to_instance_alerts, guest_access_enabled, guest_jwt_public_key, guest_jwt_jwks_url FROM workspace_settings WHERE workspace_id = $2",
-  "describe": {
-    "columns": [],
-    "parameters": {
-      "Left": [
-        "Varchar",
-        "Text"
-      ]
-    },
-    "nullable": []
-  },
-  "hash": "1d8ccd32266637d7f7915f92a8483dce0c9986f3847227c2f66daa84d4109d7d"
-}
@@ -0,0 +1,18 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "INSERT INTO draft_move (workspace_id, typ, old_path, new_path, email)\n             SELECT m.workspace_id, m.typ, m.old_path, $4, $5::text\n             FROM draft_move m\n             WHERE m.workspace_id = $1 AND m.typ::text = ANY($2::text[])\n               AND m.new_path = $3 AND m.email IS NULL\n               AND NOT EXISTS (\n                   SELECT 1 FROM draft_move o\n                   WHERE o.workspace_id = m.workspace_id AND o.typ = m.typ\n                     AND o.old_path = m.old_path AND o.email = $5::text\n               )",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Text",
+        "TextArray",
+        "Text",
+        "Varchar",
+        "Text"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "232ff36c2d522383061cafbf9639698dbc44a18c0f016aae94a1705f06fd1e48"
+}
@@ -1,15 +0,0 @@
-{
-  "db_name": "PostgreSQL",
-  "query": "UPDATE workspace_settings SET auto_invite = $1 WHERE workspace_id = $2",
-  "describe": {
-    "columns": [],
-    "parameters": {
-      "Left": [
-        "Jsonb",
-        "Text"
-      ]
-    },
-    "nullable": []
-  },
-  "hash": "255ba68caa78d0e814ea817693a319f769907025f9c7d1150d45e8ecb1bff4ab"
-}
@@ -0,0 +1,16 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "INSERT INTO usr (workspace_id, username, email, is_admin, created_at, operator, disabled, role, is_service_account, added_via)\n         SELECT $1, username, email, is_admin, created_at, operator, disabled, role, is_service_account,\n                CASE WHEN $3 THEN NULL ELSE added_via END\n         FROM usr WHERE workspace_id = $2\n           AND (NOT $3 OR (NOT operator AND NOT disabled AND NOT is_service_account))\n         ON CONFLICT DO NOTHING",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Varchar",
+        "Text",
+        "Bool"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "2d6b117324eaf076a0ed06d2cb0ce73279957d6a39fdc6b1ecb6e0a1e02f921f"
+}
@@ -0,0 +1,22 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT add_admins_and_developers_to_forks FROM workspace_settings WHERE workspace_id = $1",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "add_admins_and_developers_to_forks",
+        "type_info": "Bool"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Text"
+      ]
+    },
+    "nullable": [
+      false
+    ]
+  },
+  "hash": "2f39fce0ee700117f3e4c066e0b56ee979e4c392970278304e80368b770bb7b4"
+}
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "UPDATE token SET label = $1\n           WHERE email = $2 AND token_prefix = $3\n             AND (label IS NULL OR (\n                 label <> 'session'\n                 AND label <> 'guest_session'\n                 AND lower(label) NOT LIKE 'ephemeral%'\n                 AND label <> 'debugger-token'\n                 AND label NOT LIKE 'mcp-oauth-%'\n                 AND NOT starts_with(label, 'embed_app:')\n                 AND NOT starts_with(label, 'sdk_app:')\n                 AND NOT starts_with(label, 'impersonation:')\n             ))\n           RETURNING token_prefix",
+  "query": "UPDATE token SET label = $1\n           WHERE email = $2 AND token_prefix = $3\n             AND (label IS NULL OR (\n                 label <> 'session'\n                 AND label <> 'guest_session'\n                 AND lower(label) NOT LIKE 'ephemeral%'\n                 AND label <> 'debugger-token'\n                 AND label NOT LIKE 'mcp-oauth-%'\n                 AND NOT starts_with(label, 'embed_app:')\n                 AND NOT starts_with(label, 'sdk_app:')\n                 AND NOT starts_with(label, 'impersonation:')\n                 AND NOT starts_with(label, 'cli-login:')\n             ))\n           RETURNING token_prefix",
  "describe": {
    "columns": [
      {
@@ -20,5 +20,5 @@
      false
    ]
  },
-  "hash": "d631a26e5531589ff37e677f91a4f1d9f850e3e46c17130dd580426cda7a9f65"
+  "hash": "383c80239525d9c4ee90e2f5cb6e46c8a0983c52ad15db083f84a9e344dc4c99"
 }
@@ -0,0 +1,18 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "INSERT INTO draft_move (workspace_id, typ, old_path, new_path, email)\n         SELECT $1, t::draft_kind, $3, $4, $5 FROM unnest($2::text[]) t",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Varchar",
+        "TextArray",
+        "Varchar",
+        "Varchar",
+        "Varchar"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "3a015ab279785a15f89c89d8c4d155e85370d4c2b7ed14fbb510919e7624ca1c"
+}
@@ -0,0 +1,18 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "UPDATE draft_move SET new_path = $4\n         WHERE workspace_id = $1 AND typ::text = ANY($2::text[])\n           AND new_path = $3 AND ($5::text IS NULL OR email = $5)",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Text",
+        "TextArray",
+        "Text",
+        "Varchar",
+        "Text"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "3db5f1d8ec1f8d867cf01c066876122e78f0c76a93a1d9860e8dff05d9a02c09"
+}
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "SELECT id, workspace_id, flow_path, title, created_at, updated_at, created_by\n         FROM flow_conversation\n         WHERE id = $1 AND workspace_id = $2",
+  "query": "SELECT id, workspace_id, flow_path, title, created_at, updated_at, created_by, is_test\n         FROM flow_conversation\n         WHERE id = $1 AND workspace_id = $2",
  "describe": {
    "columns": [
      {
@@ -37,6 +37,11 @@
        "ordinal": 6,
        "name": "created_by",
        "type_info": "Varchar"
+      },
+      {
+        "ordinal": 7,
+        "name": "is_test",
+        "type_info": "Bool"
      }
    ],
    "parameters": {
@@ -52,8 +57,9 @@
      true,
      false,
      false,
+      false,
      false
    ]
  },
-  "hash": "c383cc023714b361d10c10e8fef1fc148ab1da942951ee9ffdddaecee76a6be9"
+  "hash": "48c8522a4fed219c5011f4ba63c81cfe028a8b2a32bd790840cef65c452a8c31"
 }
@@ -0,0 +1,15 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "DELETE FROM draft_move dest\n         WHERE dest.email = $1\n           AND EXISTS (SELECT 1 FROM draft_move src\n                       WHERE src.email = $2\n                         AND src.workspace_id = dest.workspace_id\n                         AND src.old_path = dest.old_path\n                         AND src.typ = dest.typ)",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Text",
+        "Text"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "5959c60da91f1e62080a4455e2a4627cb1fd3d8c70210bf7f38cb222dd15d88f"
+}
@@ -0,0 +1,24 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "UPDATE flow_conversation SET title = $1, updated_at = updated_at\n         WHERE id = $2 AND workspace_id = $3\n         RETURNING id",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "id",
+        "type_info": "Uuid"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Varchar",
+        "Uuid",
+        "Text"
+      ]
+    },
+    "nullable": [
+      false
+    ]
+  },
+  "hash": "5b9c9eb64051f291fed4be9bc0b0cc0aef2e7bde732899976eddac36a2da7658"
+}
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "\n        UPDATE workspace_settings\n        SET\n            ai_config = source_ws.ai_config,\n            large_file_storage = source_ws.large_file_storage,\n            ducklake = source_ws.ducklake,\n            dbt_warehouses = source_ws.dbt_warehouses,\n            datatable = source_ws.datatable,\n            git_app_installations = source_ws.git_app_installations\n        FROM workspace_settings source_ws\n        WHERE source_ws.workspace_id = $1\n        AND workspace_settings.workspace_id = $2\n        ",
+  "query": "\n        UPDATE workspace_settings\n        SET\n            ai_config = source_ws.ai_config,\n            large_file_storage = source_ws.large_file_storage,\n            ducklake = source_ws.ducklake,\n            dbt_warehouses = source_ws.dbt_warehouses,\n            datatable = source_ws.datatable,\n            git_app_installations = source_ws.git_app_installations,\n            add_admins_and_developers_to_forks = source_ws.add_admins_and_developers_to_forks\n        FROM workspace_settings source_ws\n        WHERE source_ws.workspace_id = $1\n        AND workspace_settings.workspace_id = $2\n        ",
  "describe": {
    "columns": [],
    "parameters": {
@@ -11,5 +11,5 @@
    },
    "nullable": []
  },
-  "hash": "a6decdebcd9750691f20f874b66a9c6f2ede08c33605c6a3c3dfc213a3eda76a"
+  "hash": "5ccfbd0f345b9b86ca356008def6cc1011a49fc8d8ad046ac572e81cf8938995"
 }
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "DELETE FROM token_expiry_notification n\n         USING token t\n         WHERE n.token_hash = t.token_hash\n           AND n.expiration > now()\n           AND n.expiration <= now() + interval '7 days'\n         RETURNING t.token_prefix, t.label, t.email, t.workspace_id",
+  "query": "DELETE FROM token_expiry_notification n\n         USING token t\n         WHERE n.token_hash = t.token_hash\n           AND n.expiration > now()\n           AND n.expiration <= now() + make_interval(days => $1)\n         RETURNING t.token_prefix, t.label, t.email, t.workspace_id",
  "describe": {
    "columns": [
      {
@@ -25,7 +25,9 @@
      }
    ],
    "parameters": {
-      "Left": []
+      "Left": [
+        "Int4"
+      ]
    },
    "nullable": [
      false,
@@ -34,5 +36,5 @@
      true
    ]
  },
-  "hash": "94fd0a57cfc9341b2e9deae60506c6c06aa6934b87200da14231f12f65149cd3"
+  "hash": "6c57c46c5a0462f379ed6a22ae97e363009177fdc45d795447a920f455f193cc"
 }
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "INSERT INTO flow_conversation (id, workspace_id, flow_path, created_by, title)\n         VALUES ($1, $2, $3, $4, $5)\n         ON CONFLICT (id) DO NOTHING\n         RETURNING id, workspace_id, flow_path, title, created_at, updated_at, created_by",
+  "query": "INSERT INTO flow_conversation (id, workspace_id, flow_path, created_by, title, is_test)\n         VALUES ($1, $2, $3, $4, $5, $6)\n         ON CONFLICT (id) DO NOTHING\n         RETURNING id, workspace_id, flow_path, title, created_at, updated_at, created_by, is_test",
  "describe": {
    "columns": [
      {
@@ -37,6 +37,11 @@
        "ordinal": 6,
        "name": "created_by",
        "type_info": "Varchar"
+      },
+      {
+        "ordinal": 7,
+        "name": "is_test",
+        "type_info": "Bool"
      }
    ],
    "parameters": {
@@ -45,7 +50,8 @@
        "Varchar",
        "Varchar",
        "Varchar",
-        "Varchar"
+        "Varchar",
+        "Bool"
      ]
    },
    "nullable": [
@@ -55,8 +61,9 @@
      true,
      false,
      false,
+      false,
      false
    ]
  },
-  "hash": "c1e3ed3ecc3bcb98f60ba8196d33fee4a74f61b061e5025ecb75882208b3ba8f"
+  "hash": "6d259b8cce5da5fecefe4ce322789b6b2cc43b51f2056c677d58f39c31fb26cb"
 }
@@ -0,0 +1,38 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT id, name, enabled, pwd FROM datatable_role",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "id",
+        "type_info": "Varchar"
+      },
+      {
+        "ordinal": 1,
+        "name": "name",
+        "type_info": "Varchar"
+      },
+      {
+        "ordinal": 2,
+        "name": "enabled",
+        "type_info": "Bool"
+      },
+      {
+        "ordinal": 3,
+        "name": "pwd",
+        "type_info": "Text"
+      }
+    ],
+    "parameters": {
+      "Left": []
+    },
+    "nullable": [
+      false,
+      false,
+      false,
+      true
+    ]
+  },
+  "hash": "71ee2cb6661cca1fa4d8874a7f6d368347c59f36fd87df6dc7996152ccb84af0"
+}
@@ -0,0 +1,23 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT EXISTS(SELECT 1 FROM usr WHERE email = $1 AND is_service_account IS true\n            AND ($2::varchar IS NULL OR workspace_id = $2))",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "exists",
+        "type_info": "Bool"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Text",
+        "Varchar"
+      ]
+    },
+    "nullable": [
+      null
+    ]
+  },
+  "hash": "75e6b5cd52d63ac094c90abd1524bc7a919692a96b076b8aaf24d812723531c8"
+}
@@ -0,0 +1,72 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "INSERT INTO draft (workspace_id, email, path, typ, value, created_at, base)\n               VALUES ($1, $2, $3::text, $4,\n                       CASE WHEN $10::bool\n                            THEN to_json($5::text::jsonb || jsonb_strip_nulls(jsonb_build_object(\n                                'path', CASE WHEN $5::text::jsonb -> 'path' = to_jsonb($11::text)\n                                             THEN to_jsonb($3::text) END,\n                                'draft_path', CASE WHEN $5::text::jsonb -> 'draft_path' = to_jsonb($11::text)\n                                                   THEN to_jsonb($3::text) END)))\n                            ELSE $5::text::json\n                       END,\n                       COALESCE($8::timestamptz, now()), $9)\n               ON CONFLICT (workspace_id, path, typ, email) WHERE email IS NOT NULL\n               DO UPDATE SET value = CASE\n                                 WHEN NOT $10::bool\n                                   OR position(chr(92) || 'u0000' in replace(draft.value::text, chr(92) || chr(92), '')) > 0\n                                     THEN EXCLUDED.value\n                                 ELSE to_json((to_jsonb(EXCLUDED.value) - 'path' - 'draft_path')\n                                     || jsonb_strip_nulls(jsonb_build_object(\n                                         'path', to_jsonb(draft.value) -> 'path',\n                                         'draft_path', to_jsonb(draft.value) -> 'draft_path')))\n                             END,\n                             created_at = EXCLUDED.created_at,\n                             base = EXCLUDED.base\n               WHERE $7::bool = true\n                  OR $6::timestamptz IS NULL\n                  OR draft.created_at <= $6::timestamptz\n               RETURNING path, created_at",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "path",
+        "type_info": "Varchar"
+      },
+      {
+        "ordinal": 1,
+        "name": "created_at",
+        "type_info": "Timestamptz"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Varchar",
+        "Varchar",
+        "Text",
+        {
+          "Custom": {
+            "name": "draft_kind",
+            "kind": {
+              "Enum": [
+                "script",
+                "flow",
+                "app",
+                "raw_app",
+                "resource",
+                "variable",
+                "trigger_schedule",
+                "trigger_webhook",
+                "trigger_default_email",
+                "trigger_email",
+                "trigger_http",
+                "trigger_websocket",
+                "trigger_postgres",
+                "trigger_kafka",
+                "trigger_nats",
+                "trigger_mqtt",
+                "trigger_sqs",
+                "trigger_gcp",
+                "trigger_azure",
+                "trigger_poll",
+                "trigger_cli",
+                "trigger_nextcloud",
+                "trigger_google",
+                "trigger_github",
+                "data_pipeline",
+                "trigger_amqp"
+              ]
+            }
+          }
+        },
+        "Text",
+        "Timestamptz",
+        "Bool",
+        "Timestamptz",
+        "Text",
+        "Bool",
+        "Text"
+      ]
+    },
+    "nullable": [
+      false,
+      false
+    ]
+  },
+  "hash": "7761959d05fff8c314d04c1bc9fb4e84a61008dbec0995ae192f1b3a1c261579"
+}
@@ -0,0 +1,18 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "DELETE FROM draft_move\n         WHERE workspace_id = $1 AND typ::text = ANY($2::text[])\n           AND old_path IN ($3, $4) AND ($5::text IS NULL OR email = $5)",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Text",
+        "TextArray",
+        "Varchar",
+        "Varchar",
+        "Text"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "797896e0152046844dbfee9d0e57e0055d742e44119bfb31ae7ff37a89b34d14"
+}
@@ -0,0 +1,29 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "\n        SELECT ws.workspace_id AS \"workspace_id!\", dt.key AS \"datatable!\"\n        FROM workspace_settings ws\n        JOIN workspace w ON w.id = ws.workspace_id AND w.deleted = false\n        CROSS JOIN LATERAL jsonb_each(COALESCE(ws.datatable->'datatables', '{}'::jsonb)) dt\n        WHERE ws.workspace_id <> $1\n          AND dt.value->'database'->>'resource_type' = 'instance'\n          AND dt.value->'database'->>'resource_path' = $2\n        ORDER BY ws.workspace_id, dt.key\n        ",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "workspace_id!",
+        "type_info": "Varchar"
+      },
+      {
+        "ordinal": 1,
+        "name": "datatable!",
+        "type_info": "Text"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Text",
+        "Text"
+      ]
+    },
+    "nullable": [
+      false,
+      null
+    ]
+  },
+  "hash": "79799b5a2e499df6c28e286c42b9ad2db940c2455ab19cc95e5198baf96d5629"
+}
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "SELECT workspace.id, workspace.name, usr.username, workspace_settings.color, workspace.parent_workspace_id,\n                workspace.is_dev_workspace, workspace.dev_workspace_label,\n                workspace.owner AS \"created_by?\",\n                CASE WHEN usr.operator THEN workspace_settings.operator_settings ELSE NULL END as operator_settings,\n                usr.disabled\n         FROM workspace\n         JOIN usr ON usr.workspace_id = workspace.id\n         JOIN workspace_settings ON workspace_settings.workspace_id = workspace.id\n         WHERE usr.email = $1 AND workspace.deleted = false",
+  "query": "SELECT workspace.id, workspace.name, usr.username, workspace_settings.color, workspace.parent_workspace_id,\n                workspace.is_dev_workspace, workspace.dev_workspace_label,\n                workspace.owner AS \"created_by?\",\n                CASE WHEN usr.operator THEN workspace_settings.operator_settings ELSE NULL END as operator_settings,\n                usr.disabled, usr.is_service_account\n         FROM workspace\n         JOIN usr ON usr.workspace_id = workspace.id\n         JOIN workspace_settings ON workspace_settings.workspace_id = workspace.id\n         WHERE usr.email = $1 AND workspace.deleted = false",
  "describe": {
    "columns": [
      {
@@ -52,6 +52,11 @@
        "ordinal": 9,
        "name": "disabled",
        "type_info": "Bool"
+      },
+      {
+        "ordinal": 10,
+        "name": "is_service_account",
+        "type_info": "Bool"
      }
    ],
    "parameters": {
@@ -69,8 +74,9 @@
      true,
      false,
      null,
+      false,
      false
    ]
  },
-  "hash": "88a134e4ca82d5ce0334977c7713021ae3e99a5a61ea1c944c1df1368746dfa5"
+  "hash": "7b33adb5cf051bc123340982b97a80a96ee43e1c91d9907d50cc2ee14111623e"
 }
@@ -0,0 +1,43 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT flow_version.id, flow_version.created_at, flow_version.created_by, deployment_metadata.deployment_msg FROM flow_version\n        LEFT JOIN deployment_metadata ON flow_version.id = deployment_metadata.flow_version\n        WHERE flow_version.path = $1 AND flow_version.workspace_id = $2\n        ORDER BY flow_version.created_at DESC\n        LIMIT $3 OFFSET $4",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "id",
+        "type_info": "Int8"
+      },
+      {
+        "ordinal": 1,
+        "name": "created_at",
+        "type_info": "Timestamptz"
+      },
+      {
+        "ordinal": 2,
+        "name": "created_by",
+        "type_info": "Varchar"
+      },
+      {
+        "ordinal": 3,
+        "name": "deployment_msg",
+        "type_info": "Text"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Text",
+        "Text",
+        "Int8",
+        "Int8"
+      ]
+    },
+    "nullable": [
+      false,
+      false,
+      false,
+      true
+    ]
+  },
+  "hash": "7ff60d155510303e307c47dcb6153ccaac72871eb68822d8e296c0d747e55f35"
+}
@@ -0,0 +1,17 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "INSERT INTO datatable_role (id, name, enabled, pwd) VALUES ($1, $2, $3, $4)",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Varchar",
+        "Varchar",
+        "Bool",
+        "Text"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "86af9d51a158ea5cb6161461ecddf2a63695f8cbf8af648da5a0a77a5b9d02ba"
+}
@@ -1,16 +0,0 @@
-{
-  "db_name": "PostgreSQL",
-  "query": "INSERT INTO draft (workspace_id, path, typ, value, created_at, email)\n         SELECT $2, path, typ,\n                CASE WHEN typ IN ('script', 'flow')\n                     THEN to_json(to_jsonb(value) - 'on_behalf_of')\n                     ELSE value END,\n                created_at, email\n         FROM draft\n         WHERE workspace_id = $1 AND (email = $3 OR email IS NULL)",
-  "describe": {
-    "columns": [],
-    "parameters": {
-      "Left": [
-        "Text",
-        "Varchar",
-        "Text"
-      ]
-    },
-    "nullable": []
-  },
-  "hash": "87afd5289e54b6accda8a399ec3c66df3d19217059f62903be3c74eadb9c15d5"
-}
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "\n        SELECT\n            workspace_id,\n            slack_team_id,\n            teams_team_id,\n            teams_team_name,\n            teams_team_guid,\n            slack_name,\n            slack_command_script,\n            teams_command_script,\n            slack_email,\n            slack_oauth_client_id,\n            slack_oauth_client_secret,\n            customer_id,\n            plan,\n            webhook,\n            ai_config,\n            dbt_warehouses,\n            large_file_storage,\n            datatable,\n            ducklake,\n            git_sync,\n            deploy_ui,\n            default_app,\n            default_scripts,\n            mute_critical_alerts,\n            color,\n            operator_settings,\n            git_app_installations,\n            auto_invite,\n            error_handler,\n            success_handler,\n            public_app_execution_limit_per_minute,\n            error_handler_fallback_to_instance_alerts,\n            guest_access_enabled,\n            guest_jwt_public_key,\n            guest_jwt_jwks_url\n        FROM\n            workspace_settings\n        WHERE\n            workspace_id = $1\n        ",
+  "query": "\n        SELECT\n            workspace_id,\n            slack_team_id,\n            teams_team_id,\n            teams_team_name,\n            teams_team_guid,\n            slack_name,\n            slack_command_script,\n            teams_command_script,\n            slack_email,\n            slack_oauth_client_id,\n            slack_oauth_client_secret,\n            customer_id,\n            plan,\n            webhook,\n            ai_config,\n            dbt_warehouses,\n            large_file_storage,\n            datatable,\n            ducklake,\n            git_sync,\n            deploy_ui,\n            default_app,\n            default_scripts,\n            mute_critical_alerts,\n            color,\n            operator_settings,\n            git_app_installations,\n            auto_invite,\n            error_handler,\n            success_handler,\n            public_app_execution_limit_per_minute,\n            error_handler_fallback_to_instance_alerts,\n            guest_access_enabled,\n            guest_jwt_public_key,\n            guest_jwt_jwks_url,\n            add_admins_and_developers_to_forks\n        FROM\n            workspace_settings\n        WHERE\n            workspace_id = $1\n        ",
  "describe": {
    "columns": [
      {
@@ -177,6 +177,11 @@
        "ordinal": 34,
        "name": "guest_jwt_jwks_url",
        "type_info": "Text"
+      },
+      {
+        "ordinal": 35,
+        "name": "add_admins_and_developers_to_forks",
+        "type_info": "Bool"
      }
    ],
    "parameters": {
@@ -219,8 +224,9 @@
      false,
      false,
      true,
-      true
+      true,
+      false
    ]
  },
-  "hash": "dc4a57df3becc610f631ef22c116450390addbfae85fecc61c991d94167e6e99"
+  "hash": "8ebe054b41793f1a7b85f1f8d29cd21d12fa1207b58bfc46249c250cdcdb5363"
 }
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "\n        SELECT\n            j.args as \"args: Json<HashMap<String, Box<RawValue>>>\",\n            js.flow_status as \"flow_status: Json<windmill_common::flow_status::FlowStatus>\"\n        FROM v2_job_status js\n        INNER JOIN v2_job j ON j.id = js.id\n        WHERE js.id = $1\n        ",
+  "query": "\n        SELECT\n            j.args as \"args: Json<HashMap<String, Box<RawValue>>>\",\n            js.flow_status as \"flow_status: Json<windmill_common::flow_status::FlowStatus>\",\n            j.runnable_path\n        FROM v2_job_status js\n        INNER JOIN v2_job j ON j.id = js.id\n        WHERE js.id = $1\n        ",
  "describe": {
    "columns": [
      {
@@ -12,6 +12,11 @@
        "ordinal": 1,
        "name": "flow_status: Json<windmill_common::flow_status::FlowStatus>",
        "type_info": "Jsonb"
+      },
+      {
+        "ordinal": 2,
+        "name": "runnable_path",
+        "type_info": "Varchar"
      }
    ],
    "parameters": {
@@ -20,9 +25,10 @@
      ]
    },
    "nullable": [
+      true,
      true,
      true
    ]
  },
-  "hash": "dd89d652154748d6d7e625e31778f6885d0ee62d29a4b8894a4b459dd215a103"
+  "hash": "9008f9abb70a9a07e38acb20bea6a710d0efd77dac4aedeb88d72240e816530b"
 }
@@ -0,0 +1,17 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "DELETE FROM draft_move WHERE workspace_id = $1 AND typ::text = ANY($2::text[])\n           AND (old_path = $3 OR (new_path = $3 AND $4::text IS NULL))",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Text",
+        "TextArray",
+        "Text",
+        "Text"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "9723edfd6de38cf0bf21123a46d88fa7d0b650b5cfd80c099f785f5e411062e5"
+}
@@ -0,0 +1,15 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "UPDATE workspace_settings SET add_admins_and_developers_to_forks = $1 WHERE workspace_id = $2",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Bool",
+        "Text"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "9bd1995747f0073b3a866d1238e78f7e6dfa9185056f7731547ff05b8176b271"
+}
@@ -0,0 +1,18 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "DELETE FROM draft_move\n         WHERE workspace_id = $1 AND typ::text = ANY($2::text[])\n           AND new_path = $4 AND old_path <> $3 AND ($5::text IS NULL OR email = $5)",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Text",
+        "TextArray",
+        "Text",
+        "Text",
+        "Text"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "9d1b202844e6935e636d1570a92d46af4419dda9410ffc46f19bc38e56efe609"
+}
@@ -1,35 +0,0 @@
-{
-  "db_name": "PostgreSQL",
-  "query": "SELECT flow_version.id, flow_version.created_at, deployment_metadata.deployment_msg FROM flow_version \n        LEFT JOIN deployment_metadata ON flow_version.id = deployment_metadata.flow_version\n        WHERE flow_version.path = $1 AND flow_version.workspace_id = $2 \n        ORDER BY flow_version.created_at DESC",
-  "describe": {
-    "columns": [
-      {
-        "ordinal": 0,
-        "name": "id",
-        "type_info": "Int8"
-      },
-      {
-        "ordinal": 1,
-        "name": "created_at",
-        "type_info": "Timestamptz"
-      },
-      {
-        "ordinal": 2,
-        "name": "deployment_msg",
-        "type_info": "Text"
-      }
-    ],
-    "parameters": {
-      "Left": [
-        "Text",
-        "Text"
-      ]
-    },
-    "nullable": [
-      false,
-      false,
-      true
-    ]
-  },
-  "hash": "a0f1c0df6bc2f1fbca50edee90e42c94445536e201b322eda6f7a90bdf38f36a"
-}
@@ -0,0 +1,15 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "UPDATE workspace_settings SET auto_invite = (COALESCE(auto_invite, '{}'::jsonb) - 'domain') || $1::jsonb WHERE workspace_id = $2",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Jsonb",
+        "Text"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "a1bcf143135ecb9f32709a56a353d3246dc238920c3e155caa9cf767246fa6fb"
+}
@@ -0,0 +1,85 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT\n                 -- The guard's own predicate, ordered own row, then legacy, then another\n                 -- user's other-kind row: each needs a different remedy, and a bare\n                 -- LIMIT 1 would name an arbitrary one.\n                 (SELECT typ::text FROM draft WHERE workspace_id = $1 AND path = $3\n                  AND typ::text = ANY($6::text[])\n                  AND (typ <> $2 OR email = $4 OR email IS NULL)\n                  ORDER BY CASE WHEN email = $4 THEN 0 WHEN email IS NULL THEN 1 ELSE 2 END\n                  LIMIT 1) as \"at_target\",\n                 EXISTS(SELECT 1 FROM draft WHERE workspace_id = $1 AND path = $3\n                        AND typ::text = ANY($6::text[]) AND email = $4) as \"at_target_own!\",\n                 EXISTS(SELECT 1 FROM draft WHERE workspace_id = $1 AND path = $3\n                        AND typ::text = ANY($6::text[]) AND email IS NULL) as \"at_target_legacy!\",\n                 EXISTS(SELECT 1 FROM draft WHERE workspace_id = $1 AND path = $5\n                        AND typ = $2 AND email = $4\n                        AND position(chr(92) || 'u0000' in replace(value::text, chr(92) || chr(92), '')) > 0\n                       ) as \"poisoned!\",\n                 EXISTS(SELECT 1 FROM draft WHERE workspace_id = $1 AND path = $5\n                        AND typ = $2 AND email IS NULL) as \"legacy!\" ",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "at_target",
+        "type_info": "Text"
+      },
+      {
+        "ordinal": 1,
+        "name": "at_target_own!",
+        "type_info": "Bool"
+      },
+      {
+        "ordinal": 2,
+        "name": "at_target_legacy!",
+        "type_info": "Bool"
+      },
+      {
+        "ordinal": 3,
+        "name": "poisoned!",
+        "type_info": "Bool"
+      },
+      {
+        "ordinal": 4,
+        "name": "legacy!",
+        "type_info": "Bool"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Text",
+        {
+          "Custom": {
+            "name": "draft_kind",
+            "kind": {
+              "Enum": [
+                "script",
+                "flow",
+                "app",
+                "raw_app",
+                "resource",
+                "variable",
+                "trigger_schedule",
+                "trigger_webhook",
+                "trigger_default_email",
+                "trigger_email",
+                "trigger_http",
+                "trigger_websocket",
+                "trigger_postgres",
+                "trigger_kafka",
+                "trigger_nats",
+                "trigger_mqtt",
+                "trigger_sqs",
+                "trigger_gcp",
+                "trigger_azure",
+                "trigger_poll",
+                "trigger_cli",
+                "trigger_nextcloud",
+                "trigger_google",
+                "trigger_github",
+                "data_pipeline",
+                "trigger_amqp"
+              ]
+            }
+          }
+        },
+        "Text",
+        "Text",
+        "Text",
+        "TextArray"
+      ]
+    },
+    "nullable": [
+      null,
+      null,
+      null,
+      null,
+      null
+    ]
+  },
+  "hash": "a1ed94ef0a6d4f7bc89ac4a2341bc06dc95d60ec92469523935f5a18a4d3f907"
+}
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "SELECT id, conversation_id, message_type as \"message_type: MessageType\", content, job_id, created_at, created_seq, step_name, success\n             FROM flow_conversation_message\n             WHERE conversation_id = $1\n               AND created_seq > $2\n             ORDER BY created_seq ASC\n             LIMIT $3\n             ",
+  "query": "SELECT id, conversation_id, message_type as \"message_type: MessageType\", content, job_id, created_at, created_seq, step_name, success, tool_arguments, tool_result, reasoning, attachments\n             FROM (\n                SELECT id, conversation_id, message_type, content, job_id, created_at, created_seq, step_name, success, tool_arguments, tool_result, reasoning, attachments\n                FROM flow_conversation_message\n                WHERE conversation_id = $1\n                ORDER BY created_seq DESC\n                LIMIT $2 OFFSET $3\n             ) AS messages\n             ORDER BY created_seq ASC\n             ",
  "describe": {
    "columns": [
      {
@@ -58,6 +58,26 @@
        "ordinal": 8,
        "name": "success",
        "type_info": "Bool"
+      },
+      {
+        "ordinal": 9,
+        "name": "tool_arguments",
+        "type_info": "Text"
+      },
+      {
+        "ordinal": 10,
+        "name": "tool_result",
+        "type_info": "Text"
+      },
+      {
+        "ordinal": 11,
+        "name": "reasoning",
+        "type_info": "Text"
+      },
+      {
+        "ordinal": 12,
+        "name": "attachments",
+        "type_info": "Jsonb"
      }
    ],
    "parameters": {
@@ -76,8 +96,12 @@
      false,
      false,
      true,
-      false
+      false,
+      true,
+      true,
+      true,
+      true
    ]
  },
-  "hash": "e8802be9203c1e88a06e337260ccca029380139f89a01a89033e36a6ed9ac082"
+  "hash": "a4a823f70b3dbe6aaf4a61c98345e94c5042fd5e6351fea139a66ecb1fb812ab"
 }
@@ -0,0 +1,15 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "UPDATE draft_move SET workspace_id = $1 WHERE workspace_id = $2",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Varchar",
+        "Text"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "a725d5b99b65a1637406179b058a1ff36620099c6b87fd82a59eae1e35303c07"
+}
@@ -0,0 +1,20 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT jsonb_object_keys(value->'databases') FROM global_settings\n         WHERE name = 'custom_instance_pg_databases'",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "jsonb_object_keys",
+        "type_info": "Text"
+      }
+    ],
+    "parameters": {
+      "Left": []
+    },
+    "nullable": [
+      null
+    ]
+  },
+  "hash": "b9842d2d8abf382bd82d8fa1de012373638be391f884f81dc387ffc465badac6"
+}
@@ -1,15 +0,0 @@
-{
-  "db_name": "PostgreSQL",
-  "query": "INSERT INTO usr (workspace_id, username, email, is_admin, created_at, operator, disabled, role, is_service_account, added_via)\n         SELECT $1, username, email, is_admin, created_at, operator, disabled, role, is_service_account, added_via\n         FROM usr WHERE workspace_id = $2\n         ON CONFLICT DO NOTHING",
-  "describe": {
-    "columns": [],
-    "parameters": {
-      "Left": [
-        "Varchar",
-        "Text"
-      ]
-    },
-    "nullable": []
-  },
-  "hash": "b98844926ff127c528ed3e7bc63bf1ebed0192be267983e8fbd18f79997e6142"
-}
@@ -0,0 +1,47 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT a.id as app_id, av.id as version_id, dm.deployment_msg as deployment_msg,\n                av.created_by as created_by, av.created_at as created_at\n        FROM app a JOIN app_version av\n             ON av.id = a.versions[array_upper(a.versions, 1)] AND av.app_id = a.id\n        LEFT JOIN deployment_metadata dm ON av.id = dm.app_version\n        WHERE a.workspace_id = $1 AND a.path = $2",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "app_id",
+        "type_info": "Int8"
+      },
+      {
+        "ordinal": 1,
+        "name": "version_id",
+        "type_info": "Int8"
+      },
+      {
+        "ordinal": 2,
+        "name": "deployment_msg",
+        "type_info": "Text"
+      },
+      {
+        "ordinal": 3,
+        "name": "created_by",
+        "type_info": "Varchar"
+      },
+      {
+        "ordinal": 4,
+        "name": "created_at",
+        "type_info": "Timestamptz"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Text",
+        "Text"
+      ]
+    },
+    "nullable": [
+      false,
+      false,
+      true,
+      false,
+      false
+    ]
+  },
+  "hash": "bd8ac169ad15346f28f561233fe340efd453512c1464a413af2393ab8bfea9ee"
+}
@@ -0,0 +1,24 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT dt.key AS \"datatable!\"\n           FROM workspace_settings ws\n           CROSS JOIN LATERAL jsonb_each(COALESCE(ws.datatable->'datatables', '{}'::jsonb)) dt\n           WHERE ws.workspace_id = $1\n             AND dt.key <> $2\n             AND NOT dt.value ? 'permissions'\n             AND dt.value->'database'->>'resource_type' = 'instance'\n             AND dt.value->'database'->>'resource_path' = $3\n           ORDER BY dt.key",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "datatable!",
+        "type_info": "Text"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Text",
+        "Text",
+        "Text"
+      ]
+    },
+    "nullable": [
+      null
+    ]
+  },
+  "hash": "d48ca62c86b1af7a9dd2450c1c28dc45020a2a553d8874c49f9eafedea5a9d40"
+}
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "SELECT id, conversation_id, message_type as \"message_type: MessageType\", content, job_id, created_at, created_seq, step_name, success\n             FROM (\n                SELECT id, conversation_id, message_type, content, job_id, created_at, created_seq, step_name, success\n                FROM flow_conversation_message\n                WHERE conversation_id = $1\n                ORDER BY created_seq DESC\n                LIMIT $2 OFFSET $3\n             ) AS messages\n             ORDER BY created_seq ASC\n             ",
+  "query": "SELECT id, conversation_id, message_type as \"message_type: MessageType\", content, job_id, created_at, created_seq, step_name, success, tool_arguments, tool_result, reasoning, attachments\n             FROM flow_conversation_message\n             WHERE conversation_id = $1\n               AND created_seq > $2\n             ORDER BY created_seq ASC\n             LIMIT $3\n             ",
  "describe": {
    "columns": [
      {
@@ -58,6 +58,26 @@
        "ordinal": 8,
        "name": "success",
        "type_info": "Bool"
+      },
+      {
+        "ordinal": 9,
+        "name": "tool_arguments",
+        "type_info": "Text"
+      },
+      {
+        "ordinal": 10,
+        "name": "tool_result",
+        "type_info": "Text"
+      },
+      {
+        "ordinal": 11,
+        "name": "reasoning",
+        "type_info": "Text"
+      },
+      {
+        "ordinal": 12,
+        "name": "attachments",
+        "type_info": "Jsonb"
      }
    ],
    "parameters": {
@@ -76,8 +96,12 @@
      false,
      false,
      true,
-      false
+      false,
+      true,
+      true,
+      true,
+      true
    ]
  },
-  "hash": "1c3473a0f9f6b6148b2c975f9f05bdefedf8a51c4e6ddf0eca367b9cc778d051"
+  "hash": "d6fa78c43b6c5f8040d7bccb29ad8627be1dac6fbe0097735a52f47c173f51c9"
 }
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "SELECT id, workspace_id, flow_path, title, created_at, updated_at, created_by\n         FROM flow_conversation\n         WHERE id = $1 AND workspace_id = $2\n         FOR UPDATE",
+  "query": "SELECT id, workspace_id, flow_path, title, created_at, updated_at, created_by, is_test\n         FROM flow_conversation\n         WHERE id = $1 AND workspace_id = $2\n         FOR UPDATE",
  "describe": {
    "columns": [
      {
@@ -37,6 +37,11 @@
        "ordinal": 6,
        "name": "created_by",
        "type_info": "Varchar"
+      },
+      {
+        "ordinal": 7,
+        "name": "is_test",
+        "type_info": "Bool"
      }
    ],
    "parameters": {
@@ -52,8 +57,9 @@
      true,
      false,
      false,
+      false,
      false
    ]
  },
-  "hash": "6f32c1feed096ff706ae359ad6a3ca33b3f82ca38289dfa4a69aa95041027d57"
+  "hash": "dd84f9dfb238d18cb74f9e43228345427131bf8008eba6920021d6a449791534"
 }
@@ -0,0 +1,15 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "UPDATE draft_move SET email = $1 WHERE email = $2",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Varchar",
+        "Text"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "e428f0cdb9077bffe2b7271d9b5f30ed81e8c26a9f1e65f423078947153ff9dd"
+}
@@ -0,0 +1,16 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "INSERT INTO draft (workspace_id, path, typ, value, created_at, email, base)\n         SELECT $2, path, typ,\n                to_json(\n                    CASE WHEN typ IN ('script', 'flow') THEN clean - 'on_behalf_of' ELSE clean END\n                    - CASE WHEN typ = 'flow' THEN 'version_id'\n                           WHEN typ IN ('app', 'raw_app') THEN 'parent_version'\n                           ELSE '' END\n                ),\n                created_at, email,\n                CASE WHEN typ = 'script' THEN base END\n         FROM (\n             SELECT d.path, d.typ, d.created_at, d.email, d.base,\n                    replace(replace(replace(d.value::text, chr(92) || chr(92), chr(1)), chr(92) || 'u0000', ''), chr(1), chr(92) || chr(92))::jsonb AS clean\n             FROM draft d\n             WHERE d.workspace_id = $1 AND (d.email = $3 OR d.email IS NULL)\n         ) s",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Text",
+        "Varchar",
+        "Text"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "e6140682ebc65083f48714ab3299fdb239ef507e23003db03aff7d35b02c540b"
+}
@@ -0,0 +1,59 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT m.new_path FROM draft_move m\n           WHERE m.workspace_id = $1 AND m.typ = $2 AND m.old_path = $3\n             AND (m.email IS NULL OR m.email = $4)\n             AND NOT EXISTS (\n                 SELECT 1 FROM draft d\n                 WHERE d.workspace_id = $1 AND d.typ = $2 AND d.path = $3\n                   AND d.email IS NOT DISTINCT FROM $4\n             )\n           ORDER BY m.email IS NULL\n           LIMIT 1",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "new_path",
+        "type_info": "Varchar"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Text",
+        {
+          "Custom": {
+            "name": "draft_kind",
+            "kind": {
+              "Enum": [
+                "script",
+                "flow",
+                "app",
+                "raw_app",
+                "resource",
+                "variable",
+                "trigger_schedule",
+                "trigger_webhook",
+                "trigger_default_email",
+                "trigger_email",
+                "trigger_http",
+                "trigger_websocket",
+                "trigger_postgres",
+                "trigger_kafka",
+                "trigger_nats",
+                "trigger_mqtt",
+                "trigger_sqs",
+                "trigger_gcp",
+                "trigger_azure",
+                "trigger_poll",
+                "trigger_cli",
+                "trigger_nextcloud",
+                "trigger_google",
+                "trigger_github",
+                "data_pipeline",
+                "trigger_amqp"
+              ]
+            }
+          }
+        },
+        "Text",
+        "Text"
+      ]
+    },
+    "nullable": [
+      false
+    ]
+  },
+  "hash": "e63da92247045f2aaaa62b9d6ee17510fd05d0b7c5311b61e7257ccd07b16da2"
+}
@@ -1,6 +1,6 @@
 {
  "db_name": "PostgreSQL",
-  "query": "\n        SELECT\n            workspace_id,\n            slack_team_id,\n            slack_name,\n            teams_team_id,\n            teams_team_name,\n            teams_team_guid,\n            mute_critical_alerts,\n            guest_access_enabled,\n            deploy_ui,\n            large_file_storage,\n            datatable\n        FROM\n            workspace_settings\n        WHERE\n            workspace_id = $1\n        ",
+  "query": "\n        SELECT\n            workspace_id,\n            slack_team_id,\n            slack_name,\n            teams_team_id,\n            teams_team_name,\n            teams_team_guid,\n            mute_critical_alerts,\n            guest_access_enabled,\n            add_admins_and_developers_to_forks,\n            deploy_ui,\n            large_file_storage,\n            datatable\n        FROM\n            workspace_settings\n        WHERE\n            workspace_id = $1\n        ",
  "describe": {
    "columns": [
      {
@@ -45,16 +45,21 @@
      },
      {
        "ordinal": 8,
+        "name": "add_admins_and_developers_to_forks",
+        "type_info": "Bool"
+      },
+      {
+        "ordinal": 9,
        "name": "deploy_ui",
        "type_info": "Jsonb"
      },
      {
-        "ordinal": 9,
+        "ordinal": 10,
        "name": "large_file_storage",
        "type_info": "Jsonb"
      },
      {
-        "ordinal": 10,
+        "ordinal": 11,
        "name": "datatable",
        "type_info": "Jsonb"
      }
@@ -73,10 +78,11 @@
      true,
      true,
      false,
+      false,
      true,
      true,
      true
    ]
  },
-  "hash": "ede15bff96152f209aff756830cbc76b5afa1af6ed324376989117b1054c3447"
+  "hash": "e6e31fdf705896c81f9a0f45d47c1b93db0406aaf278bccba07be00e9f937e2a"
 }
@@ -0,0 +1,15 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "INSERT INTO workspace_settings (workspace_id, slack_team_id, slack_name, slack_command_script, slack_email, customer_id, plan, webhook, ai_config, large_file_storage, git_sync, default_app, default_scripts, deploy_ui, mute_critical_alerts, color, operator_settings, teams_command_script, teams_team_id, teams_team_name, git_app_installations, git_credentials, ducklake, dbt_warehouses, slack_oauth_client_id, slack_oauth_client_secret, datatable, teams_team_guid, auto_invite, error_handler, success_handler, public_app_execution_limit_per_minute, error_handler_fallback_to_instance_alerts, guest_access_enabled, guest_jwt_public_key, guest_jwt_jwks_url, add_admins_and_developers_to_forks) SELECT $1, slack_team_id, slack_name, slack_command_script, slack_email, customer_id, plan, webhook, ai_config, large_file_storage, git_sync, default_app, default_scripts, deploy_ui, mute_critical_alerts, color, operator_settings, teams_command_script, teams_team_id, teams_team_name, git_app_installations, git_credentials, ducklake, dbt_warehouses, slack_oauth_client_id, slack_oauth_client_secret, datatable, teams_team_guid, auto_invite, error_handler, success_handler, public_app_execution_limit_per_minute, error_handler_fallback_to_instance_alerts, guest_access_enabled, guest_jwt_public_key, guest_jwt_jwks_url, add_admins_and_developers_to_forks FROM workspace_settings WHERE workspace_id = $2",
+  "describe": {
+    "columns": [],
+    "parameters": {
+      "Left": [
+        "Varchar",
+        "Text"
+      ]
+    },
+    "nullable": []
+  },
+  "hash": "eefa0588a6a927fd9b3f65e1df652fb2b4cf7983049d2c490940df360c7e2b30"
+}
@@ -0,0 +1,41 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT flow_version.id, flow_version.created_at, flow_version.created_by, deployment_metadata.deployment_msg FROM flow_version \n        LEFT JOIN deployment_metadata ON flow_version.id = deployment_metadata.flow_version\n        WHERE flow_version.path = $1 AND flow_version.workspace_id = $2 \n        ORDER BY flow_version.created_at DESC",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "id",
+        "type_info": "Int8"
+      },
+      {
+        "ordinal": 1,
+        "name": "created_at",
+        "type_info": "Timestamptz"
+      },
+      {
+        "ordinal": 2,
+        "name": "created_by",
+        "type_info": "Varchar"
+      },
+      {
+        "ordinal": 3,
+        "name": "deployment_msg",
+        "type_info": "Text"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Text",
+        "Text"
+      ]
+    },
+    "nullable": [
+      false,
+      false,
+      false,
+      true
+    ]
+  },
+  "hash": "ef98b0fab16d3b19997fd639b74a895cce1e8ae010307431e6787d20c7b25780"
+}
@@ -0,0 +1,31 @@
+{
+  "db_name": "PostgreSQL",
+  "query": "SELECT d.email IS NULL as \"legacy!\", COALESCE(u.username, p.username) as username\n           FROM draft d\n           LEFT JOIN usr u ON u.workspace_id = d.workspace_id AND u.email = d.email\n           LEFT JOIN password p ON p.email = d.email AND p.super_admin = true\n           WHERE d.workspace_id = $1 AND d.path = $2 AND d.typ::text = ANY($3::text[])\n             -- A row a move already brought here from the very path being renamed is this\n             -- rename's own draft, not an item in its way: the record says which owners\n             -- it moved (all of them for an item move, one for a draft-only move). Any\n             -- other row is a second item and still collides.\n             AND NOT EXISTS (\n                 SELECT 1 FROM draft_move m\n                 WHERE m.workspace_id = $1 AND m.typ::text = ANY($3::text[])\n                   AND m.old_path = $4 AND m.new_path = $2\n                   AND (m.email IS NULL OR m.email IS NOT DISTINCT FROM d.email)\n             )\n           ORDER BY 2",
+  "describe": {
+    "columns": [
+      {
+        "ordinal": 0,
+        "name": "legacy!",
+        "type_info": "Bool"
+      },
+      {
+        "ordinal": 1,
+        "name": "username",
+        "type_info": "Varchar"
+      }
+    ],
+    "parameters": {
+      "Left": [
+        "Text",
+        "Text",
+        "TextArray",
+        "Text"
+      ]
+    },
+    "nullable": [
+      null,
+      null
+    ]
+  },
+  "hash": "fc2081d9077c743bc3978ae1905db558549b2356fc6ac8f0557aa40e1345ff46"
+}
@@ -1,6 +1,6 @@
 [package]
 name = "windmill"
-version = "1.813.0"
+version = "1.815.0"
 authors.workspace = true
 edition.workspace = true

@@ -88,7 +88,7 @@ members = [
 exclude = ["./windmill-duckdb-ffi-internal", "./parsers/windmill-parser-wasm"]

 [workspace.package]
-version = "1.813.0"
+version = "1.815.0"
 authors = ["Ruben Fiszel <ruben@windmill.dev>"]
 edition = "2021"

@@ -624,7 +624,6 @@ wasm-bindgen-test = "^0"
 convert_case = "0.6.0"
 getrandom = "0.2"
 tokio-postgres = {version = "^0.7", features = ["array-impls", "with-serde_json-1", "with-chrono-0_4", "with-uuid-1", "with-bit-vec-0_6"]}
-postgres-protocol = "0.6"
 rust-postgres = { package = "tokio-postgres", git = "https://github.com/MaterializeInc/rust-postgres", rev = "78c1222577bb091d69bc22b1bc7ad01c14675abe"}
 rust-postgres-native-tls = { package = "postgres-native-tls", git = "https://github.com/MaterializeInc/rust-postgres", features = ["runtime"], rev = "78c1222577bb091d69bc22b1bc7ad01c14675abe" }
 bit-vec = "=0.6.3"
@@ -1 +1 @@
-151be033ea2bf2f768f381a8cd3b888bdd366cf9
+bc3ef08c8e4233508c023e6ee847a3cd0b8be43b
@@ -0,0 +1 @@
+ALTER TABLE draft DROP COLUMN base;
@@ -0,0 +1,14 @@
+-- The version a draft forked from, as one opaque text id whatever the kind: a
+-- script hash (hex), a flow_version.id, an app_version.id. NULL for a draft that
+-- was never forked from a deploy and for kinds that keep no lineage.
+ALTER TABLE draft ADD COLUMN base TEXT;
+
+-- A U+0000 inside a `json` value makes `->>` raise 22P05; such rows keep NULL and
+-- get their base on their next save.
+UPDATE draft SET base = CASE typ::text
+    WHEN 'script' THEN value ->> 'parent_hash'
+    WHEN 'flow' THEN value ->> 'version_id'
+    ELSE value ->> 'parent_version'
+  END
+WHERE typ::text IN ('script', 'flow', 'app', 'raw_app')
+  AND position(chr(92) || 'u0000' in replace(value::text, chr(92) || chr(92), '')) = 0;
@@ -0,0 +1 @@
+DROP TABLE draft_move;
@@ -0,0 +1,17 @@
+-- Where an item's drafts went when it moved, so a draft save still addressed to the
+-- old path (an editor left open across the move, a chat, the CLI) lands on the moved
+-- draft instead of starting a new item there. `email` NULL records a deployed item's
+-- move and applies to every user; set, it records that user's draft-only move.
+CREATE TABLE draft_move (
+    id BIGINT GENERATED ALWAYS AS IDENTITY PRIMARY KEY,
+    workspace_id VARCHAR(50) NOT NULL REFERENCES workspace(id) ON DELETE CASCADE,
+    typ DRAFT_KIND NOT NULL,
+    old_path VARCHAR(255) NOT NULL,
+    new_path VARCHAR(255) NOT NULL,
+    email VARCHAR(255)
+);
+
+CREATE INDEX draft_move_old_path_idx ON draft_move (workspace_id, typ, old_path);
+
+GRANT ALL ON draft_move TO windmill_user;
+GRANT ALL ON draft_move TO windmill_admin;
@@ -0,0 +1,3 @@
+ALTER TABLE flow_conversation_message DROP COLUMN tool_arguments;
+ALTER TABLE flow_conversation_message DROP COLUMN tool_result;
+ALTER TABLE flow_conversation_message DROP COLUMN reasoning;
@@ -0,0 +1,12 @@
+-- A chat is rebuilt from its rows without reading jobs, so every tool row carries its call:
+-- the arguments the model wrote and the text the model got back, or what the call failed
+-- with. A script or flow tool's job holds the args its input transforms produced, not the
+-- model's; an MCP tool runs inside the agent's job, whose result lists every call of the
+-- turn with nothing tying one to a row. A provider-native web search carries only its
+-- citations, the provider never returning the query.
+ALTER TABLE flow_conversation_message ADD COLUMN tool_arguments TEXT;
+ALTER TABLE flow_conversation_message ADD COLUMN tool_result TEXT;
+
+-- The thinking behind this row. The agent job keeps the turn's thinking as one string;
+-- the rows keep it per iteration, next to the answer or tool call it led to.
+ALTER TABLE flow_conversation_message ADD COLUMN reasoning TEXT;
@@ -0,0 +1 @@
+ALTER TABLE flow_conversation DROP COLUMN is_test;
@@ -0,0 +1,26 @@
+-- A chat run from the flow editor's test panel is stored exactly like one from the
+-- deployed flow, so the two were indistinguishable once written. Marking them lets the
+-- lists tell a trial apart from a real conversation.
+ALTER TABLE flow_conversation ADD COLUMN is_test BOOLEAN NOT NULL DEFAULT false;
+
+-- Existing rows: a conversation whose messages came from a flowpreview run was a test.
+-- Derived once here because the job is purged on retention, after which the origin of an
+-- old conversation is unknowable.
+--
+-- Walked to the root job rather than matched directly: an existing message row never holds
+-- the flow job itself. The rows point at the step that produced them — the AI agent's job
+-- for an answer, the tool's own job for a tool call — whose kind is never 'flowpreview'.
+--
+-- `root_job` first, matching `get_root_job_id` (windmill-worker/src/common.rs): only it
+-- reaches the top of the run. `flow_innermost_root_job` stops at the closest flow scope by
+-- design, so an agent inside a subflow would land on that subflow's 'flow' row and the
+-- conversation would read as deployed.
+UPDATE flow_conversation c
+SET is_test = true
+WHERE EXISTS (
+    SELECT 1 FROM flow_conversation_message m
+    JOIN v2_job j ON j.id = m.job_id
+    JOIN v2_job root
+      ON root.id = coalesce(j.root_job, j.flow_innermost_root_job, j.parent_job, j.id)
+    WHERE m.conversation_id = c.id AND root.kind = 'flowpreview'
+);
@@ -0,0 +1 @@
+ALTER TABLE flow_conversation_message DROP COLUMN attachments;
@@ -0,0 +1,4 @@
+-- The files a user message carried, as object-storage references: `[{input, s3, storage?,
+-- filename?}]`. Only references, never file bytes and never a presigned URL, so a
+-- transcript can show a message's files without reading its run's args.
+ALTER TABLE flow_conversation_message ADD COLUMN attachments JSONB;
@@ -1,22 +0,0 @@
-- Roles on the external cluster are live logins there; dropping the column would forget them.
-LOCK TABLE datatable_role;
-DO $$
-BEGIN
-    IF EXISTS (SELECT 1 FROM datatable_role WHERE cluster <> 'instance') THEN
-        RAISE EXCEPTION 'datatable_role holds roles on the external instance cluster. Delete them in instance settings first.';
-    END IF;
-    -- Before this, only data tables on Windmill's own cluster could be under roles, and a role
-    -- block left with just `admin` survives deleting every external role.
-    IF EXISTS (
-        SELECT 1 FROM workspace_settings ws,
-            jsonb_each(CASE WHEN jsonb_typeof(ws.datatable->'datatables') = 'object'
-                THEN ws.datatable->'datatables' ELSE '{}'::jsonb END) dt
-        WHERE dt.value->'database'->>'resource_type' = 'external_instance'
-          AND dt.value ? 'permissions'
-    ) THEN
-        RAISE EXCEPTION 'external instance data tables are still under roles. Turn their roles off first.';
-    END IF;
-END $$;
-ALTER TABLE datatable_role DROP CONSTRAINT datatable_role_cluster_name_key;
-ALTER TABLE datatable_role ADD CONSTRAINT datatable_role_name_key UNIQUE (name);
-ALTER TABLE datatable_role DROP COLUMN cluster;
@@ -1,8 +0,0 @@
-- A data table role is a Postgres login on one cluster: Windmill's own ('instance'), or the external
-- instance cluster ('external_instance'). Role names are the cluster's own key, so they are unique
-- per cluster rather than across the instance.
-ALTER TABLE datatable_role
-    ADD COLUMN cluster VARCHAR(20) NOT NULL DEFAULT 'instance'
-        CHECK (cluster IN ('instance', 'external_instance'));
-ALTER TABLE datatable_role DROP CONSTRAINT datatable_role_name_key;
-ALTER TABLE datatable_role ADD CONSTRAINT datatable_role_cluster_name_key UNIQUE (cluster, name);
@@ -0,0 +1 @@
+ALTER TABLE workspace_settings DROP COLUMN add_admins_and_developers_to_forks;
@@ -0,0 +1 @@
+ALTER TABLE workspace_settings ADD COLUMN add_admins_and_developers_to_forks BOOLEAN NOT NULL DEFAULT false;
@@ -6191,7 +6191,7 @@ checksum = "712e227841d057c1ee1cd2fb22fa7e5a5461ae8e48fa2ca79ec42cfc1931183f"

 [[package]]
 name = "windmill-common"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "aho-corasick",
 "anyhow",
@@ -6274,7 +6274,7 @@ dependencies = [

 [[package]]
 name = "windmill-macros"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "proc-macro2",
 "quote",
@@ -6286,7 +6286,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "convert_case",
 "serde",
@@ -6295,7 +6295,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-bash"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "lazy_static",
@@ -6307,7 +6307,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-csharp"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "serde_json",
@@ -6319,7 +6319,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-go"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "gosyn",
@@ -6331,7 +6331,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-graphql"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "lazy_static",
@@ -6343,7 +6343,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-java"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "serde_json",
@@ -6355,7 +6355,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-nu"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "nu-parser",
@@ -6366,7 +6366,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-php"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "itertools 0.14.0",
@@ -6377,7 +6377,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-py"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "itertools 0.14.0",
@@ -6389,7 +6389,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-py-asset"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "rustpython-ast",
@@ -6400,7 +6400,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-py-imports"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "async-recursion",
@@ -6422,7 +6422,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-r"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "serde_json",
@@ -6434,7 +6434,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-ruby"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "lazy_static",
@@ -6448,7 +6448,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-rust"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "convert_case",
@@ -6465,7 +6465,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-sql"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "lazy_static",
@@ -6478,7 +6478,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-sql-asset"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "serde",
@@ -6490,7 +6490,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-ts"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "lazy_static",
@@ -6508,7 +6508,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-ts-asset"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "serde-wasm-bindgen",
@@ -6524,7 +6524,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-wac"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "rustpython-ast",
@@ -6540,7 +6540,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-wasm"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "getrandom 0.2.17",
@@ -6572,7 +6572,7 @@ dependencies = [

 [[package]]
 name = "windmill-parser-yaml"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "lazy_static",
@@ -6586,7 +6586,7 @@ dependencies = [

 [[package]]
 name = "windmill-types"
-version = "1.813.0"
+version = "1.815.0"
 dependencies = [
 "anyhow",
 "bitflags",
@@ -12,7 +12,7 @@ resolver = "2"
 members = ["."]

 [workspace.package]
-version = "1.813.0"
+version = "1.815.0"
 edition = "2021"
 authors = ["Ruben Fiszel <ruben@windmill.dev>"]

@@ -52,7 +52,8 @@ use windmill_common::{
        INSTANCE_EVENTS_WEBHOOK_SETTING, INSTANCE_PYTHON_VERSION_SETTING,
        JOB_DEFAULT_TIMEOUT_SECS_SETTING, JOB_ISOLATION_SETTING, JWT_SECRET_SETTING,
        KEEP_JOB_DIR_SETTING, LICENSE_KEY_SETTING, MAVEN_REPOS_SETTING, MAVEN_SETTINGS_XML_SETTING,
-        MONITOR_LOGS_ON_OBJECT_STORE_SETTING, NO_DEFAULT_MAVEN_SETTING,
+        MCP_DISABLE_TOKEN_QUERY_PARAM_SETTING, MONITOR_LOGS_ON_OBJECT_STORE_SETTING,
+        NO_DEFAULT_MAVEN_SETTING,
        NPM_CONFIG_REGISTRY_SETTING, NSJAIL_TMPFS_SIZE_MB_SETTING, NSJAIL_TMP_BACKING_SETTING,
        NUGET_CONFIG_SETTING, OAUTH_SETTING, OTEL_SETTING, OTEL_TRACES_RETENTION_SECS_SETTING,
        OTEL_TRACING_PROXY_SETTING, PIP_INDEX_URL_SETTING, POWERSHELL_REPO_PAT_SETTING,
@@ -126,7 +127,8 @@ use windmill_worker::{

 use crate::monitor::{
    initial_load, load_concurrency_key_max_queued, load_disable_password_login,
-    load_fork_workspace_tag_append_fork_suffix, load_keep_job_dir, load_metrics_debug_enabled,
+    load_fork_workspace_tag_append_fork_suffix, load_keep_job_dir,
+    load_mcp_disable_token_query_param, load_metrics_debug_enabled,
    load_preview_tags_override, load_require_preexisting_user, load_retention_period_overrides,
    load_tag_per_workspace_enabled, load_tag_per_workspace_workspaces,
    load_workspace_fairness_duration_secs, load_workspace_fairness_enabled,
@@ -2164,6 +2166,9 @@ async fn process_notify_event(
                DISABLE_PASSWORD_LOGIN_SETTING => {
                    load_disable_password_login(db).await;
                }
+                MCP_DISABLE_TOKEN_QUERY_PARAM_SETTING => {
+                    load_mcp_disable_token_query_param(db).await;
+                }
                EXPOSE_METRICS_SETTING => {
                    tracing::info!("Metrics setting changed, restarting");
                    spawn_graceful_killpill(tx, db, 30, "metrics setting change", server_mode)
@@ -45,7 +45,10 @@ use windmill_common::otel_oss::{
 use windmill_common::{
    agent_workers::DECODED_AGENT_TOKEN,
    apps::APP_WORKSPACED_ROUTE,
-    auth::{create_token_for_owner, ephemeral_script_token_label, job_token_expiry_secs},
+    auth::{
+        create_token_for_owner, ephemeral_script_token_label, job_token_expiry_secs,
+        TOKEN_EXPIRY_WARNING_DAYS,
+    },
    ee_oss::CriticalErrorChannel,
    email_oss::send_email_if_possible,
    error,
@@ -62,6 +65,7 @@ use windmill_common::{
        FORK_WORKSPACE_TAG_APPEND_FORK_SUFFIX_SETTING, HUB_API_SECRET_SETTING,
        HUB_BASE_URL_SETTING, INSTANCE_PYTHON_VERSION_SETTING, JOB_DEFAULT_TIMEOUT_SECS_SETTING,
        JOB_ISOLATION_SETTING, JWT_SECRET_SETTING, KEEP_JOB_DIR_SETTING, LICENSE_KEY_SETTING,
+        MCP_DISABLE_TOKEN_QUERY_PARAM, MCP_DISABLE_TOKEN_QUERY_PARAM_SETTING,
        MONITOR_LOGS_ON_OBJECT_STORE_SETTING, NPMRC_SETTING, NPM_CONFIG_REGISTRY_SETTING,
        NSJAIL_TMPFS_SIZE_MB_SETTING, NSJAIL_TMP_BACKING_SETTING, NUGET_CONFIG_SETTING,
        OTEL_SETTING, OTEL_TRACES_RETENTION_SECS_SETTING, OTEL_TRACING_PROXY_SETTING,
@@ -285,6 +289,15 @@ pub async fn initial_load(
    );

    if let Some(db) = conn.as_sql() {
+        // Outside the `server_mode` block below: a `MODE=mcp` process serves the MCP routes
+        // with `server_mode` false and would otherwise never read this at all. That mode
+        // joins no monitor loop, so there — as for every global setting, `base_url`
+        // included — this pass is the only read, and a change lands on restart.
+        pass.setting(
+            MCP_DISABLE_TOKEN_QUERY_PARAM_SETTING,
+            false,
+            |v| async move { apply_mcp_disable_token_query_param(v) },
+        );
        pass.setting(DEFAULT_TAGS_PER_WORKSPACE_SETTING, false, |v| async move {
            apply_tag_per_workspace_enabled(v)
        });
@@ -1614,6 +1627,23 @@ pub fn apply_disable_password_login(value: Option<serde_json::Value>) {
    };
 }

+pub async fn load_mcp_disable_token_query_param(db: &DB) {
+    match load_value_from_global_settings(db, MCP_DISABLE_TOKEN_QUERY_PARAM_SETTING).await {
+        Ok(v) => apply_mcp_disable_token_query_param(v),
+        Err(e) => tracing::error!("Error loading mcp_disable_token_query_param setting: {e:#}"),
+    };
+}
+
+pub fn apply_mcp_disable_token_query_param(value: Option<serde_json::Value>) {
+    match value {
+        Some(serde_json::Value::Bool(t)) => {
+            MCP_DISABLE_TOKEN_QUERY_PARAM.store(t, Ordering::Relaxed)
+        }
+        None => MCP_DISABLE_TOKEN_QUERY_PARAM.store(false, Ordering::Relaxed),
+        _ => (),
+    };
+}
+
 struct LogFile {
    file_path: String,
    hostname: String,
@@ -2198,7 +2228,7 @@ async fn cleanup_scheduled_job_deletions(db: &Pool<Postgres>) {
 }

 pub async fn check_expiring_tokens(db: &DB) {
-    // Find tokens expiring within 7 days that still have a pending notification row.
+    // Find tokens expiring within the warning window that still have a pending notification row.
    // The notification table stores token_hash (not plaintext) so the join works
    // even after the hash migration makes token.token nullable.
    let expiring_tokens_r = sqlx::query_as!(
@@ -2207,8 +2237,9 @@ pub async fn check_expiring_tokens(db: &DB) {
         USING token t
         WHERE n.token_hash = t.token_hash
           AND n.expiration > now()
-           AND n.expiration <= now() + interval '7 days'
+           AND n.expiration <= now() + make_interval(days => $1)
         RETURNING t.token_prefix, t.label, t.email, t.workspace_id",
+        TOKEN_EXPIRY_WARNING_DAYS,
    )
    .fetch_all(db)
    .await;
@@ -99,9 +99,9 @@ email_trigger: path(char), local_part(char), workspaced_local_part(bool), script
 favorite: usr(char), workspace_id(char), path(char), favorite_kind(favorite_kind)
 flow: workspace_id(char), path(char), summary(text), description(text), value(jsonb), edited_by(char), edited_at(ts), archived(bool), schema(json), extra_perms(jsonb), dependency_job(uuid), draft_only(bool), tag(char), ws_error_handler_muted(bool), dedicated_worker(bool), timeout(int), visible_to_runner_only(bool), concurrency_key(char), versions(bigint[]), on_behalf_of(varchar), on_behalf_of_email(text), lock_error_logs(text), labels(text[])
  FK: (workspace_id) -> workspace(id)
-flow_conversation: id(uuid), workspace_id(char), flow_path(char), title(char), created_at(ts), updated_at(ts), created_by(char)
+flow_conversation: id(uuid), workspace_id(char), flow_path(char), title(char), created_at(ts), updated_at(ts), created_by(char), is_test(bool)
  FK: (workspace_id) -> workspace(id)
-flow_conversation_message: id(uuid), conversation_id(uuid), message_type(message_type), content(text), job_id(uuid), created_at(ts), created_seq(int8), step_name(char), success(bool)
+flow_conversation_message: id(uuid), conversation_id(uuid), message_type(message_type), content(text), job_id(uuid), created_at(ts), created_seq(int8), step_name(char), success(bool), tool_arguments(text), tool_result(text), reasoning(text), attachments(jsonb)
  FK: (conversation_id) -> flow_conversation(id) | (job_id) -> v2_job(id)
 flow_iterator_data: job_id(uuid), itered(jsonb)
 flow_node: id(bigint), workspace_id(char), hash(bigint), path(char), lock(text), code(text), flow(jsonb), hash_v2(char(64))
@@ -234,7 +234,7 @@ workspace_protection_rule: workspace_id(char), name(char), rules(int), bypass_gr
  FK: (workspace_id) -> workspace(id)
 workspace_runnable_dependencies: flow_path(char), runnable_path(char), script_hash(bigint), runnable_is_flow(bool), workspace_id(char), app_path(char), id(bigint), runnable_is_agent(bool)
  FK: (app_path, workspace_id) -> app(path, workspace_id) | (flow_path, workspace_id) -> flow(path, workspace_id)
-workspace_settings: workspace_id(char), slack_team_id(char), slack_name(char), slack_command_script(char), slack_email(char), customer_id(char), plan(char), webhook(text), ai_config(jsonb), large_file_storage(jsonb), git_sync(jsonb), default_app(char), default_scripts(jsonb), deploy_ui(jsonb), mute_critical_alerts(bool), color(char), operator_settings(jsonb), teams_command_script(text), teams_team_id(text), teams_team_name(text), git_app_installations(jsonb), ducklake(jsonb), slack_oauth_client_id(char), slack_oauth_client_secret(char), datatable(jsonb), teams_team_guid(text), auto_invite(jsonb), error_handler(jsonb), success_handler(jsonb), public_app_execution_limit_per_minute(int), dbt_warehouses(jsonb), guest_access_enabled(bool), guest_jwt_public_key(text), guest_jwt_jwks_url(text), ai_sessions_backup_generation(int)
+workspace_settings: workspace_id(char), slack_team_id(char), slack_name(char), slack_command_script(char), slack_email(char), customer_id(char), plan(char), webhook(text), ai_config(jsonb), large_file_storage(jsonb), git_sync(jsonb), default_app(char), default_scripts(jsonb), deploy_ui(jsonb), mute_critical_alerts(bool), color(char), operator_settings(jsonb), teams_command_script(text), teams_team_id(text), teams_team_name(text), git_app_installations(jsonb), ducklake(jsonb), slack_oauth_client_id(char), slack_oauth_client_secret(char), datatable(jsonb), teams_team_guid(text), auto_invite(jsonb), error_handler(jsonb), success_handler(jsonb), public_app_execution_limit_per_minute(int), dbt_warehouses(jsonb), guest_access_enabled(bool), guest_jwt_public_key(text), guest_jwt_jwks_url(text), ai_sessions_backup_generation(int), add_admins_and_developers_to_forks(bool)
  FK: (workspace_id) -> workspace(id)
 zombie_job_counter: job_id(uuid), counter(int)
  FK: (job_id) -> v2_job(id)
@@ -0,0 +1,193 @@
+//! The deployed order of an app's versions is the order they were appended to
+//! `app.versions`, not the order of their `created_at`.
+//!
+//! `app_version.created_at` defaults to `now()`, which in Postgres is the
+//! transaction's start time, while the append happens under the app row's lock.
+//! Two deploys that overlap therefore land in one order and carry timestamps in
+//! the other. The head the editor guards against, and the sequence the diff
+//! picker numbers, both have to follow the array.
+//!
+//! Users from the `base` fixture: test-user (admin, token SECRET_TOKEN).
+
+use serde_json::json;
+use sqlx::{Pool, Postgres};
+use windmill_test_utils::*;
+
+const TOKEN: &str = "SECRET_TOKEN";
+
+#[sqlx::test(fixtures("base"))]
+async fn test_app_head_follows_the_append_order_not_the_timestamps(
+    db: Pool<Postgres>,
+) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let ws = format!(
+        "http://localhost:{}/api/w/test-workspace",
+        server.addr.port()
+    );
+    let client = reqwest::Client::new();
+
+    let res = client
+        .post(format!("{ws}/apps/create"))
+        .header("Authorization", format!("Bearer {TOKEN}"))
+        .json(&json!({
+            "path": "u/test-user/order_app",
+            "summary": "ordered",
+            "value": {},
+            "policy": { "execution_mode": "publisher", "triggerables": {} }
+        }))
+        .send()
+        .await?;
+    assert!(res.status().is_success(), "{}", res.text().await?);
+
+    let first: i64 = sqlx::query_scalar(
+        "SELECT versions[array_upper(versions, 1)] FROM app
+         WHERE workspace_id = 'test-workspace' AND path = 'u/test-user/order_app'",
+    )
+    .fetch_one(&db)
+    .await?;
+
+    // The overlapping deploy: appended after `first`, so it is the version that
+    // landed, but stamped before it, so a timestamp sort puts it underneath.
+    let second: i64 = sqlx::query_scalar(
+        "INSERT INTO app_version (app_id, value, created_by, created_at, raw_app)
+         SELECT app_id, value, 'racer', created_at - interval '1 hour', raw_app
+         FROM app_version WHERE id = $1
+         RETURNING id",
+    )
+    .bind(first)
+    .fetch_one(&db)
+    .await?;
+    sqlx::query(
+        "UPDATE app SET versions = array_append(versions, $1::bigint)
+         WHERE workspace_id = 'test-workspace' AND path = 'u/test-user/order_app'",
+    )
+    .bind(second)
+    .execute(&db)
+    .await?;
+
+    let head: serde_json::Value = client
+        .get(format!(
+            "{ws}/apps/get_latest_version/u/test-user/order_app"
+        ))
+        .header("Authorization", format!("Bearer {TOKEN}"))
+        .send()
+        .await?
+        .json()
+        .await?;
+    assert_eq!(
+        head["version"], second,
+        "the head is the version appended last, not the newest timestamp: {head}"
+    );
+
+    let history: Vec<serde_json::Value> = client
+        .get(format!("{ws}/apps/history/p/u/test-user/order_app"))
+        .header("Authorization", format!("Bearer {TOKEN}"))
+        .send()
+        .await?
+        .json()
+        .await?;
+    let listed: Vec<i64> = history
+        .iter()
+        .map(|v| v["version"].as_i64().unwrap())
+        .collect();
+    assert_eq!(
+        listed,
+        vec![second, first],
+        "the history lists in deployed order, so it leads with the head"
+    );
+
+    // Enough versions that a page cannot hold them, so "asked for nothing" and "asked for
+    // a page" are visibly different answers.
+    let mut appended = vec![second, first];
+    for _ in 0..24 {
+        let extra: i64 = sqlx::query_scalar(
+            "INSERT INTO app_version (app_id, value, created_by, created_at, raw_app)
+             SELECT app_id, value, 'bulk', created_at, raw_app FROM app_version WHERE id = $1
+             RETURNING id",
+        )
+        .bind(first)
+        .fetch_one(&db)
+        .await?;
+        sqlx::query(
+            "UPDATE app SET versions = array_append(versions, $1::bigint)
+             WHERE workspace_id = 'test-workspace' AND path = 'u/test-user/order_app'",
+        )
+        .bind(extra)
+        .execute(&db)
+        .await?;
+        appended.insert(0, extra);
+    }
+
+    let versions_at = |query: &str| {
+        let url = format!("{ws}/apps/history/p/u/test-user/order_app{query}");
+        let client = client.clone();
+        async move {
+            let rows: Vec<serde_json::Value> = client
+                .get(url)
+                .header("Authorization", format!("Bearer {TOKEN}"))
+                .send()
+                .await?
+                .json()
+                .await?;
+            Ok::<_, anyhow::Error>(
+                rows.iter()
+                    .map(|v| v["version"].as_i64().unwrap())
+                    .collect::<Vec<_>>(),
+            )
+        }
+    };
+
+    // The deployment-history panel and the CLI read this endpoint without paging, so
+    // asking for no page has to keep answering with the whole history.
+    assert_eq!(
+        versions_at("").await?,
+        appended,
+        "an unpaginated request still answers whole"
+    );
+    assert_eq!(
+        versions_at("?per_page=10").await?,
+        appended[..10],
+        "a page holds what was asked for, newest first"
+    );
+    assert_eq!(
+        versions_at("?per_page=10&page=2").await?,
+        appended[10..20],
+        "the next page carries on where the first left off, skipping nothing"
+    );
+    // A page past the end runs off it rather than overflowing into one. (The clamp on an
+    // asked-for size is pinned where it lives, in `paginate_optional`'s own test.)
+    assert!(
+        versions_at("?per_page=10&page=99999999")
+            .await?
+            .is_empty(),
+        "a page past the end is empty"
+    );
+    // Naming only the page still asks for a page, rather than an offset into a listing
+    // sized to hold everything — which would answer nothing at all.
+    assert_eq!(
+        versions_at("?page=1").await?,
+        appended,
+        "a first page with no size named holds the history"
+    );
+
+    // A version that never entered the deployed sequence — one a restore or a fork copy
+    // could leave behind — still has to be reachable, after the ones that did.
+    let stray: i64 = sqlx::query_scalar(
+        "INSERT INTO app_version (app_id, value, created_by, created_at, raw_app)
+         SELECT app_id, value, 'stray', created_at, raw_app FROM app_version WHERE id = $1
+         RETURNING id",
+    )
+    .bind(first)
+    .fetch_one(&db)
+    .await?;
+    let mut with_stray = appended.clone();
+    with_stray.push(stray);
+    assert_eq!(
+        versions_at("").await?,
+        with_stray,
+        "a version outside app.versions is listed, and after the deployed sequence"
+    );
+
+    Ok(())
+}
@@ -521,10 +521,8 @@ def main():
            .await
            .unwrap();

-        assert_eq!(
-            res.text().await.unwrap(),
-            "app f/rel/root_app updated (npath: \"f/rel/root_app_renamed\")"
-        );
+        let deployed: serde_json::Value = res.json().await.unwrap();
+        assert_eq!(deployed["path"], "f/rel/root_app_renamed", "{deployed}");

        let mut completed = listen_for_completed_jobs(&db).await;
        in_test_worker(&db, completed.next(), port).await;
@@ -0,0 +1,74 @@
+//! Moving a draft must carry both of its path keys.
+//!
+//! A draft value holds a typed path and a mirror the editors keep beside it while
+//! it differs from the row's path (`path`/`draft_path`; which is which depends on
+//! the kind). The loaders prefer the mirror, so a move that rewrote only the typed
+//! key left the mirror naming the old location: reopening the item restored the
+//! old path, and the next autosave wrote it back — undoing the move silently.
+//!
+//! This pins `move_draft`, including that a draft with no mirror never gains one.
+//! A deploy's carry follows the keys only where they still name the old path
+//! (`drafts_save_follows_move` pins that).
+
+use serde_json::Value;
+use sqlx::{Pool, Postgres};
+
+use windmill_test_utils::*;
+
+async fn move_to(port: u16, from: &str, to: &str) -> anyhow::Result<String> {
+    let resp = reqwest::Client::new()
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/move/script/{from}"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&serde_json::json!({ "new_path": to }))
+        .send()
+        .await?;
+    // Asserted here so a rejected move fails as itself, rather than as a JSON
+    // parse error in the read-back below.
+    let status = resp.status();
+    let body = resp.text().await?;
+    assert!(status.is_success(), "move {from} -> {to} failed: {body}");
+    Ok(body)
+}
+
+/// The stored draft value at `path`, read back through the API so this test needs
+/// no `sqlx::query!` (which would want an offline cache entry of its own).
+async fn value_at(port: u16, path: &str) -> anyhow::Result<Value> {
+    let body: Value = reqwest::Client::new()
+        .get(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/get_own/script/{path}"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .send()
+        .await?
+        .json()
+        .await?;
+    Ok(body
+        .get("value")
+        .cloned()
+        .unwrap_or_else(|| panic!("no draft at {path}: {body}")))
+}
+
+#[sqlx::test(fixtures("base", "drafts_move_mirror"))]
+async fn test_move_carries_both_path_keys(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+
+    move_to(port, "u/test-user/draft_mirror", "u/test-user/renamed").await?;
+    move_to(port, "u/test-user/draft_plain", "u/test-user/plain2").await?;
+
+    // The mirror follows: left at `u/test-user/friendly` it would win at load and
+    // walk the item back there.
+    let moved = value_at(port, "u/test-user/renamed").await?;
+    assert_eq!(moved["path"], "u/test-user/renamed");
+    assert_eq!(moved["draft_path"], "u/test-user/renamed");
+
+    // A draft that never had a mirror must not be given one.
+    let plain = value_at(port, "u/test-user/plain2").await?;
+    assert_eq!(plain["path"], "u/test-user/plain2");
+    assert_eq!(plain.get("draft_path"), None, "mirror injected: {plain}");
+
+    Ok(())
+}
@@ -0,0 +1,250 @@
+//! A rename onto a path that already holds a draft is refused.
+//!
+//! Nothing deployed can sit at a rename's destination (the deploy conflicts on
+//! that), but a draft can: a never-deployed item, or a draft left on an archived
+//! script. Moving onto it would merge two items or strand a row, so the rename
+//! itself fails, in its own transaction, and the source stays deployed. The
+//! destination draft here is the deployer's own, which is the same collision.
+
+use serde_json::{json, Value};
+use sqlx::{Pool, Postgres};
+
+use windmill_test_utils::*;
+
+/// Hex form of script hash 7010, the way the API takes a parent hash.
+const HEAD_HASH: &str = "0000000000001b62";
+
+#[sqlx::test(fixtures("base", "drafts_move_taken"))]
+async fn test_rename_onto_a_draft_is_refused(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+    let client = reqwest::Client::new();
+
+    let resp = client
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/scripts/create"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({
+            "path": "u/test-user/mvtaken_b",
+            "parent_hash": HEAD_HASH,
+            "summary": "A",
+            "description": "",
+            "content": "export function main() { return 1 }",
+            "language": "deno",
+            "schema": {}
+        }))
+        .send()
+        .await?;
+    let status = resp.status();
+    let body = resp.text().await?;
+    assert_eq!(status, 400, "rename onto a draft was not refused: {body}");
+    assert!(
+        body.contains("already has a draft"),
+        "unexpected refusal: {body}"
+    );
+
+    // The whole deploy rolled back: the source is still the live head, and the
+    // draft at the destination is untouched.
+    let head: Value = client
+        .get(format!(
+            "http://localhost:{port}/api/w/test-workspace/scripts/get/p/u/test-user/mvtaken_a"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .send()
+        .await?
+        .json()
+        .await?;
+    assert_eq!(head["hash"], HEAD_HASH, "source was replaced: {head}");
+    assert_eq!(head["archived"], false, "source was archived: {head}");
+
+    let draft: Value = client
+        .get(format!("http://localhost:{port}/api/w/test-workspace/drafts/get_own/script/u/test-user/mvtaken_b"))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .send()
+        .await?
+        .json()
+        .await?;
+    assert_eq!(
+        draft["value"]["summary"], "B",
+        "destination draft changed: {draft}"
+    );
+
+    Ok(())
+}
+
+/// A legacy (ownerless) draft occupies its path too: a deploy there deletes it together
+/// with the caller's own row, so a move that parks a second draft beside it would discard
+/// edits the caller never saw. Only an admin can clear it, so the refusal says so.
+#[sqlx::test(fixtures("base", "drafts_move_taken"))]
+async fn test_draft_move_refuses_a_legacy_destination(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+
+    let resp = reqwest::Client::new()
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/move/script/u/test-user/mvtaken_b"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({ "new_path": "u/test-user/mvtaken_legacy" }))
+        .send()
+        .await?;
+    let status = resp.status();
+    let body = resp.text().await?;
+    assert_eq!(status, 400, "move onto a legacy draft was allowed: {body}");
+    assert!(
+        body.contains("legacy workspace draft") && body.contains("workspace admin"),
+        "the refusal did not point at the one remedy: {body}"
+    );
+
+    // Both rows stayed where they were: the caller's own, and the legacy one the list
+    // synthesizes under the caller's name.
+    let list: Vec<Value> = reqwest::Client::new()
+        .get(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/list"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .send()
+        .await?
+        .json()
+        .await?;
+    let mut at = list
+        .iter()
+        .filter_map(|d| Some((d["kind"].as_str()?, d["path"].as_str()?)))
+        .filter(|(_, p)| p.starts_with("u/test-user/mvtaken_b") || p.ends_with("mvtaken_legacy"))
+        .collect::<Vec<_>>();
+    at.sort();
+    assert_eq!(
+        at,
+        vec![
+            ("script", "u/test-user/mvtaken_b"),
+            ("script", "u/test-user/mvtaken_legacy")
+        ],
+        "{list:?}"
+    );
+    Ok(())
+}
+
+/// A classic app and a raw app deploy into the same table, so a draft-only move onto
+/// the other kind's draft must be refused: deploying either path afterwards deletes
+/// the caller's drafts of both kinds, taking the loser's item with it.
+#[sqlx::test(fixtures("base", "drafts_move_taken"))]
+async fn test_draft_move_refuses_the_other_app_kind(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+
+    let resp = reqwest::Client::new()
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/move/raw_app/u/test-user/mvtaken_raw"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({ "new_path": "u/test-user/mvtaken_app" }))
+        .send()
+        .await?;
+    let status = resp.status();
+    let body = resp.text().await?;
+    assert_eq!(
+        status, 400,
+        "move onto a classic app draft was allowed: {body}"
+    );
+    assert!(
+        body.contains("already have a draft at 'u/test-user/mvtaken_app' (app)"),
+        "the refusal did not name the occupying kind: {body}"
+    );
+
+    // Both drafts are untouched.
+    let list: Vec<Value> = reqwest::Client::new()
+        .get(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/list"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .send()
+        .await?
+        .json()
+        .await?;
+    let mut at = list
+        .iter()
+        .filter(|d| matches!(d["kind"].as_str(), Some("app") | Some("raw_app")))
+        .filter_map(|d| Some((d["kind"].as_str()?, d["path"].as_str()?)))
+        .collect::<Vec<_>>();
+    at.sort();
+    assert_eq!(
+        at,
+        vec![
+            ("app", "u/test-user/mvtaken_app"),
+            ("raw_app", "u/test-user/mvtaken_raw")
+        ],
+        "{list:?}"
+    );
+    Ok(())
+}
+
+/// Teammates' drafts of one item share its path by design, so another user's row is no
+/// obstacle — except across the app pair, where the two kinds are different items on one
+/// deployed path: deploying either strands the other, and deleting the app takes both.
+#[sqlx::test(fixtures("base", "drafts_move_taken"))]
+async fn test_draft_move_refuses_another_users_other_app_kind(
+    db: Pool<Postgres>,
+) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+    let move_to = |kind: &'static str, from: &'static str, to: &'static str| async move {
+        let resp = reqwest::Client::new()
+            .post(format!(
+                "http://localhost:{port}/api/w/test-workspace/drafts/move/{kind}/{from}"
+            ))
+            .header("Authorization", "Bearer SECRET_TOKEN")
+            .json(&json!({ "new_path": to }))
+            .send()
+            .await?;
+        Ok::<_, anyhow::Error>((resp.status(), resp.text().await?))
+    };
+
+    let (status, body) = move_to(
+        "app",
+        "u/test-user/mvtaken_app",
+        "u/test-user/mvtaken_theirs",
+    )
+    .await?;
+    assert_eq!(
+        status, 400,
+        "a classic app was moved onto another user's raw app: {body}"
+    );
+    assert!(
+        body.contains("holds another user's raw app draft"),
+        "the refusal did not name the occupant: {body}"
+    );
+
+    // And the other direction, where the occupant reads as the classic kind.
+    let (status, body) = move_to(
+        "raw_app",
+        "u/test-user/mvtaken_raw",
+        "u/test-user/mvtaken_app_theirs",
+    )
+    .await?;
+    assert_eq!(
+        status, 400,
+        "a raw app was moved onto another user's classic app: {body}"
+    );
+    assert!(
+        body.contains("holds another user's app draft"),
+        "the refusal did not name the occupant: {body}"
+    );
+
+    // The same-kind case is the ordinary one: two users' drafts of one raw app.
+    let (status, body) = move_to(
+        "raw_app",
+        "u/test-user/mvtaken_raw",
+        "u/test-user/mvtaken_theirs",
+    )
+    .await?;
+    assert!(
+        status.is_success(),
+        "a raw app was refused beside another user's raw-app draft: {body}"
+    );
+    Ok(())
+}
@@ -78,3 +78,46 @@ async fn test_draft_write_strips_nul(db: Pool<Postgres>) -> anyhow::Result<()> {

    Ok(())
 }
+
+/// The lineage the server derives for `draft.base` comes from the same value: read before
+/// the sanitizer, a NUL in it costs the draft its base, which then reads as up to date
+/// whatever the head is.
+#[sqlx::test(fixtures("drafts_nul"))]
+async fn test_draft_base_is_derived_from_sanitized_value(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+    let base = format!("http://localhost:{port}/api/w/dnul-ws");
+
+    let resp = authed(client().post(format!(
+        "{base}/drafts/update/script/u/dnul-admin/poisoned_base"
+    )))
+    .json(&json!({
+        "value": {
+            "summary": "s",
+            "path": "u/dnul-admin/poisoned_base",
+            "content": "x",
+            "parent_hash": "0000000000001b62\u{0}"
+        }
+    }))
+    .send()
+    .await?;
+    assert_eq!(
+        resp.status(),
+        200,
+        "save refused: {}",
+        resp.text().await.unwrap_or_default()
+    );
+
+    let stored: Option<String> = sqlx::query_scalar(
+        "SELECT base FROM draft WHERE workspace_id = 'dnul-ws' AND path = 'u/dnul-admin/poisoned_base'",
+    )
+    .fetch_one(&db)
+    .await?;
+    assert_eq!(
+        stored.as_deref(),
+        Some("0000000000001b62"),
+        "the NUL cost the draft its base: {stored:?}"
+    );
+    Ok(())
+}
@@ -0,0 +1,587 @@
+//! A draft save addressed to a path its item moved away from lands on the moved draft.
+//!
+//! A move carries every draft on the item to the new path and records where they
+//! went. An editor left open across it still saves to the old path; the server puts
+//! the save on the moved draft, keeps the path keys the move gave it, and names the
+//! new path so the editor can follow. Without the record the save would plant a
+//! phantom draft-only item at the old location.
+
+use serde_json::{json, Value};
+use sqlx::{Pool, Postgres};
+
+use windmill_test_utils::*;
+
+/// Hex form of script hash 7030, the way the API takes a parent hash.
+const HEAD_HASH: &str = "0000000000001b76";
+
+async fn own_draft_value(port: u16, path: &str) -> anyhow::Result<Value> {
+    let draft: Value = reqwest::Client::new()
+        .get(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/get_own/script/{path}"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .send()
+        .await?
+        .json()
+        .await?;
+    Ok(draft["value"].clone())
+}
+
+async fn own_draft_paths(port: u16) -> anyhow::Result<Vec<String>> {
+    let list: Vec<Value> = reqwest::Client::new()
+        .get(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/list"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .send()
+        .await?
+        .json()
+        .await?;
+    Ok(list
+        .iter()
+        .filter(|d| d["kind"] == "script")
+        .filter_map(|d| d["path"].as_str().map(String::from))
+        .filter(|p| p.starts_with("u/test-user/follow_"))
+        .collect())
+}
+
+#[sqlx::test(fixtures("base", "drafts_save_follows_move"))]
+async fn test_save_follows_a_rename(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+    let client = reqwest::Client::new();
+
+    // Rename from Home: a redeploy of the deployed content at the new path that
+    // keeps the deployer's own draft, so the draft is carried rather than consumed.
+    let resp = client
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/scripts/create"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({
+            "path": "u/test-user/follow_b",
+            "parent_hash": HEAD_HASH,
+            "summary": "A",
+            "description": "",
+            "content": "export function main() { return 1 }",
+            "language": "deno",
+            "schema": {},
+            "skip_draft_deletion": true
+        }))
+        .send()
+        .await?;
+    assert_eq!(resp.status(), 201, "rename failed: {}", resp.text().await?);
+    assert_eq!(own_draft_paths(port).await?, vec!["u/test-user/follow_b"]);
+    // A script draft's `path` is where deploying it lands, so it moves with the row.
+    assert_eq!(
+        own_draft_value(port, "u/test-user/follow_b").await?["path"],
+        "u/test-user/follow_b"
+    );
+
+    // The editor is still on the old path and writes that path back into the value.
+    let saved: Value = client
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/update/script/u/test-user/follow_a"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({
+            "value": {
+                "path": "u/test-user/follow_a",
+                "parent_hash": HEAD_HASH,
+                "summary": "A",
+                "content": "edited after the move"
+            }
+        }))
+        .send()
+        .await?
+        .json()
+        .await?;
+    assert_eq!(saved["status"], "saved", "save refused: {saved}");
+    assert_eq!(
+        saved["path"], "u/test-user/follow_b",
+        "save did not follow the row: {saved}"
+    );
+
+    // The write landed on the carried row; nothing reappeared at the old path.
+    assert_eq!(own_draft_paths(port).await?, vec!["u/test-user/follow_b"]);
+    let draft = own_draft_value(port, "u/test-user/follow_b").await?;
+    assert_eq!(draft["content"], "edited after the move", "{draft}");
+    assert_eq!(draft["path"], "u/test-user/follow_b", "{draft}");
+
+    Ok(())
+}
+
+/// A draft-only move rewrites both path keys. The owner's open editor still carries
+/// the typed path it had, which names neither the old nor the new path; the moved
+/// draft's own keys have to win, or the save walks the item back.
+#[sqlx::test(fixtures("base", "drafts_save_follows_move"))]
+async fn test_save_follows_a_draft_only_move(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+    let client = reqwest::Client::new();
+
+    let resp = client
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/move/script/u/test-user/draft_store"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({ "new_path": "u/test-user/moved" }))
+        .send()
+        .await?;
+    assert!(
+        resp.status().is_success(),
+        "move failed: {}",
+        resp.text().await?
+    );
+
+    let saved: Value = client
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/update/script/u/test-user/draft_store"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({
+            "value": {
+                "path": "u/test-user/friendly",
+                "draft_path": "u/test-user/friendly",
+                "summary": "D",
+                "content": "edited after the move"
+            }
+        }))
+        .send()
+        .await?
+        .json()
+        .await?;
+    assert_eq!(saved["path"], "u/test-user/moved", "{saved}");
+
+    let draft = own_draft_value(port, "u/test-user/moved").await?;
+    assert_eq!(draft["content"], "edited after the move", "{draft}");
+    assert_eq!(draft["path"], "u/test-user/moved", "{draft}");
+    assert_eq!(draft["draft_path"], "u/test-user/moved", "{draft}");
+    Ok(())
+}
+
+/// Rename `from` to `to` the way Home does: redeploy the deployed content at the new
+/// path, keeping the deployer's own draft so it is carried rather than consumed.
+/// Returns the new head's hash.
+async fn rename(port: u16, from_hash: &str, to: &str) -> anyhow::Result<String> {
+    let resp = reqwest::Client::new()
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/scripts/create"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({
+            "path": to,
+            "parent_hash": from_hash,
+            "summary": "A",
+            "description": "",
+            "content": "export function main() { return 1 }",
+            "language": "deno",
+            "schema": {},
+            "skip_draft_deletion": true
+        }))
+        .send()
+        .await?;
+    let status = resp.status();
+    let hash = resp.text().await?;
+    assert_eq!(status, 201, "rename to {to} failed: {hash}");
+    Ok(hash)
+}
+
+/// Save the draft as an editor still bound to `url_path` would. Returns the path the
+/// save landed at.
+async fn save_at(port: u16, url_path: &str, content: &str) -> anyhow::Result<String> {
+    let saved: Value = reqwest::Client::new()
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/update/script/{url_path}"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({
+            "value": { "path": url_path, "summary": "A", "content": content, "language": "deno" }
+        }))
+        .send()
+        .await?
+        .json()
+        .await?;
+    assert_eq!(saved["status"], "saved", "save refused: {saved}");
+    Ok(saved["path"].as_str().unwrap_or_default().to_string())
+}
+
+/// A record is kept to one hop, and a move back to the path it left ends it: both are
+/// three statements whose order decides the answer.
+#[sqlx::test(fixtures("base", "drafts_save_follows_move"))]
+async fn test_move_records_stay_one_hop(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+
+    let b = rename(port, HEAD_HASH, "u/test-user/follow_b").await?;
+    let _c = rename(port, &b, "u/test-user/follow_c").await?;
+    assert_eq!(
+        save_at(port, "u/test-user/follow_a", "after two moves").await?,
+        "u/test-user/follow_c",
+        "a save at the first path did not reach the last"
+    );
+    assert_eq!(own_draft_paths(port).await?, vec!["u/test-user/follow_c"]);
+    Ok(())
+}
+
+#[sqlx::test(fixtures("base", "drafts_save_follows_move"))]
+async fn test_move_back_ends_the_record(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+
+    let b = rename(port, HEAD_HASH, "u/test-user/follow_b").await?;
+    let _a = rename(port, &b, "u/test-user/follow_a").await?;
+    assert_eq!(
+        save_at(port, "u/test-user/follow_a", "after moving back").await?,
+        "u/test-user/follow_a",
+        "a save was routed off the path the item moved back to"
+    );
+    assert_eq!(own_draft_paths(port).await?, vec!["u/test-user/follow_a"]);
+    Ok(())
+}
+
+/// A rename carries a teammate's row too: both its path keys follow, and the version
+/// it forked from does not move. A restamp there would clear their out-of-date prompt
+/// and let them deploy over the mover's version believing they were current.
+#[sqlx::test(fixtures("base", "drafts_save_follows_move"))]
+async fn test_a_teammates_draft_follows_with_its_base(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+
+    rename(port, HEAD_HASH, "u/test-user/follow_b").await?;
+
+    // Read from the pool: the teammate's row is another user's, and this asserts on
+    // `base`, which no endpoint exposes for someone else's draft.
+    let row: (String, String, Option<String>) = sqlx::query_as(
+        "SELECT value::jsonb ->> 'path', value::jsonb ->> 'draft_path', base
+         FROM draft WHERE workspace_id = 'test-workspace' AND typ = 'script'
+           AND email = 'test2@windmill.dev'",
+    )
+    .fetch_one(&db)
+    .await?;
+    assert_eq!(row.0, "u/test-user/follow_b", "typed path did not follow");
+    assert_eq!(row.1, "u/test-user/follow_b", "mirror did not follow");
+    assert_eq!(
+        row.2.as_deref(),
+        Some(HEAD_HASH),
+        "the teammate's base was restamped by someone else's rename"
+    );
+    Ok(())
+}
+
+/// An item move and then the owner's own move of what is left: the two records have
+/// different scopes, so the owner's move has to extend the chain in its own scope or
+/// a save addressed to the first path stops at the abandoned middle one.
+#[sqlx::test(fixtures("base", "drafts_save_follows_move"))]
+async fn test_an_owner_move_extends_an_item_move(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+
+    rename(port, HEAD_HASH, "u/test-user/follow_b").await?;
+    // Archiving the script at the new path leaves the carried draft as a draft-only
+    // item, which its owner can move through `/drafts/move`.
+    sqlx::query("UPDATE script SET archived = true WHERE path = 'u/test-user/follow_b'")
+        .execute(&db)
+        .await?;
+    let resp = reqwest::Client::new()
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/move/script/u/test-user/follow_b"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({ "new_path": "u/test-user/follow_c" }))
+        .send()
+        .await?;
+    assert!(
+        resp.status().is_success(),
+        "move failed: {}",
+        resp.text().await?
+    );
+
+    assert_eq!(
+        save_at(port, "u/test-user/follow_a", "after both moves").await?,
+        "u/test-user/follow_c",
+        "a save at the first path stopped at the path the owner's move left"
+    );
+    assert_eq!(own_draft_paths(port).await?, vec!["u/test-user/follow_c"]);
+    Ok(())
+}
+
+/// Redeploying at a path an owner's move routed away from ends that route: the live item
+/// owns its path again, and its saves must not follow the draft that left.
+#[sqlx::test(fixtures("base", "drafts_save_follows_move"))]
+async fn test_redeploy_at_a_routed_path_ends_the_route(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+
+    // `move_draft` ignores archived rows, so an archived script's draft can be moved away.
+    sqlx::query("UPDATE script SET archived = true WHERE path = 'u/test-user/follow_a'")
+        .execute(&db)
+        .await?;
+    let resp = reqwest::Client::new()
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/move/script/u/test-user/follow_a"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({ "new_path": "u/test-user/follow_b" }))
+        .send()
+        .await?;
+    assert!(
+        resp.status().is_success(),
+        "move failed: {}",
+        resp.text().await?
+    );
+
+    // Unarchiving redeploys at the same path, with the archived version as parent.
+    rename(port, HEAD_HASH, "u/test-user/follow_a").await?;
+
+    assert_eq!(
+        save_at(port, "u/test-user/follow_a", "for the live script").await?,
+        "u/test-user/follow_a",
+        "a save for the redeployed script followed the draft that moved away"
+    );
+    Ok(())
+}
+
+/// A draft written before the NUL sanitizer still has to follow a move: its path keys are
+/// what a deploy of it would land on, so the carry rewrites them, sanitizing the value it
+/// could not otherwise parse.
+#[sqlx::test(fixtures("base", "drafts_save_follows_move"))]
+async fn test_a_poisoned_draft_follows_a_rename(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+
+    // The teammate's row, rewritten the way a pre-sanitizer client left one: a real NUL
+    // escape in the content, both path keys naming the path the item is about to leave.
+    sqlx::query(
+        r#"UPDATE draft SET value = '{"path": "u/test-user/follow_a", "draft_path": "u/test-user/follow_a",
+             "parent_hash": "0000000000001b76", "summary": "A", "content": "a\u0000b"}'
+           WHERE email = 'test2@windmill.dev'"#,
+    )
+    .execute(&db)
+    .await?;
+
+    rename(port, HEAD_HASH, "u/test-user/follow_b").await?;
+
+    let row: (String, String, String) = sqlx::query_as(
+        "SELECT value::jsonb ->> 'path', value::jsonb ->> 'draft_path', value::jsonb ->> 'content'
+         FROM draft WHERE email = 'test2@windmill.dev'",
+    )
+    .fetch_one(&db)
+    .await?;
+    assert_eq!(row.0, "u/test-user/follow_b", "typed path did not follow");
+    assert_eq!(row.1, "u/test-user/follow_b", "mirror did not follow");
+    assert_eq!(row.2, "ab", "the NUL survived the rewrite");
+    Ok(())
+}
+
+/// A discard whose answer is lost is retried, and the retry finds nothing to delete. It
+/// still has to name where the item went: the editor asking is on the path it left, and a
+/// reload there lands on nothing.
+#[sqlx::test(fixtures("base", "drafts_save_follows_move"))]
+async fn test_a_retried_discard_still_names_the_destination(
+    db: Pool<Postgres>,
+) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+    let client = reqwest::Client::new();
+
+    rename(port, HEAD_HASH, "u/test-user/follow_b").await?;
+
+    let discard = |token: &'static str| {
+        let client = client.clone();
+        async move {
+            client
+                .post(format!(
+                    "http://localhost:{port}/api/w/test-workspace/drafts/update/script/u/test-user/follow_a"
+                ))
+                .header("Authorization", format!("Bearer {token}"))
+                .json(&json!({ "value": null }))
+                .send()
+                .await?
+                .json::<Value>()
+                .await
+        }
+    };
+
+    let first = discard("SECRET_TOKEN").await?;
+    assert_eq!(first["status"], "saved", "{first}");
+    assert_eq!(first["path"], "u/test-user/follow_b", "{first}");
+
+    let retry = discard("SECRET_TOKEN").await?;
+    assert_eq!(
+        retry["status"], "saved",
+        "the retry was not idempotent: {retry}"
+    );
+    assert_eq!(retry["path"], "u/test-user/follow_b", "{retry}");
+
+    // A third user has no draft on this item, so their discard deletes nothing and the
+    // destination is a move record and nothing else: it is answered only to a caller who
+    // can read that path.
+    let other = discard("SECRET_TOKEN_3").await?;
+    assert_eq!(other["status"], "saved", "{other}");
+    assert_eq!(
+        other["path"],
+        Value::Null,
+        "a user who cannot read the destination was told where the item went: {other}"
+    );
+    Ok(())
+}
+
+/// The legacy workspace-level row is carried by a rename like any other draft, and the
+/// record that routes saves to it covers every caller — so discarding it from a page that
+/// still names the old path has to reach it where it went.
+#[sqlx::test(fixtures("base", "drafts_save_follows_move"))]
+async fn test_a_legacy_discard_follows_a_rename(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+
+    sqlx::query(
+        r#"INSERT INTO draft (workspace_id, path, typ, value, email)
+           VALUES ('test-workspace', 'u/test-user/follow_a', 'script',
+                   '{"path": "u/test-user/follow_a", "summary": "legacy", "content": "x"}', NULL)"#,
+    )
+    .execute(&db)
+    .await?;
+
+    rename(port, HEAD_HASH, "u/test-user/follow_b").await?;
+
+    let resp = reqwest::Client::new()
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/update/script/u/test-user/follow_a"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({ "value": null, "legacy": true }))
+        .send()
+        .await?;
+    assert!(
+        resp.status().is_success(),
+        "discard failed: {}",
+        resp.text().await?
+    );
+
+    let left: i64 = sqlx::query_scalar(
+        "SELECT count(*) FROM draft WHERE workspace_id = 'test-workspace' AND email IS NULL",
+    )
+    .fetch_one(&db)
+    .await?;
+    assert_eq!(
+        left, 0,
+        "the legacy draft survived a discard aimed at its old path"
+    );
+    Ok(())
+}
+
+/// Deploying a draft that a move carried off an archived script: its parent is still the
+/// version at the old path, so the deploy renames from there and carries what is left over
+/// — onto the very draft being deployed. That row is this deploy's own, not an item in its
+/// way, or the deploy is refused and every retry refuses again.
+#[sqlx::test(fixtures("base", "drafts_save_follows_move"))]
+async fn test_a_moved_draft_deploys_at_its_new_path(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+
+    sqlx::query("UPDATE script SET archived = true WHERE path = 'u/test-user/follow_a'")
+        .execute(&db)
+        .await?;
+    let resp = reqwest::Client::new()
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/move/script/u/test-user/follow_a"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({ "new_path": "u/test-user/follow_b" }))
+        .send()
+        .await?;
+    assert!(
+        resp.status().is_success(),
+        "move failed: {}",
+        resp.text().await?
+    );
+
+    // Deploy it where it now lives, still parented on the archived version it forked from.
+    let resp = reqwest::Client::new()
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/scripts/create"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({
+            "path": "u/test-user/follow_b",
+            "parent_hash": HEAD_HASH,
+            "summary": "A",
+            "description": "",
+            "content": "export function main() { return 2 }",
+            "language": "deno",
+            "schema": {}
+        }))
+        .send()
+        .await?;
+    let status = resp.status();
+    let body = resp.text().await?;
+    assert_eq!(status, 201, "the moved draft could not be deployed: {body}");
+    Ok(())
+}
+
+/// A route is only as good as the item it points at: when an unrelated item claims the
+/// destination, a save still addressed to the old path must stay where it is rather than
+/// land on that item's draft.
+#[sqlx::test(fixtures("base", "drafts_save_follows_move"))]
+async fn test_a_reused_destination_ends_the_route(db: Pool<Postgres>) -> anyhow::Result<()> {
+    initialize_tracing().await;
+    let server = ApiServer::start(db.clone()).await?;
+    let port = server.addr.port();
+
+    sqlx::query("UPDATE script SET archived = true WHERE path = 'u/test-user/follow_a'")
+        .execute(&db)
+        .await?;
+    let resp = reqwest::Client::new()
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/drafts/move/script/u/test-user/follow_a"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({ "new_path": "u/test-user/follow_b" }))
+        .send()
+        .await?;
+    assert!(
+        resp.status().is_success(),
+        "move failed: {}",
+        resp.text().await?
+    );
+    // Someone else's item takes the destination, and the moved draft goes with the
+    // deploy that consumes it.
+    sqlx::query("DELETE FROM draft WHERE workspace_id = 'test-workspace' AND path = 'u/test-user/follow_b'")
+        .execute(&db)
+        .await?;
+    let resp = reqwest::Client::new()
+        .post(format!(
+            "http://localhost:{port}/api/w/test-workspace/scripts/create"
+        ))
+        .header("Authorization", "Bearer SECRET_TOKEN")
+        .json(&json!({
+            "path": "u/test-user/follow_b",
+            "summary": "unrelated",
+            "description": "",
+            "content": "export function main() { return 3 }",
+            "language": "deno",
+            "schema": {}
+        }))
+        .send()
+        .await?;
+    assert_eq!(resp.status(), 201, "create failed: {}", resp.text().await?);
+
+    assert_eq!(
+        save_at(port, "u/test-user/follow_a", "after the destination was reused").await?,
+        "u/test-user/follow_a",
+        "a save was routed onto the item that now owns the destination"
+    );
+    Ok(())
+}
@@ -0,0 +1,16 @@
+-- Fixture for the two-path-key move test.
+--
+-- Two draft-only scripts owned by test-user, both parked at a generated storage
+-- key the way a session-created draft is. One carries the `draft_path` mirror the
+-- session editor writes while the typed path differs from that key; the other
+-- carries no mirror at all, which is what a plain draft looks like.
+
+INSERT INTO draft (workspace_id, path, typ, value, email) VALUES
+    ('test-workspace', 'u/test-user/draft_mirror', 'script',
+     '{"path": "u/test-user/friendly", "draft_path": "u/test-user/friendly",
+       "content": "x", "language": "bun", "summary": "S"}',
+     'test@windmill.dev'),
+    ('test-workspace', 'u/test-user/draft_plain', 'script',
+     '{"path": "u/test-user/draft_plain",
+       "content": "x", "language": "bun", "summary": "S"}',
+     'test@windmill.dev');
--- a/Show More
+++ b/Show More
				`@@ -0,0 +1 @@`
				`ALTER TABLE flow_conversation DROP COLUMN is_test;`
				`@@ -0,0 +1 @@`
				`ALTER TABLE flow_conversation_message DROP COLUMN attachments;`
				`@@ -0,0 +1 @@`
				`ALTER TABLE workspace_settings DROP COLUMN add_admins_and_developers_to_forks;`
				`@@ -0,0 +1 @@`
				`ALTER TABLE workspace_settings ADD COLUMN add_admins_and_developers_to_forks BOOLEAN NOT NULL DEFAULT false;`