Update web driver to handle mutiple tasks per agent

Visagan Guruparan · Visagan Guruparan · commit d6672cbdddd2 · 2025-03-19T15:59:11.000-05:00
diff --git a/src/Infrastructure/BotSharp.Abstraction/Infrastructures/Enums/StateConst.cs b/src/Infrastructure/BotSharp.Abstraction/Infrastructures/Enums/StateConst.cs
@@ -16,4 +16,5 @@ public class StateConst
 
     public const string SUB_CONVERSATION_ID = "sub_conversation_id";
     public const string ORIGIN_CONVERSATION_ID = "origin_conversation_id";
+    public const string WEB_DRIVER_TASK_ID = "web_driver_task_id";
 }
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Functions/ChangeCheckboxFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Functions/ChangeCheckboxFn.cs
@@ -34,7 +34,7 @@ public async Task<bool> Execute(RoleDialogModel message)
         message.Data = await _browser.ScreenshotAsync(new MessageInfo
         {
             AgentId = message.CurrentAgentId,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         }, path);
 
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Functions/ChangeListValueFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Functions/ChangeListValueFn.cs
@@ -34,7 +34,7 @@ public async Task<bool> Execute(RoleDialogModel message)
         message.Data = await _browser.ScreenshotAsync(new MessageInfo
         {
             AgentId = message.CurrentAgentId,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         }, path);
 
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Functions/CheckRadioButtonFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Functions/CheckRadioButtonFn.cs
@@ -34,7 +34,7 @@ public async Task<bool> Execute(RoleDialogModel message)
         message.Data = await _browser.ScreenshotAsync(new MessageInfo
         {
             AgentId = message.CurrentAgentId,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         }, path);
 
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Functions/ClickButtonFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Functions/ClickButtonFn.cs
@@ -34,7 +34,7 @@ public async Task<bool> Execute(RoleDialogModel message)
         message.Data = await _browser.ScreenshotAsync(new MessageInfo
         {
             AgentId = message.CurrentAgentId,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         }, path);
 
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Functions/ClickElementFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Functions/ClickElementFn.cs
@@ -34,7 +34,7 @@ public async Task<bool> Execute(RoleDialogModel message)
         message.Data = await _browser.ScreenshotAsync(new MessageInfo
         {
             AgentId = message.CurrentAgentId,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         }, path);
 
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Functions/ExtractDataFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Functions/ExtractDataFn.cs
@@ -28,7 +28,7 @@ public async Task<bool> Execute(RoleDialogModel message)
         message.Data = await _browser.ScreenshotAsync(new MessageInfo
         {
             AgentId = message.CurrentAgentId,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         }, path);
 
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Functions/GoToPageFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Functions/GoToPageFn.cs
@@ -31,7 +31,7 @@ public async Task<bool> Execute(RoleDialogModel message)
         var result = await _browser.GoToPage(new MessageInfo
         {
             AgentId = message.CurrentAgentId,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         }, new PageActionArgs
         {
@@ -45,7 +45,7 @@ public async Task<bool> Execute(RoleDialogModel message)
         message.Data = await _browser.ScreenshotAsync(new MessageInfo
         {
             AgentId = message.CurrentAgentId,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         }, path);
 
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Functions/HttpRequestFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Functions/HttpRequestFn.cs
@@ -1,3 +1,5 @@
+using BotSharp.Plugin.WebDriver.Services;
+
 namespace BotSharp.Plugin.WebDriver.Functions;
 
 public class HttpRequestFn : IFunctionCallback
@@ -20,12 +22,13 @@ public async Task<bool> Execute(RoleDialogModel message)
         var args = JsonSerializer.Deserialize<HttpRequestParams>(message.FunctionArgs);
 
         var agentService = _services.GetRequiredService<IAgentService>();
+        var webDriverService = _services.GetRequiredService<WebDriverService>();
         var agent = await agentService.LoadAgent(message.CurrentAgentId);
         var result = await _browser.SendHttpRequest(new MessageInfo
         {
             AgentId = agent.Id,
             MessageId = message.MessageId,
-            ContextId = convService.ConversationId
+            ContextId = webDriverService.GetMessageContext(message)
         }, args);
 
         message.Content = result.IsSuccess ? 
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Functions/InputUserPasswordFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Functions/InputUserPasswordFn.cs
@@ -33,7 +33,7 @@ public async Task<bool> Execute(RoleDialogModel message)
         message.Data = await _browser.ScreenshotAsync(new MessageInfo
         {
             AgentId = message.CurrentAgentId,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         }, path);
 
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Functions/InputUserTextFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Functions/InputUserTextFn.cs
@@ -39,7 +39,7 @@ public async Task<bool> Execute(RoleDialogModel message)
         message.Data = await _browser.ScreenshotAsync(new MessageInfo
         {
             AgentId = message.CurrentAgentId,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         }, path);
 
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Functions/OpenBrowserFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Functions/OpenBrowserFn.cs
@@ -31,7 +31,7 @@ public async Task<bool> Execute(RoleDialogModel message)
         var msgInfo = new MessageInfo
         {
             AgentId = message.CurrentAgentId,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         };
         var result = await _browser.LaunchBrowser(msgInfo, new BrowserActionArgs
@@ -58,7 +58,7 @@ public async Task<bool> Execute(RoleDialogModel message)
         message.Data = await _browser.ScreenshotAsync(new MessageInfo
         {
             AgentId = message.CurrentAgentId,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         }, path);
 
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Functions/ScreenshotFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Functions/ScreenshotFn.cs
@@ -24,7 +24,7 @@ public async Task<bool> Execute(RoleDialogModel message)
         message.Data = await _browser.ScreenshotAsync(new MessageInfo
         {
             AgentId = message.CurrentAgentId,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         }, path);
         message.Content = "Took screenshot completed. You can take another screenshot if needed.";
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Functions/ScrollPageFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Functions/ScrollPageFn.cs
@@ -20,12 +20,13 @@ public async Task<bool> Execute(RoleDialogModel message)
         var args = JsonSerializer.Deserialize<BrowsingContextIn>(message.FunctionArgs);
 
         var agentService = _services.GetRequiredService<IAgentService>();
+        var webDriverService = _services.GetRequiredService<WebDriverService>();
         var agent = await agentService.LoadAgent(message.CurrentAgentId);
 
         message.Data = await _browser.ScrollPage(new MessageInfo
         {
             AgentId = agent.Id,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         }, new PageActionArgs
         {
@@ -35,13 +36,12 @@ public async Task<bool> Execute(RoleDialogModel message)
     
         message.Content = "Scrolled. You can scroll more if needed.";
 
-        var webDriverService = _services.GetRequiredService<WebDriverService>();
         var path = webDriverService.GetScreenshotFilePath(message.MessageId);
 
         message.Data = await _browser.ScreenshotAsync(new MessageInfo
         {
             AgentId = message.CurrentAgentId,
-            ContextId = convService.ConversationId,
+            ContextId = webDriverService.GetMessageContext(message),
             MessageId = message.MessageId
         }, path);
 
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Services/WebDriverService.GetMessageContext.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Services/WebDriverService.GetMessageContext.cs
@@ -0,0 +1,20 @@
+using BotSharp.Abstraction.Infrastructures.Enums;
+
+namespace BotSharp.Plugin.WebDriver.Services
+{
+    public partial class WebDriverService
+    {
+        public string GetMessageContext(RoleDialogModel message)
+        {
+            var states = _services.GetService<IConversationStateService>();
+            var convService = _services.GetRequiredService<IConversationService>();
+            var webDriverTaskId = states.GetState(StateConst.WEB_DRIVER_TASK_ID, "");
+            var contextId = message.CurrentAgentId;
+            if (!string.IsNullOrWhiteSpace(webDriverTaskId))
+            {
+                contextId = webDriverTaskId;
+            }
+            return contextId;
+        }
+    }
+}
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/UtilFunctions/UtilWebActionOnElementFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/UtilFunctions/UtilWebActionOnElementFn.cs
@@ -35,17 +35,17 @@ public async Task<bool> Execute(RoleDialogModel message)
         var conv = _services.GetRequiredService<IConversationService>();
 
         var browser = _services.GetRequiredService<IWebBrowser>();
+        var webDriverService = _services.GetRequiredService<WebDriverService>();
         var msg = new MessageInfo
         {
             AgentId = message.CurrentAgentId,
             MessageId = message.MessageId,
-            ContextId = message.CurrentAgentId,
+            ContextId = webDriverService.GetMessageContext(message),
         };
         var result = await browser.ActionOnElement(msg, locatorArgs, actionArgs);
 
         message.Content = $"{actionArgs.Action} executed {(result.IsSuccess ? "success" : "failed")}";
 
-        var webDriverService = _services.GetRequiredService<WebDriverService>();
         var path = webDriverService.GetScreenshotFilePath(message.MessageId);
 
         message.Data = await browser.ScreenshotAsync(msg, path);
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/UtilFunctions/UtilWebCloseBrowserFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/UtilFunctions/UtilWebCloseBrowserFn.cs
@@ -18,20 +18,19 @@ public UtilWebCloseBrowserFn(
     public async Task<bool> Execute(RoleDialogModel message)
     {
         var conv = _services.GetRequiredService<IConversationService>();
-
+        var webDriverService = _services.GetRequiredService<WebDriverService>();
         var browser = _services.GetRequiredService<IWebBrowser>();
         var msg = new MessageInfo
         {
             AgentId = message.CurrentAgentId,
             MessageId = message.MessageId,
-            ContextId = message.CurrentAgentId,
+            ContextId = webDriverService.GetMessageContext(message)
         };
 
         await browser.CloseBrowser(message.CurrentAgentId);
 
         message.Content = $"Browser closed.";
 
-        var webDriverService = _services.GetRequiredService<WebDriverService>();
         var path = webDriverService.GetScreenshotFilePath(message.MessageId);
 
         message.Data = await browser.ScreenshotAsync(msg, path);
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/UtilFunctions/UtilWebGoToPageFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/UtilFunctions/UtilWebGoToPageFn.cs
@@ -29,13 +29,13 @@ public async Task<bool> Execute(RoleDialogModel message)
         args.WaitTime = _webDriver.DefaultWaitTime;
 
         var conv = _services.GetRequiredService<IConversationService>();
-
+        var webDriverService = _services.GetRequiredService<WebDriverService>();
         var browser = _services.GetRequiredService<IWebBrowser>();
         var msg = new MessageInfo
         {
             AgentId = message.CurrentAgentId,
             MessageId = message.MessageId,
-            ContextId = message.CurrentAgentId,
+            ContextId = webDriverService.GetMessageContext(message)
         };
         if (!args.KeepBrowserOpen)
         {
@@ -50,7 +50,6 @@ public async Task<bool> Execute(RoleDialogModel message)
 
         message.Content = $"Open web page successfully.";
 
-        var webDriverService = _services.GetRequiredService<WebDriverService>();
         var path = webDriverService.GetScreenshotFilePath(message.MessageId);
 
         message.Data = await browser.ScreenshotAsync(msg, path);
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/UtilFunctions/UtilWebLocateElementFn.cs b/src/Plugins/BotSharp.Plugin.WebDriver/UtilFunctions/UtilWebLocateElementFn.cs
@@ -22,17 +22,17 @@ public async Task<bool> Execute(RoleDialogModel message)
         locatorArgs.Highlight = true;
 
         var browser = _services.GetRequiredService<IWebBrowser>();
+        var webDriverService = _services.GetRequiredService<WebDriverService>();
         var msg = new MessageInfo
         {
             AgentId = message.CurrentAgentId,
             MessageId = message.MessageId,
-            ContextId = message.CurrentAgentId,
+            ContextId = webDriverService.GetMessageContext(message)
         };
         var result = await browser.LocateElement(msg, locatorArgs);
 
         message.Content = $"Locating element {(result.IsSuccess ? "success" : "failed")}";
 
-        var webDriverService = _services.GetRequiredService<WebDriverService>();
         var path = webDriverService.GetScreenshotFilePath(message.MessageId);
 
         message.Data = await browser.ScreenshotAsync(msg, path);
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/WebDriverPlugin.cs b/src/Plugins/BotSharp.Plugin.WebDriver/WebDriverPlugin.cs
@@ -23,7 +23,7 @@ public void RegisterDI(IServiceCollection services, IConfiguration config)
             var settingService = provider.GetRequiredService<ISettingService>();
             return settings;
         });
-
+        services.AddSingleton<WebBrowsingSettings>();
         services.AddScoped<PlaywrightWebDriver>();
         services.AddSingleton<PlaywrightInstance>();
 

Original file line number	Diff line number	Diff line change
`@@ -16,4 +16,5 @@ public class StateConst`
`16`	`16`
`17`	`17`	`public const string SUB_CONVERSATION_ID = "sub_conversation_id";`
`18`	`18`	`public const string ORIGIN_CONVERSATION_ID = "origin_conversation_id";`
	`19`	`+ public const string WEB_DRIVER_TASK_ID = "web_driver_task_id";`
`19`	`20`	`}`
Original file line number	Diff line number	Diff line change
`@@ -34,7 +34,7 @@ public async Task<bool> Execute(RoleDialogModel message)`
`34`	`34`	`message.Data = await _browser.ScreenshotAsync(new MessageInfo`
`35`	`35`	`{`
`36`	`36`	`AgentId = message.CurrentAgentId,`
`37`		`- ContextId = convService.ConversationId,`
	`37`	`+ ContextId = webDriverService.GetMessageContext(message),`
`38`	`38`	`MessageId = message.MessageId`
`39`	`39`	`}, path);`
`40`	`40`
Original file line number	Diff line number	Diff line change
`@@ -28,7 +28,7 @@ public async Task<bool> Execute(RoleDialogModel message)`
`28`	`28`	`message.Data = await _browser.ScreenshotAsync(new MessageInfo`
`29`	`29`	`{`
`30`	`30`	`AgentId = message.CurrentAgentId,`
`31`		`- ContextId = convService.ConversationId,`
	`31`	`+ ContextId = webDriverService.GetMessageContext(message),`
`32`	`32`	`MessageId = message.MessageId`
`33`	`33`	`}, path);`
`34`	`34`
Original file line number	Diff line number	Diff line change
`@@ -31,7 +31,7 @@ public async Task<bool> Execute(RoleDialogModel message)`
`31`	`31`	`var result = await _browser.GoToPage(new MessageInfo`
`32`	`32`	`{`
`33`	`33`	`AgentId = message.CurrentAgentId,`
`34`		`- ContextId = convService.ConversationId,`
	`34`	`+ ContextId = webDriverService.GetMessageContext(message),`
`35`	`35`	`MessageId = message.MessageId`
`36`	`36`	`}, new PageActionArgs`
`37`	`37`	`{`
`@@ -45,7 +45,7 @@ public async Task<bool> Execute(RoleDialogModel message)`
`45`	`45`	`message.Data = await _browser.ScreenshotAsync(new MessageInfo`
`46`	`46`	`{`
`47`	`47`	`AgentId = message.CurrentAgentId,`
`48`		`- ContextId = convService.ConversationId,`
	`48`	`+ ContextId = webDriverService.GetMessageContext(message),`
`49`	`49`	`MessageId = message.MessageId`
`50`	`50`	`}, path);`
`51`	`51`
Original file line number	Diff line number	Diff line change
`@@ -33,7 +33,7 @@ public async Task<bool> Execute(RoleDialogModel message)`
`33`	`33`	`message.Data = await _browser.ScreenshotAsync(new MessageInfo`
`34`	`34`	`{`
`35`	`35`	`AgentId = message.CurrentAgentId,`
`36`		`- ContextId = convService.ConversationId,`
	`36`	`+ ContextId = webDriverService.GetMessageContext(message),`
`37`	`37`	`MessageId = message.MessageId`
`38`	`38`	`}, path);`
`39`	`39`