donobu 2.14.0 → 2.15.0

package/dist/esm/assets/donobu-virtual-mouse.svg

@@ -0,0 +1,11 @@
+<svg xmlns="http://www.w3.org/2000/svg"
+     width="32"
+     height="32"
+     viewBox="0 0 24 24"
+     fill="oklch(13.09% 0.005 165.18)"
+     stroke="oklch(83.04% 0.1698 83.99)"
+     stroke-width="2"
+     stroke-linecap="round"
+     stroke-linejoin="round">
+  <path d="M4.037 4.688a.495.495 0 0 1 .651-.651l16 6.5a.5.5 0 0 1-.063.947l-6.124 1.58a2 2 0 0 0-1.438 1.435l-1.579 6.126a.5.5 0 0 1-.947.063z" />
+</svg>

package/dist/esm/assets/donobu.css

@@ -0,0 +1,62 @@
+.donobu-message {
+  z-index: 2147483647;
+  background-color: black;
+  color: white;
+  padding: 8px;
+  border-radius: 5px;
+  font-size: 12px;
+  font-family:
+    'Source Sans Pro', 'Helvetica Neue', Helvetica, Arial, sans-serif;
+  display: block;
+  max-width: 300px;
+  white-space: pre-wrap;
+  box-shadow: 2px 2px 10px rgba(0, 0, 0, 0.2);
+  position: absolute;
+}
+.donobu-virtual-mouse {
+  width: 24px;
+  height: 24px;
+  background-image: url('/donobu-virtual-mouse.svg');
+  background-repeat: no-repeat;
+  background-size: contain;
+  position: absolute;
+  z-index: 2147483646;
+  transition-property: left, top;
+  transition-timing-function: ease-in-out;
+  filter: drop-shadow(1px 1px 1px rgba(0, 0, 0, 0.5));
+}
+.donobu-virtual-mouse:hover {
+  transform: scale(1.1);
+}
+/* Ripple Effect */
+.donobu-virtual-mouse::after {
+  content: '';
+  position: absolute;
+  left: 50%;
+  top: 50%;
+  width: 24px;
+  height: 24px;
+  background: oklch(83.04% 0.1698 83.99); /* Donobu Yellow */
+  opacity: 0;
+  border-radius: 50%;
+  transform: translate(-50%, -50%) scale(0);
+  filter: blur(4px);
+  pointer-events: none;
+}
+.donobu-virtual-mouse.rippling::after {
+  animation: ripple-pop 0.6s ease-out;
+}
+@keyframes ripple-pop {
+  0% {
+    transform: translate(-50%, -50%) scale(0.5);
+    opacity: 0.3;
+  }
+  40% {
+    transform: translate(-50%, -50%) scale(1);
+    opacity: 0.5;
+  }
+  100% {
+    transform: translate(-50%, -50%) scale(2);
+    opacity: 0;
+  }
+}

package/dist/esm/assets/generated/parameter-schemas.json

@@ -150,32 +150,87 @@
     "additionalProperties": false,
     "required": ["annotation", "rationale", "whyThisAnnotation"]
   },
-  "HandleBrowserDialogToolCoreParameters": {
-    "description": "The parameters for handling browser dialogs",
+  "InputTextToolCoreParameters": {
     "type": "object",
     "properties": {
       "text": {
-        "description": "If this is a \"confirm\" dialog, a string of \"true\" will accept the dialog and a string of \"false\" will dismiss it.\nHowever, if this is a \"prompt\" dialog, any non-null string will accept it, and a null value will dismiss it.",
-        "type": ["string", "null"]
+        "description": "The text to input.",
+        "type": "string"
+      },
+      "finalizeWithSubmit": {
+        "description": "Attempt to submit the data after inputting the text (i.e. hitting 'Enter' at the end).\nThis can be useful if doing something like using a webpage's search box, etc.",
+        "type": "boolean"
       }
     },
     "additionalProperties": false,
-    "required": ["text"]
+    "required": ["finalizeWithSubmit", "text"]
   },
-  "HandleBrowserDialogToolGptParameters": {
+  "SelectorBasedInputTextToolParameters": {
     "type": "object",
     "properties": {
+      "selector": {
+        "description": "A query structure used to resolve specific a HTML element in a webpage.",
+        "type": "object",
+        "properties": {
+          "element": {
+            "description": "An prioritized list of element selectors to use to attempt to resolve the\ndesired element, going from most specific to least specific. The first\nelement selector that is successfully used, wins, and the rest are ignored.",
+            "type": "array",
+            "items": {
+              "type": "string"
+            }
+          },
+          "frame": {
+            "description": "A selector for the corresponding frame of the element to select.\nIf null, this means the primary frame of the page is used.",
+            "type": "string"
+          }
+        },
+        "additionalProperties": false,
+        "required": ["element", "frame"]
+      },
+      "text": {
+        "description": "The text to input.",
+        "type": "string"
+      },
+      "finalizeWithSubmit": {
+        "description": "Attempt to submit the data after inputting the text (i.e. hitting 'Enter' at the end).\nThis can be useful if doing something like using a webpage's search box, etc.",
+        "type": "boolean"
+      }
+    },
+    "additionalProperties": false,
+    "required": ["finalizeWithSubmit", "selector", "text"]
+  },
+  "AnnotationBasedInputTextToolParameters": {
+    "type": "object",
+    "properties": {
+      "whyThisAnnotation": {
+        "description": "The reason why this particular numbered annotation was chosen.",
+        "type": "string"
+      },
+      "annotation": {
+        "description": "The numbered annotation of the element to click.",
+        "type": "string"
+      },
       "rationale": {
         "description": "The reason why this particular action is being taken. It MUST relate to the overall objective.",
         "type": "string"
       },
       "text": {
-        "description": "If this is a \"confirm\" dialog, a string of \"true\" will accept the dialog and a string of \"false\" will dismiss it.\nHowever, if this is a \"prompt\" dialog, any non-null string will accept it, and a null value will dismiss it.",
-        "type": ["string", "null"]
+        "description": "The text to input.",
+        "type": "string"
+      },
+      "finalizeWithSubmit": {
+        "description": "Attempt to submit the data after inputting the text (i.e. hitting 'Enter' at the end).\nThis can be useful if doing something like using a webpage's search box, etc.",
+        "type": "boolean"
       }
     },
     "additionalProperties": false,
-    "required": ["rationale", "text"]
+    "required": [
+      "annotation",
+      "finalizeWithSubmit",
+      "rationale",
+      "text",
+      "whyThisAnnotation"
+    ]
   },
   "InputRandomizedEmailAddressToolCoreParameters": {
     "type": "object",
@@ -259,55 +314,6 @@
       "whyThisAnnotation"
     ]
   },
-  "MarkObjectiveCompleteToolCoreParameters": {
-    "description": "Parameters for marking an objective as complete",
-    "type": "object",
-    "additionalProperties": false,
-    "required": [],
-    "properties": {}
-  },
-  "MarkObjectiveCompleteToolGptParameters": {
-    "description": "Parameters for marking an objective as complete",
-    "type": "object",
-    "properties": {
-      "details": {
-        "description": "Relevant details for the completed objective.",
-        "type": "string"
-      },
-      "rationale": {
-        "description": "The reason why this particular action is being taken. It MUST relate to the overall objective.",
-        "type": "string"
-      }
-    },
-    "additionalProperties": false,
-    "required": ["details", "rationale"]
-  },
-  "GoToWebpageToolCoreParameters": {
-    "type": "object",
-    "properties": {
-      "url": {
-        "description": "The URL to navigate to. Must use the http or https (example is https://google.com).",
-        "type": "string"
-      }
-    },
-    "additionalProperties": false,
-    "required": ["url"]
-  },
-  "GoToWebpageToolGptParameters": {
-    "type": "object",
-    "properties": {
-      "rationale": {
-        "description": "The reason why this particular action is being taken. It MUST relate to the overall objective.",
-        "type": "string"
-      },
-      "url": {
-        "description": "The URL to navigate to. Must use the http or https (example is https://google.com).",
-        "type": "string"
-      }
-    },
-    "additionalProperties": false,
-    "required": ["rationale", "url"]
-  },
   "PressKeyToolCoreParameters": {
     "type": "object",
     "properties": {
@@ -372,158 +378,152 @@
     "additionalProperties": false,
     "required": ["annotation", "key", "rationale", "whyThisAnnotation"]
   },
-  "RunAccessibilityTestToolCoreParameters": {
+  "HandleBrowserDialogToolCoreParameters": {
+    "description": "The parameters for handling browser dialogs",
+    "type": "object",
+    "properties": {
+      "text": {
+        "description": "If this is a \"confirm\" dialog, a string of \"true\" will accept the dialog and a string of \"false\" will dismiss it.\nHowever, if this is a \"prompt\" dialog, any non-null string will accept it, and a null value will dismiss it.",
+        "type": ["string", "null"]
+      }
+    },
+    "additionalProperties": false,
+    "required": ["text"]
+  },
+  "HandleBrowserDialogToolGptParameters": {
+    "type": "object",
+    "properties": {
+      "rationale": {
+        "description": "The reason why this particular action is being taken. It MUST relate to the overall objective.",
+        "type": "string"
+      },
+      "text": {
+        "description": "If this is a \"confirm\" dialog, a string of \"true\" will accept the dialog and a string of \"false\" will dismiss it.\nHowever, if this is a \"prompt\" dialog, any non-null string will accept it, and a null value will dismiss it.",
+        "type": ["string", "null"]
+      }
+    },
+    "additionalProperties": false,
+    "required": ["rationale", "text"]
+  },
+  "MarkObjectiveCompleteToolCoreParameters": {
+    "description": "Parameters for marking an objective as complete",
     "type": "object",
     "additionalProperties": false,
     "required": [],
     "properties": {}
   },
-  "RunAccessibilityTestToolGptParameters": {
+  "MarkObjectiveCompleteToolGptParameters": {
+    "description": "Parameters for marking an objective as complete",
     "type": "object",
     "properties": {
+      "details": {
+        "description": "Relevant details for the completed objective.",
+        "type": "string"
+      },
       "rationale": {
         "description": "The reason why this particular action is being taken. It MUST relate to the overall objective.",
         "type": "string"
       }
     },
     "additionalProperties": false,
-    "required": ["rationale"]
+    "required": ["details", "rationale"]
   },
-  "ScrollPageToolCoreParameters": {
+  "GoToWebpageToolCoreParameters": {
     "type": "object",
     "properties": {
-      "direction": {
-        "description": "The direction to scroll - either UP or DOWN",
-        "enum": ["DOWN", "UP"],
+      "url": {
+        "description": "The URL to navigate to. Must use the http or https (example is https://google.com).",
         "type": "string"
       }
     },
     "additionalProperties": false,
-    "required": ["direction"]
+    "required": ["url"]
   },
-  "ScrollPageToolGptParameters": {
+  "GoToWebpageToolGptParameters": {
     "type": "object",
     "properties": {
       "rationale": {
         "description": "The reason why this particular action is being taken. It MUST relate to the overall objective.",
         "type": "string"
       },
-      "direction": {
-        "description": "The direction to scroll - either UP or DOWN",
-        "enum": ["DOWN", "UP"],
+      "url": {
+        "description": "The URL to navigate to. Must use the http or https (example is https://google.com).",
         "type": "string"
       }
     },
     "additionalProperties": false,
-    "required": ["direction", "rationale"]
+    "required": ["rationale", "url"]
   },
-  "AssertToolCoreParameters": {
+  "RunAccessibilityTestToolCoreParameters": {
+    "type": "object",
+    "additionalProperties": false,
+    "required": [],
+    "properties": {}
+  },
+  "RunAccessibilityTestToolGptParameters": {
     "type": "object",
     "properties": {
-      "assertionToTestFor": {
-        "description": "The assertion that will be tested for, using plain English.\n\nExamples:\n- 'Assert that this webpage includes a product summary.'\n- 'Assert there are more than 3 search results.'\n- 'Assert that the webpage title is \"Hello World\".'\n- 'Assert that the theme of the website is in dark mode.'\n\nThe raw assertion text should avoid embedding state about the webpage\nsince the assertion is being evaluated by the tool itself, and the tool\nwill have access to the webpage's content, title, URL, etc.\n\nExamples of bad assertions:\n- 'Assert that the 3 search results is greater than 2.'\n- 'Assert that the first search result (\"banana\") is not \"strawberry\".'\n\nExamples of the bad assertions being rephrased as good assertions:\n- 'Assert that there are more than 2 search results.'\n- 'Assert that the first search result is not \"strawberry\".'",
+      "rationale": {
+        "description": "The reason why this particular action is being taken. It MUST relate to the overall objective.",
         "type": "string"
       }
     },
     "additionalProperties": false,
-    "required": ["assertionToTestFor"]
+    "required": ["rationale"]
   },
-  "AssertToolGptParameters": {
+  "ScrollPageToolCoreParameters": {
     "type": "object",
     "properties": {
-      "rationale": {
-        "description": "The reason why this particular action is being taken. It MUST relate to the overall objective.",
-        "type": "string"
-      },
-      "assertionToTestFor": {
-        "description": "The assertion that will be tested for, using plain English.\n\nExamples:\n- 'Assert that this webpage includes a product summary.'\n- 'Assert there are more than 3 search results.'\n- 'Assert that the webpage title is \"Hello World\".'\n- 'Assert that the theme of the website is in dark mode.'\n\nThe raw assertion text should avoid embedding state about the webpage\nsince the assertion is being evaluated by the tool itself, and the tool\nwill have access to the webpage's content, title, URL, etc.\n\nExamples of bad assertions:\n- 'Assert that the 3 search results is greater than 2.'\n- 'Assert that the first search result (\"banana\") is not \"strawberry\".'\n\nExamples of the bad assertions being rephrased as good assertions:\n- 'Assert that there are more than 2 search results.'\n- 'Assert that the first search result is not \"strawberry\".'",
+      "direction": {
+        "description": "The direction to scroll - either UP or DOWN",
+        "enum": ["DOWN", "UP"],
         "type": "string"
       }
     },
     "additionalProperties": false,
-    "required": ["assertionToTestFor", "rationale"]
+    "required": ["direction"]
   },
-  "InputTextToolCoreParameters": {
+  "ScrollPageToolGptParameters": {
     "type": "object",
     "properties": {
-      "text": {
-        "description": "The text to input.",
+      "rationale": {
+        "description": "The reason why this particular action is being taken. It MUST relate to the overall objective.",
         "type": "string"
       },
-      "finalizeWithSubmit": {
-        "description": "Attempt to submit the data after inputting the text (i.e. hitting 'Enter' at the end).\nThis can be useful if doing something like using a webpage's search box, etc.",
-        "type": "boolean"
+      "direction": {
+        "description": "The direction to scroll - either UP or DOWN",
+        "enum": ["DOWN", "UP"],
+        "type": "string"
       }
     },
     "additionalProperties": false,
-    "required": ["finalizeWithSubmit", "text"]
+    "required": ["direction", "rationale"]
   },
-  "SelectorBasedInputTextToolParameters": {
+  "AssertToolCoreParameters": {
     "type": "object",
     "properties": {
-      "selector": {
-        "description": "A query structure used to resolve specific a HTML element in a webpage.",
-        "type": "object",
-        "properties": {
-          "element": {
-            "description": "An prioritized list of element selectors to use to attempt to resolve the\ndesired element, going from most specific to least specific. The first\nelement selector that is successfully used, wins, and the rest are ignored.",
-            "type": "array",
-            "items": {
-              "type": "string"
-            }
-          },
-          "frame": {
-            "description": "A selector for the corresponding frame of the element to select.\nIf null, this means the primary frame of the page is used.",
-            "type": "string"
-          }
-        },
-        "additionalProperties": false,
-        "required": ["element", "frame"]
-      },
-      "text": {
-        "description": "The text to input.",
+      "assertionToTestFor": {
+        "description": "The assertion that will be tested for, using plain English.\n\nExamples:\n- 'Assert that this webpage includes a product summary.'\n- 'Assert there are more than 3 search results.'\n- 'Assert that the webpage title is \"Hello World\".'\n- 'Assert that the theme of the website is in dark mode.'\n\nThe raw assertion text should avoid embedding state about the webpage\nsince the assertion is being evaluated by the tool itself, and the tool\nwill have access to the webpage's content, title, URL, etc.\n\nExamples of bad assertions:\n- 'Assert that the 3 search results is greater than 2.'\n- 'Assert that the first search result (\"banana\") is not \"strawberry\".'\n\nExamples of the bad assertions being rephrased as good assertions:\n- 'Assert that there are more than 2 search results.'\n- 'Assert that the first search result is not \"strawberry\".'",
         "type": "string"
-      },
-      "finalizeWithSubmit": {
-        "description": "Attempt to submit the data after inputting the text (i.e. hitting 'Enter' at the end).\nThis can be useful if doing something like using a webpage's search box, etc.",
-        "type": "boolean"
       }
     },
     "additionalProperties": false,
-    "required": ["finalizeWithSubmit", "selector", "text"]
+    "required": ["assertionToTestFor"]
   },
-  "AnnotationBasedInputTextToolParameters": {
+  "AssertToolGptParameters": {
     "type": "object",
     "properties": {
-      "whyThisAnnotation": {
-        "description": "The reason why this particular numbered annotation was chosen.",
-        "type": "string"
-      },
-      "annotation": {
-        "description": "The numbered annotation of the element to click.",
-        "type": "string"
-      },
       "rationale": {
         "description": "The reason why this particular action is being taken. It MUST relate to the overall objective.",
         "type": "string"
       },
-      "text": {
-        "description": "The text to input.",
+      "assertionToTestFor": {
+        "description": "The assertion that will be tested for, using plain English.\n\nExamples:\n- 'Assert that this webpage includes a product summary.'\n- 'Assert there are more than 3 search results.'\n- 'Assert that the webpage title is \"Hello World\".'\n- 'Assert that the theme of the website is in dark mode.'\n\nThe raw assertion text should avoid embedding state about the webpage\nsince the assertion is being evaluated by the tool itself, and the tool\nwill have access to the webpage's content, title, URL, etc.\n\nExamples of bad assertions:\n- 'Assert that the 3 search results is greater than 2.'\n- 'Assert that the first search result (\"banana\") is not \"strawberry\".'\n\nExamples of the bad assertions being rephrased as good assertions:\n- 'Assert that there are more than 2 search results.'\n- 'Assert that the first search result is not \"strawberry\".'",
         "type": "string"
-      },
-      "finalizeWithSubmit": {
-        "description": "Attempt to submit the data after inputting the text (i.e. hitting 'Enter' at the end).\nThis can be useful if doing something like using a webpage's search box, etc.",
-        "type": "boolean"
       }
     },
     "additionalProperties": false,
-    "required": [
-      "annotation",
-      "finalizeWithSubmit",
-      "rationale",
-      "text",
-      "whyThisAnnotation"
-    ]
+    "required": ["assertionToTestFor", "rationale"]
   },
   "AnalyzePageTextToolCoreParameters": {
     "type": "object",
@@ -1962,107 +1962,6 @@
     "additionalProperties": false,
     "required": ["rationale", "text"]
   },
-  "ToolParameters": {
-    "anyOf": [
-      {
-        "type": "object",
-        "properties": {
-          "_type": {},
-          "_output": {},
-          "_input": {},
-          "_def": {},
-          "description": {
-            "type": "string"
-          },
-          "~standard": {
-            "description": "The Standard Schema properties interface.",
-            "type": "object",
-            "properties": {
-              "version": {
-                "description": "The version number of the standard.",
-                "type": "number",
-                "const": 1
-              },
-              "vendor": {
-                "description": "The vendor name of the schema library.",
-                "type": "string"
-              },
-              "validate": {
-                "description": "Validates unknown input values.",
-                "type": "object",
-                "additionalProperties": false
-              },
-              "types": {
-                "description": "Inferred types associated with the schema.",
-                "type": "object",
-                "properties": {
-                  "input": {
-                    "description": "The input type of the schema."
-                  },
-                  "output": {
-                    "description": "The output type of the schema."
-                  }
-                },
-                "additionalProperties": false,
-                "required": ["input", "output"]
-              }
-            },
-            "additionalProperties": false,
-            "required": ["validate", "vendor", "version"]
-          },
-          "spa": {
-            "description": "Alias of safeParseAsync",
-            "type": "object",
-            "additionalProperties": false
-          }
-        },
-        "additionalProperties": false,
-        "required": [
-          "_def",
-          "_input",
-          "_output",
-          "_type",
-          "description",
-          "spa",
-          "~standard"
-        ]
-      },
-      {
-        "additionalProperties": false,
-        "type": "object",
-        "properties": {
-          "validate": {
-            "description": "Optional. Validates that the structure of a value matches this schema,\nand returns a typed version of the value if it does.",
-            "type": "object",
-            "additionalProperties": false
-          },
-          "__@validatorSymbol@19388": {
-            "description": "Used to mark validator functions so we can support both Zod and custom schemas.",
-            "type": "boolean",
-            "const": true
-          },
-          "_type": {
-            "description": "Schema type for inference."
-          },
-          "jsonSchema": {
-            "$ref": "#/definitions/JSONSchema7",
-            "description": "The JSON Schema for the schema. It is passed to the providers."
-          },
-          "__@schemaSymbol@19393": {
-            "description": "Used to mark schemas so we can support both Zod and custom schemas.",
-            "type": "boolean",
-            "const": true
-          }
-        },
-        "required": [
-          "__@schemaSymbol@19393",
-          "__@validatorSymbol@19388",
-          "_type",
-          "jsonSchema"
-        ]
-      }
-    ]
-  },
   "AggregateExtractedStreetviewDataToolCoreParameters": {
     "type": "object",
     "additionalProperties": false,

package/dist/esm/assets/generated/version

@@ -1 +1 @@
-1087
+1107

package/dist/esm/assets/smart-selector-generator.js

@@ -32,7 +32,7 @@
       return `'${value}'`;
     }
 
-    // If the string has single quotes but no double quotes, wrap in double quotes
+    // If the string has no double quotes, wrap in double quotes
     if (!value.includes('"')) {
       return `"${value}"`;
     }
@@ -91,9 +91,24 @@
         selectors.add(selector),
       );
 
+      // Generate DOM position-based selectors
+      this.getDOMPositionSelectors().forEach((selector) =>
+        selectors.add(selector),
+      );
+
       // Combine selectors for robustness
       this.getCombinedSelector().forEach((selector) => selectors.add(selector));
 
+      function selectorTiebreakerPriority(sel) {
+        // highest-to-lowest: aria-label, placeholder, text-based XPath, DOM position, id, other
+        if (/\[aria-label\s*=/.test(sel) || /@aria-label/.test(sel)) return 0;
+        if (/\[placeholder\s*=/.test(sel) || /@placeholder/.test(sel)) return 1;
+        if (sel.startsWith('//')) return 2;
+        if (sel.includes(' > ')) return 3; // DOM position selector
+        if (/^#/.test(sel)) return 4;
+        return 5;
+      }
+
       const rankedSelectors = Array.from(selectors)
         .map((selector) => {
           try {
@@ -104,9 +119,18 @@
           }
         })
         .filter((a) => a !== null && a.count !== 0)
-        .sort((a, b) => a.count - b.count);
+        .sort((a, b) => {
+          // 1. fewer matches -> better
+          if (a.count !== b.count) return a.count - b.count;
+          // 2. tie-break on priority class
+          return (
+            selectorTiebreakerPriority(a.selector) -
+            selectorTiebreakerPriority(b.selector)
+          );
+        })
+        .map((a) => a.selector);
 
-      return rankedSelectors.map((a) => a.selector);
+      return rankedSelectors;
     }
 
     getIdSelector() {
@@ -243,6 +267,40 @@
       return [];
     }
 
+    getDOMPositionSelectors() {
+      const selectors = [];
+
+      // Generate full DOM position-based selector without depth limit
+      let currentElement = this.element;
+      let path = [];
+
+      // Build the complete path from the element up to html
+      while (currentElement && currentElement !== document) {
+        let tagName = currentElement.tagName.toLowerCase();
+        const parent = currentElement.parentElement;
+
+        if (parent) {
+          const siblings = Array.from(parent.children).filter(
+            (child) => child.tagName.toLowerCase() === tagName,
+          );
+          if (siblings.length > 1) {
+            const index = siblings.indexOf(currentElement) + 1;
+            tagName += `:nth-of-type(${index})`;
+          }
+        }
+
+        path.unshift(tagName);
+        currentElement = currentElement.parentElement;
+      }
+
+      // Create the full selector string with ">" separators
+      if (path.length > 1) {
+        selectors.push(path.join(' > '));
+      }
+
+      return selectors;
+    }
+
     getCombinedSelector() {
       const classSelector = this.getClassSelector()[0];
       return classSelector ? [`${this.getTagSelector()}${classSelector}`] : [];