Merge pull request #411 from Qtoss-Alpha/master

Optimize WebDriver with context id.
SciSharp · Apr 12, 2024 · 1992a70 · 1992a70
2 parents 887619d + db709b2
commit 1992a70
Show file tree

Hide file tree

Showing 27 changed files with 111 additions and 85 deletions.
diff --git a/src/Infrastructure/BotSharp.Abstraction/Browsing/IWebBrowser.cs b/src/Infrastructure/BotSharp.Abstraction/Browsing/IWebBrowser.cs
@@ -4,8 +4,8 @@ namespace BotSharp.Abstraction.Browsing;
 
 public interface IWebBrowser
 {
-    Task<BrowserActionResult> LaunchBrowser(string conversationId, string? url);
-    Task<BrowserActionResult> ScreenshotAsync(string conversationId, string path);
+    Task<BrowserActionResult> LaunchBrowser(string contextId, string? url, bool openIfNotExist = true);
+    Task<BrowserActionResult> ScreenshotAsync(string contextId, string path);
     Task<BrowserActionResult> ScrollPageAsync(BrowserActionParams actionParams);
 
     Task<BrowserActionResult> ActionOnElement(MessageInfo message, ElementLocatingArgs location, ElementActionArgs action);
@@ -19,10 +19,11 @@ public interface IWebBrowser
     Task<BrowserActionResult> ChangeListValue(BrowserActionParams actionParams);
     Task<BrowserActionResult> CheckRadioButton(BrowserActionParams actionParams);
     Task<BrowserActionResult> ChangeCheckbox(BrowserActionParams actionParams);
-    Task<BrowserActionResult> GoToPage(string conversationId, string url);
+    Task<BrowserActionResult> GoToPage(string contextId, string url, bool openNewTab = false);
     Task<string> ExtractData(BrowserActionParams actionParams);
-    Task<T> EvaluateScript<T>(string conversationId, string script);
-    Task CloseBrowser(string conversationId);
-    Task<BrowserActionResult> SendHttpRequest(HttpRequestParams actionParams);
+    Task<T> EvaluateScript<T>(string contextId, string script);
+    Task CloseBrowser(string contextId);
+    Task CloseCurrentPage(string contextId);
+    Task<BrowserActionResult> SendHttpRequest(string contextId, HttpRequestParams actionParams);
     Task<string> GetAttributeValue(MessageInfo message, ElementLocatingArgs location, BrowserActionResult result);
 }
diff --git a/src/Infrastructure/BotSharp.Abstraction/Browsing/Models/BrowserActionParams.cs b/src/Infrastructure/BotSharp.Abstraction/Browsing/Models/BrowserActionParams.cs
@@ -4,14 +4,14 @@ public class BrowserActionParams
 {
     public Agent Agent { get; set; }
     public BrowsingContextIn Context { get; set; }
-    public string ConversationId { get; set; }
+    public string ContextId { get; set; }
     public string MessageId { get; set; }
 
-    public BrowserActionParams(Agent agent, BrowsingContextIn context, string conversationId, string messageId)
+    public BrowserActionParams(Agent agent, BrowsingContextIn context, string contextId, string messageId)
     {
         Agent = agent;
         Context = context;
-        ConversationId = conversationId;
+        ContextId = contextId;
         MessageId = messageId;
     }
 }
diff --git a/src/Infrastructure/BotSharp.Abstraction/Browsing/Models/MessageInfo.cs b/src/Infrastructure/BotSharp.Abstraction/Browsing/Models/MessageInfo.cs
@@ -3,6 +3,6 @@ namespace BotSharp.Abstraction.Browsing.Models;
 public class MessageInfo
 {
     public string AgentId { get; set; }
-    public string ConversationId { get; set; }
+    public string ContextId { get; set; }
     public string MessageId { get; set; }
 }
diff --git a/src/Plugins/BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightInstance.cs b/src/Plugins/BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightInstance.cs
@@ -6,31 +6,30 @@ public class PlaywrightInstance : IDisposable
 {
     IPlaywright _playwright;
     Dictionary<string, IBrowserContext> _contexts = new Dictionary<string, IBrowserContext>();
+    public Dictionary<string, IBrowserContext> Contexts => _contexts;
 
     public IPage GetPage(string id)
     {
         InitInstance(id).Wait();
         return _contexts[id].Pages.LastOrDefault();
     }
 
-    public async Task InitInstance(string id)
+    public async Task<IBrowserContext> InitInstance(string id)
     {
         if (_playwright == null)
         {
             _playwright = await Playwright.CreateAsync();
         }
-        await InitContext(id);
+        return await InitContext(id);
     }
 
-    public async Task InitContext(string id)
+    public async Task<IBrowserContext> InitContext(string id)
     {
         if (_contexts.ContainsKey(id))
-            return;
-#if DEBUG
-        string tempFolderPath = $"{Path.GetTempPath()}\\playwright";
-#else
+            return _contexts[id];
+
         string tempFolderPath = $"{Path.GetTempPath()}\\playwright\\{id}";
-#endif
+
         _contexts[id] = await _playwright.Chromium.LaunchPersistentContextAsync(tempFolderPath, new BrowserTypeLaunchPersistentContextOptions
         {
 #if DEBUG
@@ -65,6 +64,8 @@ public async Task InitContext(string id)
             Serilog.Log.Warning($"Playwright browser context is closed");
             _contexts.Remove(id);
         };
+
+        return _contexts[id];
     }
 
     public async Task<IPage> NewPage(string id)
@@ -92,6 +93,14 @@ public async Task Close(string id)
         }
     }
 
+    public async Task CloseCurrentPage(string id)
+    {
+        if (_contexts.ContainsKey(id))
+        {
+            await GetPage(id).CloseAsync();
+        }
+    }
+
     public void Dispose()
     {
         _contexts.Clear();

diff --git a/...BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.ActionOnElement.cs b/...BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.ActionOnElement.cs
@@ -4,7 +4,7 @@ public partial class PlaywrightWebDriver
 {
     public async Task<BrowserActionResult> ActionOnElement(MessageInfo message, ElementLocatingArgs location, ElementActionArgs action)
     {
-        await _instance.Wait(message.ConversationId);
+        await _instance.Wait(message.ContextId);
         var result = await LocateElement(message, location);
         if (result.IsSuccess)
         {

diff --git a/.../BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.ChangeCheckbox.cs b/.../BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.ChangeCheckbox.cs
@@ -1,5 +1,3 @@
-using System.Text.RegularExpressions;
-
 namespace BotSharp.Plugin.WebDriver.Drivers.PlaywrightDriver;
 
 public partial class PlaywrightWebDriver
@@ -8,7 +6,7 @@ public async Task<BrowserActionResult> ChangeCheckbox(BrowserActionParams action
     {
         var result = new BrowserActionResult();
 
-        await _instance.Wait(actionParams.ConversationId);
+        await _instance.Wait(actionParams.ContextId);
 
         // Retrieve the page raw html and infer the element path
         var regexExpression = actionParams.Context.MatchRule.ToLower() switch
@@ -19,7 +17,7 @@ public async Task<BrowserActionResult> ChangeCheckbox(BrowserActionParams action
             _ => $"^{actionParams.Context.ElementText}$"
         };
         var regex = new Regex(regexExpression, RegexOptions.IgnoreCase);
-        var elements = _instance.GetPage(actionParams.ConversationId).GetByText(regex);
+        var elements = _instance.GetPage(actionParams.ContextId).GetByText(regex);
         var count = await elements.CountAsync();
 
         var errorMessage = $"Can't locate element by keyword {actionParams.Context.ElementText}";
@@ -55,7 +53,7 @@ public async Task<BrowserActionResult> ChangeCheckbox(BrowserActionParams action
         }
         else
         {
-            elements = _instance.GetPage(actionParams.ConversationId).Locator($"#{id}");
+            elements = _instance.GetPage(actionParams.ContextId).Locator($"#{id}");
         }
         count = await elements.CountAsync();
 

diff --git a/...BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.ChangeListValue.cs b/...BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.ChangeListValue.cs
@@ -5,10 +5,10 @@ public partial class PlaywrightWebDriver
     public async Task<BrowserActionResult> ChangeListValue(BrowserActionParams actionParams)
     {
         var result = new BrowserActionResult();
-        await _instance.Wait(actionParams.ConversationId);
+        await _instance.Wait(actionParams.ContextId);
 
         // Retrieve the page raw html and infer the element path
-        var body = await _instance.GetPage(actionParams.ConversationId).QuerySelectorAsync("body");
+        var body = await _instance.GetPage(actionParams.ContextId).QuerySelectorAsync("body");
 
         var str = new List<string>();
         var inputs = await body.QuerySelectorAllAsync("select");
@@ -61,7 +61,7 @@ public async Task<BrowserActionResult> ChangeListValue(BrowserActionParams actio
             string.Join("", str),
             actionParams.Context.ElementName,
             actionParams.MessageId);
-        ILocator element = Locator(actionParams.ConversationId, htmlElementContextOut);
+        ILocator? element = Locator(actionParams.ContextId, htmlElementContextOut);
 
         try
         {
@@ -70,11 +70,11 @@ public async Task<BrowserActionResult> ChangeListValue(BrowserActionParams actio
             if (!isVisible)
             {
                 // Select the element you want to make visible (replace with your own selector)
-                var control = await _instance.GetPage(actionParams.ConversationId)
+                var control = await _instance.GetPage(actionParams.ContextId)
                     .QuerySelectorAsync($"#{htmlElementContextOut.ElementId}");
 
                 // Show the element by modifying its CSS styles
-                await _instance.GetPage(actionParams.ConversationId)
+                await _instance.GetPage(actionParams.ContextId)
                     .EvaluateAsync(@"(element) => {
                         element.style.display = 'block';
                         element.style.visibility = 'visible';
@@ -92,11 +92,11 @@ await element.SelectOptionAsync(new SelectOptionValue
             if (!isVisible)
             {
                 // Select the element you want to make visible (replace with your own selector)
-                var control = await _instance.GetPage(actionParams.ConversationId)
+                var control = await _instance.GetPage(actionParams.ContextId)
                     .QuerySelectorAsync($"#{htmlElementContextOut.ElementId}");
 
                 // Show the element by modifying its CSS styles
-                await _instance.GetPage(actionParams.ConversationId).EvaluateAsync(@"(element) => {
+                await _instance.GetPage(actionParams.ContextId).EvaluateAsync(@"(element) => {
                     element.style.display = 'none';
                     element.style.visibility = 'hidden';
                 }", control);

diff --git a/...otSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.CheckRadioButton.cs b/...otSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.CheckRadioButton.cs
@@ -5,7 +5,7 @@ public partial class PlaywrightWebDriver
     public async Task<BrowserActionResult> CheckRadioButton(BrowserActionParams actionParams)
     {
         var result = new BrowserActionResult();
-        await _instance.Wait(actionParams.ConversationId);
+        await _instance.Wait(actionParams.ContextId);
 
         // Retrieve the page raw html and infer the element path
         var regexExpression = actionParams.Context.MatchRule.ToLower() switch
@@ -16,7 +16,7 @@ public async Task<BrowserActionResult> CheckRadioButton(BrowserActionParams acti
             _ => $"^{actionParams.Context.ElementText}$"
         };
         var regex = new Regex(regexExpression, RegexOptions.IgnoreCase);
-        var elements = _instance.GetPage(actionParams.ConversationId).GetByText(regex);
+        var elements = _instance.GetPage(actionParams.ContextId).GetByText(regex);
         var count = await elements.CountAsync();
 
         var errorMessage = $"Can't locate element by keyword {actionParams.Context.ElementText}";

diff --git a/...ins/BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.ClickButton.cs b/...ins/BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.ClickButton.cs
@@ -5,10 +5,10 @@ public partial class PlaywrightWebDriver
     public async Task<BrowserActionResult> ClickButton(BrowserActionParams actionParams)
     {
         var result = new BrowserActionResult();
-        await _instance.Wait(actionParams.ConversationId);
+        await _instance.Wait(actionParams.ContextId);
 
         // Find by text exactly match
-        var elements = _instance.GetPage(actionParams.ConversationId)
+        var elements = _instance.GetPage(actionParams.ContextId)
             .GetByRole(AriaRole.Button, new PageGetByRoleOptions
             {
                 Name = actionParams.Context.ElementName
@@ -17,7 +17,7 @@ public async Task<BrowserActionResult> ClickButton(BrowserActionParams actionPar
 
         if (count == 0)
         {
-            elements = _instance.GetPage(actionParams.ConversationId)
+            elements = _instance.GetPage(actionParams.ContextId)
                 .GetByRole(AriaRole.Link, new PageGetByRoleOptions
                 {
                     Name = actionParams.Context.ElementName
@@ -27,7 +27,7 @@ public async Task<BrowserActionResult> ClickButton(BrowserActionParams actionPar
 
         if (count == 0)
         {
-            elements = _instance.GetPage(actionParams.ConversationId)
+            elements = _instance.GetPage(actionParams.ContextId)
                 .GetByText(actionParams.Context.ElementName);
             count = await elements.CountAsync();
         }
@@ -36,12 +36,12 @@ public async Task<BrowserActionResult> ClickButton(BrowserActionParams actionPar
         {
             // Infer element if not found
             var driverService = _services.GetRequiredService<WebDriverService>();
-            var html = await FilteredButtonHtml(actionParams.ConversationId);
+            var html = await FilteredButtonHtml(actionParams.ContextId);
             var htmlElementContextOut = await driverService.InferElement(actionParams.Agent,
                 html,
                 actionParams.Context.ElementName,
                 actionParams.MessageId);
-            elements = Locator(actionParams.ConversationId, htmlElementContextOut);
+            elements = Locator(actionParams.ContextId, htmlElementContextOut);
 
             if (elements == null)
             {
@@ -54,7 +54,7 @@ public async Task<BrowserActionResult> ClickButton(BrowserActionParams actionPar
         try
         {
             await elements.ClickAsync();
-            await _instance.Wait(actionParams.ConversationId);
+            await _instance.Wait(actionParams.ContextId);
 
             result.IsSuccess = true;
         }
@@ -67,12 +67,12 @@ public async Task<BrowserActionResult> ClickButton(BrowserActionParams actionPar
         return result;
     }
 
-    private async Task<string> FilteredButtonHtml(string conversationId)
+    private async Task<string> FilteredButtonHtml(string contextId)
     {
         var driverService = _services.GetRequiredService<WebDriverService>();
 
         // Retrieve the page raw html and infer the element path
-        var body = await _instance.GetPage(conversationId).QuerySelectorAsync("body");
+        var body = await _instance.GetPage(contextId).QuerySelectorAsync("body");
 
         var str = new List<string>();
         /*var anchors = await body.QuerySelectorAllAsync("a");

diff --git a/...ns/BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.ClickElement.cs b/...ns/BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.ClickElement.cs
@@ -5,9 +5,9 @@ public partial class PlaywrightWebDriver
     public async Task<BrowserActionResult> ClickElement(BrowserActionParams actionParams)
     {
         var result = new BrowserActionResult();
-        await _instance.Wait(actionParams.ConversationId);
+        await _instance.Wait(actionParams.ContextId);
 
-        var page = _instance.GetPage(actionParams.ConversationId);
+        var page = _instance.GetPage(actionParams.ContextId);
         ILocator locator = default;
         int count = 0;
 
@@ -51,7 +51,7 @@ public async Task<BrowserActionResult> ClickElement(BrowserActionParams actionPa
             await locator.ClickAsync();
 
             // Triggered ajax
-            await _instance.Wait(actionParams.ConversationId);
+            await _instance.Wait(actionParams.ContextId);
 
             result.IsSuccess = true;
         }

diff --git a/...ns/BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.CloseBrowser.cs b/...ns/BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.CloseBrowser.cs
@@ -2,8 +2,8 @@ namespace BotSharp.Plugin.WebDriver.Drivers.PlaywrightDriver;
 
 public partial class PlaywrightWebDriver
 {
-    public async Task CloseBrowser(string conversationId)
+    public async Task CloseBrowser(string contextId)
     {
-        await _instance.Close(conversationId);
+        await _instance.Close(contextId);
     }
 }
diff --git a/...otSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.CloseCurrentPage.cs b/...otSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.CloseCurrentPage.cs
@@ -0,0 +1,9 @@
+namespace BotSharp.Plugin.WebDriver.Drivers.PlaywrightDriver;
+
+public partial class PlaywrightWebDriver
+{
+    public async Task CloseCurrentPage(string contextId)
+    {
+        await _instance.CloseCurrentPage(contextId);
+    }
+}
diff --git a/...lugins/BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.DoAction.cs b/...lugins/BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.DoAction.cs
@@ -1,12 +1,10 @@
-using BotSharp.Abstraction.Browsing.Enums;
-
 namespace BotSharp.Plugin.WebDriver.Drivers.PlaywrightDriver;
 
 public partial class PlaywrightWebDriver
 {
     public async Task DoAction(MessageInfo message, ElementActionArgs action, BrowserActionResult result)
     {
-        var page = _instance.GetPage(message.ConversationId);
+        var page = _instance.GetPage(message.ContextId);
         ILocator locator = page.Locator(result.Selector);
 
         if (action.Action == BroswerActionEnum.Click)

diff --git a/.../BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.EvaluateScript.cs b/.../BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.EvaluateScript.cs
@@ -2,10 +2,10 @@ namespace BotSharp.Plugin.WebDriver.Drivers.PlaywrightDriver;
 
 public partial class PlaywrightWebDriver
 {
-    public async Task<T> EvaluateScript<T>(string conversationId, string script)
+    public async Task<T> EvaluateScript<T>(string contextId, string script)
     {
-        await _instance.Wait(conversationId);
+        await _instance.Wait(contextId);
 
-        return await _instance.GetPage(conversationId).EvaluateAsync<T>(script);
+        return await _instance.GetPage(contextId).EvaluateAsync<T>(script);
     }
 }
diff --git a/...ins/BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.ExtractData.cs b/...ins/BotSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.ExtractData.cs
@@ -4,12 +4,12 @@ public partial class PlaywrightWebDriver
 {
     public async Task<string> ExtractData(BrowserActionParams actionParams)
     {
-        await _instance.Wait(actionParams.ConversationId);
+        await _instance.Wait(actionParams.ContextId);
 
         await Task.Delay(3000);
 
         // Retrieve the page raw html and infer the element path
-        var body = await _instance.GetPage(actionParams.ConversationId).QuerySelectorAsync("body");
+        var body = await _instance.GetPage(actionParams.ContextId).QuerySelectorAsync("body");
         var content = await body.InnerTextAsync();
 
         var driverService = _services.GetRequiredService<WebDriverService>();

diff --git a/...tSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.GetAttributeValue.cs b/...tSharp.Plugin.WebDriver/Drivers/PlaywrightDriver/PlaywrightWebDriver.GetAttributeValue.cs
@@ -4,7 +4,7 @@ public partial class PlaywrightWebDriver
 {
     public async Task<string> GetAttributeValue(MessageInfo message, ElementLocatingArgs location, BrowserActionResult result)
     {
-        var page = _instance.GetPage(message.ConversationId);
+        var page = _instance.GetPage(message.ContextId);
         ILocator locator = page.Locator(result.Selector);
         var value = string.Empty;