dhalsim 2.3.0 → 2.4.0

package/dist/index.js

@@ -15962,7 +15962,7 @@ function createGadgetsByName(names, config2) {
 }
 
 // src/subagents/dhalsim.ts
-import { Gadget as Gadget13, z as z14, getHostExports, resolveSubagentModel, resolveValue } from "llmist";
+import { Gadget as Gadget13, z as z14, getHostExports, resolveSubagentModel, resolveValue, TaskCompletionSignal } from "llmist";
 
 // src/state/page-state.ts
 var DEFAULT_CONFIG = {
@@ -16553,25 +16553,7 @@ ${states.join("\n\n")}
 };
 
 // src/subagents/prompts.ts
-var GADGET_LIST_WITH_USER_ASSISTANCE = `## Available Gadgets
-- ReportResult: **REQUIRED** - Call this to return your findings when task is complete
-- Navigate: Go to a URL
-- Click: Click an element (auto-waits for element to be actionable)
-- Fill: Fill a form input
-- FillForm: Fill multiple fields and submit
-- Select: Select dropdown option
-- Check: Toggle checkboxes
-- GetFullPageContent: Read page text content
-- Screenshot: Capture the page (use when you need to show visual results)
-- DismissOverlays: Auto-dismiss cookie banners
-- Scroll: Scroll the page
-- WaitForElement: Wait for an element to appear
-- Wait: General wait
-- RequestUserAssistance: Ask user for help with CAPTCHAs, 2FA codes, or other human-only challenges`;
-var GADGET_LIST_WITHOUT_USER_ASSISTANCE = `## Available Gadgets
-- ReportResult: **REQUIRED** - Call this to return your findings when task is complete
-- Navigate: Go to a URL
-- Click: Click an element (auto-waits for element to be actionable)
+var GADGETS_CORE = `- Click: Click an element (auto-waits for element to be actionable)
 - Fill: Fill a form input
 - FillForm: Fill multiple fields and submit
 - Select: Select dropdown option
@@ -16582,8 +16564,44 @@ var GADGET_LIST_WITHOUT_USER_ASSISTANCE = `## Available Gadgets
 - Scroll: Scroll the page
 - WaitForElement: Wait for an element to appear
 - Wait: General wait`;
+function buildGadgetList(options) {
+  const lines = ["## Available Gadgets", "- ReportResult: **REQUIRED** - Call this to return your findings when task is complete"];
+  if (options.includeNavigation) {
+    lines.push("- Navigate: Go to a URL");
+  }
+  lines.push(GADGETS_CORE);
+  if (options.includeUserAssistance) {
+    lines.push("- RequestUserAssistance: Ask user for help with CAPTCHAs, 2FA codes, or other human-only challenges");
+  }
+  return lines.join("\n");
+}
 function createDhalsimSystemPrompt(options) {
-  const gadgetList = options.includeUserAssistance ? GADGET_LIST_WITH_USER_ASSISTANCE : GADGET_LIST_WITHOUT_USER_ASSISTANCE;
+  const includeNavigation = !options.disableNavigation;
+  const gadgetList = buildGadgetList({
+    includeNavigation,
+    includeUserAssistance: options.includeUserAssistance
+  });
+  const criticalRules = includeNavigation ? `## CRITICAL Rules
+1. You have ONE page (p1) already open. Use Navigate to go to URLs.
+2. ONLY use selectors exactly as shown in <CurrentBrowserState>
+3. NEVER guess selectors - use GetFullPageContent if you need more info
+4. Focus on completing the task efficiently - avoid unnecessary actions
+5. If a selector matches multiple elements, you'll get an error with a "suggestions" array containing valid selectors. USE ONE OF THESE SUGGESTIONS DIRECTLY - don't guess or modify them.
+6. For batch extraction: GetFullPageContent returns ALL matches when a selector matches multiple elements (as "texts" array). Use this instead of querying each element separately.` : `## CRITICAL Rules
+1. You have ONE page (p1) open at the target URL. You cannot navigate to other URLs.
+2. ONLY use selectors exactly as shown in <CurrentBrowserState>
+3. NEVER guess selectors - use GetFullPageContent if you need more info
+4. Focus on completing the task efficiently - avoid unnecessary actions
+5. If a selector matches multiple elements, you'll get an error with a "suggestions" array containing valid selectors. USE ONE OF THESE SUGGESTIONS DIRECTLY - don't guess or modify them.
+6. For batch extraction: GetFullPageContent returns ALL matches when a selector matches multiple elements (as "texts" array). Use this instead of querying each element separately.`;
+  const efficientPattern = includeNavigation ? `## Efficient Pattern
+On first call: Navigate and DismissOverlays are ALREADY done. Take action immediately.
+After any Navigate call: DismissOverlays, then interact with elements.
+
+If an action doesn't produce expected results, use GetFullPageContent to diagnose before retrying.` : `## Efficient Pattern
+On first call: DismissOverlays is ALREADY done. Take action immediately.
+
+If an action doesn't produce expected results, use GetFullPageContent to diagnose before retrying.`;
   return `You are a browser automation agent focused on completing a specific web task.
 
 ## Browser State (<CurrentBrowserState>)
@@ -16597,19 +16615,9 @@ This is your source of truth for what's on screen. It contains:
 - CHECKBOXES: Checkbox/radio inputs
 - MENUITEMS: Dropdown options (only visible when dropdown is open)
 
-## CRITICAL Rules
-1. You have ONE page (p1) already open. Use Navigate to go to URLs.
-2. ONLY use selectors exactly as shown in <CurrentBrowserState>
-3. NEVER guess selectors - use GetFullPageContent if you need more info
-4. Focus on completing the task efficiently - avoid unnecessary actions
-5. If a selector matches multiple elements, you'll get an error with a "suggestions" array containing valid selectors. USE ONE OF THESE SUGGESTIONS DIRECTLY - don't guess or modify them.
-6. For batch extraction: GetFullPageContent returns ALL matches when a selector matches multiple elements (as "texts" array). Use this instead of querying each element separately.
+${criticalRules}
 
-## Efficient Pattern
-On first call: Navigate and DismissOverlays are ALREADY done. Take action immediately.
-After any Navigate call: DismissOverlays, then interact with elements.
-
-If an action doesn't produce expected results, use GetFullPageContent to diagnose before retrying.
+${efficientPattern}
 
 ## Dropdown/Toggle Behavior
 Dropdowns are TOGGLES - clicking the same trigger twice will close it!
@@ -16660,10 +16668,8 @@ var ReportResult = class extends Gadget13({
     )
   })
 }) {
-  result = null;
   execute(params) {
-    this.result = params.result;
-    return "Result reported successfully.";
+    throw new TaskCompletionSignal(params.result);
   }
 };
 var Dhalsim = class extends Gadget13({
@@ -16680,7 +16686,8 @@ Use this for web research, data extraction, form filling, or any web-based task.
     headless: z14.boolean().optional().describe("Run browser in headless mode (default: true, configurable via CLI)"),
     timeoutMs: z14.number().optional().describe("Overall timeout in ms (default: 300000 = 5 min, 0 = disabled, configurable via CLI)"),
     disableCache: z14.boolean().optional().describe("Disable browser cache for lower memory usage (default: false, configurable via CLI)"),
-    navigationTimeoutMs: z14.number().optional().describe("Navigation timeout in ms (default: 60000, configurable via CLI)")
+    navigationTimeoutMs: z14.number().optional().describe("Navigation timeout in ms (default: 60000, configurable via CLI)"),
+    disableNavigation: z14.boolean().optional().describe("Disable Navigate gadget to restrict agent to the initial URL (default: false, configurable via CLI)")
   }),
   timeoutMs: 3e5
   // 5 minutes - web browsing can take time
@@ -16688,12 +16695,14 @@ Use this for web research, data extraction, form filling, or any web-based task.
   customSessionManager;
   customSystemPrompt;
   userAssistanceEnabled;
+  navigationDisabled;
   customUserAssistanceCallback;
   constructor(options) {
     super();
     this.customSessionManager = options?.sessionManager;
     this.customSystemPrompt = options?.systemPrompt;
     this.userAssistanceEnabled = options?.userAssistance;
+    this.navigationDisabled = options?.disableNavigation;
     this.customUserAssistanceCallback = options?.onUserAssistance;
     if (options?.timeoutMs !== void 0) {
       this.timeoutMs = options.timeoutMs === 0 ? void 0 : options.timeoutMs;
@@ -16724,8 +16733,13 @@ Use this for web research, data extraction, form filling, or any web-based task.
       subagentKey: "navigationTimeoutMs",
       defaultValue: 6e4
     });
+    const disableNavigation = resolveValue(ctx, "BrowseWeb", {
+      runtime: params.disableNavigation,
+      subagentKey: "disableNavigation",
+      defaultValue: this.navigationDisabled ?? false
+    });
     const userAssistanceEnabled = this.userAssistanceEnabled ?? (this.customUserAssistanceCallback !== void 0 || ctx?.requestHumanInput !== void 0);
-    logger13?.debug(`[BrowseWeb] User assistance enabled=${userAssistanceEnabled}`);
+    logger13?.debug(`[BrowseWeb] User assistance enabled=${userAssistanceEnabled}, disableNavigation=${disableNavigation}`);
     const collectedMedia = [];
     const manager = this.customSessionManager ?? new BrowserSessionManager(logger13);
     const isOwnedManager = !this.customSessionManager;
@@ -16771,7 +16785,8 @@ Use this for web research, data extraction, form filling, or any web-based task.
       const gadgets = [
         reportResult,
         // First so it's prominent in the list
-        new Navigate(manager),
+        // Conditionally include Navigate (excluded when disableNavigation is true)
+        ...disableNavigation ? [] : [new Navigate(manager)],
         new Click(manager),
         new Fill(manager),
         new FillForm(manager),
@@ -16788,7 +16803,7 @@ Use this for web research, data extraction, form filling, or any web-based task.
       ];
       const { AgentBuilder, LLMist } = getHostExports(ctx);
       const client = new LLMist();
-      const systemPrompt = this.customSystemPrompt ?? createDhalsimSystemPrompt({ includeUserAssistance: userAssistanceEnabled });
+      const systemPrompt = this.customSystemPrompt ?? createDhalsimSystemPrompt({ includeUserAssistance: userAssistanceEnabled, disableNavigation });
       const builder = new AgentBuilder(client).withModel(model).withSystem(systemPrompt).withMaxIterations(maxIterations).withGadgets(...gadgets).withTrailingMessage((trailingCtx) => [
         pageStateScanner.getCachedState(),
         "",
@@ -16844,6 +16859,7 @@ Use this for web research, data extraction, form filling, or any web-based task.
 Task: ${task}`);
       logger13?.debug(`[BrowseWeb] Starting agent loop model=${model} maxIterations=${maxIterations}`);
       let finalResult = "";
+      let reportedResult;
       for await (const event of agent.run()) {
         if (ctx?.signal?.aborted) {
           break;
@@ -16854,15 +16870,15 @@ Task: ${task}`);
               collectedMedia.push(media);
             }
           }
-          if (reportResult.result !== null) {
-            break;
+          if (event.result.gadgetName === "ReportResult" && event.result.breaksLoop) {
+            reportedResult = event.result.result;
           }
         } else if (event.type === "text") {
           finalResult = event.content;
         }
       }
       return {
-        result: reportResult.result || finalResult || "Task completed but no result text was generated.",
+        result: reportedResult || finalResult || "Task completed but no result text was generated.",
         media: collectedMedia.length > 0 ? collectedMedia : void 0
       };
     } finally {