@jarvis-agent/core 0.3.0 → 0.3.1

package/dist/index.esm.js

@@ -39669,6 +39669,24 @@ Output:
   "changed": true,
   "changeInfo": "New message received in the group chat. The message content is: 'Hello, how are you?'"
 }`;
+const watch_text_system_prompt = `You are a page content analyzer. Given a page content and a condition description, determine if the condition is currently met on the page.
+Return ONLY a JSON object, no other text.
+- "changed": true means the condition IS met (e.g. the target button exists, the status has changed to the expected value)
+- "changed": false means the condition is NOT yet met
+
+## Example
+Condition: Monitor for a "Retry" button appearing on the page
+### Condition not met
+Output:
+{
+  "changed": false
+}
+### Condition met
+Output:
+{
+  "changed": true,
+  "changeInfo": "The 'Retry' button is present on the page at index 127-128"
+}`;
 class WatchTriggerTool {
     constructor() {
         this.name = TOOL_NAME$a;
@@ -39733,12 +39751,34 @@ class WatchTriggerTool {
                 ],
             };
         }
+        const rlm = new RetryLanguageModel(agentContext.context.config.llms, agentContext.agent.Llms, agentContext.context.config.globalConfig?.streamFirstTimeout, agentContext.context.config.globalConfig?.streamTokenTimeout, agentContext);
+        const useVision = this.isVisionModel(rlm);
+        // Initial condition check (text-based, works with all models)
+        const pageContent = await this.get_page_content(agentContext);
+        const initialCheck = await this.is_condition_met(rlm, pageContent, task_description, agentContext);
+        if (initialCheck.changed) {
+            return {
+                content: [
+                    {
+                        type: "text",
+                        text: initialCheck.changeInfo || "Condition already met on page.",
+                    },
+                ],
+            };
+        }
+        // Enter monitoring loop
         await this.init_eko_observer(agentContext);
-        const image1 = await this.get_screenshot(agentContext);
         const start = new Date().getTime();
         const timeout = (args.timeout || 5) * 60000;
         const frequency = Math.max(500, (args.frequency || 1) * 1000);
-        const rlm = new RetryLanguageModel(agentContext.context.config.llms, agentContext.agent.Llms, agentContext.context.config.globalConfig?.streamFirstTimeout, agentContext.context.config.globalConfig?.streamTokenTimeout, agentContext);
+        let image1;
+        let content1;
+        if (useVision) {
+            image1 = await this.get_screenshot(agentContext);
+        }
+        else {
+            content1 = pageContent;
+        }
         while (new Date().getTime() - start < timeout) {
             await agentContext.context.checkAborted();
             await new Promise((resolve) => setTimeout(resolve, frequency));
@@ -39747,17 +39787,32 @@ class WatchTriggerTool {
                 continue;
             }
             await this.init_eko_observer(agentContext);
-            const image2 = await this.get_screenshot(agentContext);
-            const changeResult = await this.is_dom_change(agentContext, rlm, image1, image2, task_description);
-            if (changeResult.changed) {
-                return {
-                    content: [
-                        {
-                            type: "text",
-                            text: changeResult.changeInfo || "DOM change detected.",
-                        },
-                    ],
-                };
+            if (useVision) {
+                // Vision model: compare screenshots
+                const image2 = await this.get_screenshot(agentContext);
+                const changeResult = await this.is_dom_change(agentContext, rlm, image1, image2, task_description);
+                if (changeResult.changed) {
+                    return {
+                        content: [
+                            { type: "text", text: changeResult.changeInfo || "DOM change detected." },
+                        ],
+                    };
+                }
+            }
+            else {
+                // Text model: compare page content
+                const content2 = await this.get_page_content(agentContext);
+                if (content2 === content1)
+                    continue;
+                const changeResult = await this.is_condition_met(rlm, content2, task_description, agentContext);
+                content1 = content2;
+                if (changeResult.changed) {
+                    return {
+                        content: [
+                            { type: "text", text: changeResult.changeInfo || "Condition met." },
+                        ],
+                    };
+                }
             }
         }
         return {
@@ -39769,6 +39824,65 @@ class WatchTriggerTool {
             ],
         };
     }
+    /** Check if the primary LLM supports vision */
+    isVisionModel(rlm) {
+        const names = rlm.Names;
+        const llms = rlm.Llms;
+        if (!names || names.length === 0)
+            return false;
+        const config = llms[names[0]];
+        if (!config)
+            return false;
+        const provider = String(config.provider || "").toLowerCase();
+        const model = String(config.model || "").toLowerCase();
+        if (provider === "deepseek" || model.includes("deepseek"))
+            return false;
+        if (provider === "anthropic")
+            return true;
+        if (provider === "google")
+            return true;
+        if (model.includes("gpt-4o") || model.includes("gpt-4-vision") || model.includes("gpt-4-turbo"))
+            return true;
+        if (model.includes("claude") || model.includes("gemini"))
+            return true;
+        return false;
+    }
+    /** Get page text content via extract_page_content */
+    async get_page_content(agentContext) {
+        const extract = agentContext.agent["extract_page_content"];
+        if (!extract)
+            return "";
+        const result = await extract.call(agentContext.agent, agentContext);
+        return result?.page_content || "";
+    }
+    /** Check if condition is met using text-based LLM analysis */
+    async is_condition_met(rlm, pageContent, task_description, agentContext) {
+        try {
+            const request = {
+                messages: [
+                    { role: "system", content: watch_text_system_prompt },
+                    {
+                        role: "user",
+                        content: [
+                            {
+                                type: "text",
+                                text: `Condition: ${task_description}\n\nPage content:\n${pageContent.slice(0, 30000)}`,
+                            },
+                        ],
+                    },
+                ],
+                abortSignal: agentContext.context.controller.signal,
+            };
+            const result = await rlm.call(request);
+            let resultText = result.text || "{}";
+            resultText = resultText.substring(resultText.indexOf("{"), resultText.lastIndexOf("}") + 1);
+            return JSON.parse(resultText);
+        }
+        catch (error) {
+            Log.error("Error in is_condition_met:", error);
+        }
+        return { changed: false };
+    }
     async get_screenshot(agentContext) {
         const screenshot = agentContext.agent["screenshot"];
         const imageResult = (await screenshot.call(agentContext.agent, agentContext));
@@ -40178,7 +40292,7 @@ class ActivateSkillTool {
             required: ["name"],
         };
     }
-    async execute(args) {
+    async execute(args, ..._rest) {
         const name = args.name;
         if (!global.skillService) {
             return {
@@ -40475,11 +40589,11 @@ monitor changes in webpage DOM elements, when executing to the watch node, requi
 <if ${TOOL_NAME$6}Tool>
 * SKILLS
 You can use the \`${TOOL_NAME$6}\` tool to load domain-specific skill instructions when they would help complete the current task.
+</if>
 <if skills>
 Available skills:
 {{skills}}
 </if>
-</if>
 
 <if mainTask>
 Main task: {{mainTask}}