@vosjs/cli 0.32.0 → 0.34.0

package/dist/cli.js

@@ -495,7 +495,7 @@ async function cmdCheck(argv) {
   return result.ok ? EXIT_OK : EXIT_ERROR;
 }
 async function delegate(argv, viaAlias = false) {
-  const { run } = await import("./run-PYK3XJNH.js");
+  const { run } = await import("./run-EOIIBITT.js");
   if (viaAlias && argv[0]) {
     process.stderr.write(
       `note: "vos voila ${argv[0]}" is now "vos ${argv[0]}".

package/dist/index.d.ts

@@ -89,6 +89,22 @@ interface PluginManifest {
 }
 declare const manifest: PluginManifest;
 
+interface PaceReport {
+    askedMs: number;
+    gestureMs: number;
+    wallMs: number;
+    /** Wall time neither asked nor a gesture: what the page and the round trips cost. */
+    overheadMs: number;
+    overheadPct: number;
+    /** The steps whose wall ran past their ask plus gesture by more than a third. */
+    slow: {
+        step: number;
+        do: string;
+        askedMs: number;
+        wallMs: number;
+    }[];
+}
+
 /**
  * The action script — the declarative recipe an agent (or human) writes to
  * drive a take. Small on purpose: selectors + a handful of verbs. The recorder
@@ -112,9 +128,12 @@ type ActionStep = ({
     do: 'hover';
     selector: string;
     ms?: number;
-} | {
+}
+/** `ms` is the settle after the press (default 150). */
+ | {
     do: 'click';
     selector: string;
+    ms?: number;
 }
 /**
  * Type into `selector`. The recorder clicks the field first, which is what
@@ -129,9 +148,14 @@ type ActionStep = ({
     text: string;
     delayMs?: number;
     focus?: boolean;
-} | {
+    /** the settle after the last character (default 150) */
+    ms?: number;
+}
+/** `ms` is the settle after the scroll lands (default 200). */
+ | {
     do: 'scroll';
     dy: number;
+    ms?: number;
 } | {
     do: 'move';
     x: number;
@@ -205,6 +229,8 @@ interface RecordResult {
     events: CursorEvent[];
     frames: FrameRec[];
     meta: RecordingMeta;
+    /** The take's pace: what the script asked, what the gestures added, what the page cost. */
+    pace: PaceReport;
     /** steps whose selector never became visible — the take continued without them. */
     skipped: SkippedStep[];
     /** the initial goto never reached networkidle (recording proceeded anyway). */

package/dist/index.js

@@ -30,7 +30,7 @@ import {
   validateActions,
   waitForPageDone,
   writeSyncState
-} from "./chunk-QJORXTJM.js";
+} from "./chunk-43FCXWNS.js";
 import {
   BrowserUnavailableError,
   configDuration,

package/dist/{run-PYK3XJNH.js → run-EOIIBITT.js}

@@ -1,9 +1,9 @@
 #!/usr/bin/env node
 import {
   run
-} from "./chunk-QJORXTJM.js";
+} from "./chunk-43FCXWNS.js";
 import "./chunk-NHKHTIDR.js";
 export {
   run
 };
-//# sourceMappingURL=run-PYK3XJNH.js.map
+//# sourceMappingURL=run-EOIIBITT.js.map

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vosjs/cli",
-  "version": "0.32.0",
+  "version": "0.34.0",
   "description": "The vos CLI: record the real product from a scripted browser flow, auto-zoom from the cursor track, cut as data in doc.json, render deterministic video and stills, deliver a release's media per destination spec, and sync with vos.so. One binary, every verb, MIT.",
   "type": "module",
   "license": "MIT",
@@ -53,9 +53,9 @@
     "@vosjs/elements": "^0.8.1",
     "@vosjs/render-core": "^0.2.8",
     "@vosjs/shared": "^0.4.1",
-    "@vosjs/studio-core": "^0.24.0",
-    "@vosjs/timeline": "^0.4.1",
-    "@vosjs/tween": "^0.8.2"
+    "@vosjs/studio-core": "^0.25.0",
+    "@vosjs/tween": "^0.8.2",
+    "@vosjs/timeline": "^0.4.1"
   },
   "devDependencies": {
     "@types/node": "^22",

package/schema/actions.schema.json

@@ -4,9 +4,7 @@
   "title": "vosso actions script",
   "description": "Declarative browser flow executed by `vos record` — the recipe an agent writes after exploring a site. The recorder synthesizes the cursor track from these steps.",
   "type": "object",
-  "required": [
-    "steps"
-  ],
+  "required": ["steps"],
   "additionalProperties": false,
   "properties": {
     "url": {
@@ -16,10 +14,7 @@
     },
     "viewport": {
       "type": "object",
-      "required": [
-        "width",
-        "height"
-      ],
+      "required": ["width", "height"],
       "additionalProperties": false,
       "properties": {
         "width": {
@@ -41,10 +36,7 @@
         "oneOf": [
           {
             "type": "object",
-            "required": [
-              "do",
-              "ms"
-            ],
+            "required": ["do", "ms"],
             "additionalProperties": false,
             "properties": {
               "do": {
@@ -68,10 +60,7 @@
           },
           {
             "type": "object",
-            "required": [
-              "do",
-              "selector"
-            ],
+            "required": ["do", "selector"],
             "additionalProperties": false,
             "properties": {
               "do": {
@@ -100,15 +89,17 @@
           },
           {
             "type": "object",
-            "required": [
-              "do",
-              "selector"
-            ],
+            "required": ["do", "selector"],
             "additionalProperties": false,
             "properties": {
               "do": {
                 "const": "click"
               },
+              "ms": {
+                "type": "number",
+                "minimum": 0,
+                "description": "the settle after the press, ms (default 150): how long the recorder holds before the next step. The script's own wait steps carry the pauses that mean something; this is only the hand lifting."
+              },
               "selector": {
                 "type": "string"
               },
@@ -126,17 +117,18 @@
           },
           {
             "type": "object",
-            "required": [
-              "do",
-              "selector",
-              "text"
-            ],
+            "required": ["do", "selector", "text"],
             "additionalProperties": false,
             "description": "Typing is a zoom signal — a type step earns a held frame on its field (the click into it opens the span, the camera releases after the last character). ~0.6s+ of typing qualifies, so pace delayMs like a person (60-90ms) for a camera moment.",
             "properties": {
               "do": {
                 "const": "type"
               },
+              "ms": {
+                "type": "number",
+                "minimum": 0,
+                "description": "the settle after the last character, ms (default 150)"
+              },
               "selector": {
                 "type": "string"
               },
@@ -167,15 +159,17 @@
           },
           {
             "type": "object",
-            "required": [
-              "do",
-              "dy"
-            ],
+            "required": ["do", "dy"],
             "additionalProperties": false,
             "properties": {
               "do": {
                 "const": "scroll"
               },
+              "ms": {
+                "type": "number",
+                "minimum": 0,
+                "description": "the settle after the scroll lands, ms (default 200)"
+              },
               "dy": {
                 "type": "number"
               },
@@ -193,11 +187,7 @@
           },
           {
             "type": "object",
-            "required": [
-              "do",
-              "x",
-              "y"
-            ],
+            "required": ["do", "x", "y"],
             "additionalProperties": false,
             "properties": {
               "do": {
@@ -225,11 +215,7 @@
           },
           {
             "type": "object",
-            "required": [
-              "do",
-              "tx",
-              "ty"
-            ],
+            "required": ["do", "tx", "ty"],
             "additionalProperties": false,
             "description": "Press-move-release: drag an element, slide a range input, move a clip. Start = selector center when given, else (x, y); end = (tx, ty), eased over ms.",
             "properties": {