Update web driver to handle mutiple tasks per agent

This commit is contained in:
Visagan Guruparan 2025-03-19 15:59:11 -05:00
parent 24657bd3b5
commit d6672cbddd
20 changed files with 50 additions and 28 deletions

View file

@ -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";
}

View file

@ -34,7 +34,7 @@ public class ChangeCheckboxFn : IFunctionCallback
message.Data = await _browser.ScreenshotAsync(new MessageInfo
{
AgentId = message.CurrentAgentId,
ContextId = convService.ConversationId,
ContextId = webDriverService.GetMessageContext(message),
MessageId = message.MessageId
}, path);

View file

@ -34,7 +34,7 @@ public class ChangeListValueFn : IFunctionCallback
message.Data = await _browser.ScreenshotAsync(new MessageInfo
{
AgentId = message.CurrentAgentId,
ContextId = convService.ConversationId,
ContextId = webDriverService.GetMessageContext(message),
MessageId = message.MessageId
}, path);

View file

@ -34,7 +34,7 @@ public class CheckRadioButtonFn : IFunctionCallback
message.Data = await _browser.ScreenshotAsync(new MessageInfo
{
AgentId = message.CurrentAgentId,
ContextId = convService.ConversationId,
ContextId = webDriverService.GetMessageContext(message),
MessageId = message.MessageId
}, path);

View file

@ -34,7 +34,7 @@ public class ClickButtonFn : IFunctionCallback
message.Data = await _browser.ScreenshotAsync(new MessageInfo
{
AgentId = message.CurrentAgentId,
ContextId = convService.ConversationId,
ContextId = webDriverService.GetMessageContext(message),
MessageId = message.MessageId
}, path);

View file

@ -34,7 +34,7 @@ public class ClickElementFn : IFunctionCallback
message.Data = await _browser.ScreenshotAsync(new MessageInfo
{
AgentId = message.CurrentAgentId,
ContextId = convService.ConversationId,
ContextId = webDriverService.GetMessageContext(message),
MessageId = message.MessageId
}, path);

View file

@ -28,7 +28,7 @@ public class ExtractDataFn : IFunctionCallback
message.Data = await _browser.ScreenshotAsync(new MessageInfo
{
AgentId = message.CurrentAgentId,
ContextId = convService.ConversationId,
ContextId = webDriverService.GetMessageContext(message),
MessageId = message.MessageId
}, path);

View file

@ -31,7 +31,7 @@ public class GoToPageFn : IFunctionCallback
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 class GoToPageFn : IFunctionCallback
message.Data = await _browser.ScreenshotAsync(new MessageInfo
{
AgentId = message.CurrentAgentId,
ContextId = convService.ConversationId,
ContextId = webDriverService.GetMessageContext(message),
MessageId = message.MessageId
}, path);

View file

@ -1,3 +1,5 @@
using BotSharp.Plugin.WebDriver.Services;
namespace BotSharp.Plugin.WebDriver.Functions;
public class HttpRequestFn : IFunctionCallback
@ -20,12 +22,13 @@ public class HttpRequestFn : IFunctionCallback
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 ?

View file

@ -33,7 +33,7 @@ public class InputUserPasswordFn : IFunctionCallback
message.Data = await _browser.ScreenshotAsync(new MessageInfo
{
AgentId = message.CurrentAgentId,
ContextId = convService.ConversationId,
ContextId = webDriverService.GetMessageContext(message),
MessageId = message.MessageId
}, path);

View file

@ -39,7 +39,7 @@ public class InputUserTextFn : IFunctionCallback
message.Data = await _browser.ScreenshotAsync(new MessageInfo
{
AgentId = message.CurrentAgentId,
ContextId = convService.ConversationId,
ContextId = webDriverService.GetMessageContext(message),
MessageId = message.MessageId
}, path);

View file

@ -31,7 +31,7 @@ public class OpenBrowserFn : IFunctionCallback
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 class OpenBrowserFn : IFunctionCallback
message.Data = await _browser.ScreenshotAsync(new MessageInfo
{
AgentId = message.CurrentAgentId,
ContextId = convService.ConversationId,
ContextId = webDriverService.GetMessageContext(message),
MessageId = message.MessageId
}, path);

View file

@ -24,7 +24,7 @@ public class ScreenshotFn : IFunctionCallback
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.";

View file

@ -20,12 +20,13 @@ public class ScrollPageFn : IFunctionCallback
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 class ScrollPageFn : IFunctionCallback
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);

View file

@ -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;
}
}
}

View file

@ -35,17 +35,17 @@ public class UtilWebActionOnElementFn : IFunctionCallback
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);

View file

@ -18,20 +18,19 @@ public class UtilWebCloseBrowserFn : IFunctionCallback
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);

View file

@ -29,13 +29,13 @@ public class UtilWebGoToPageFn : IFunctionCallback
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 class UtilWebGoToPageFn : IFunctionCallback
message.Content = $"Open web page successfully.";
var webDriverService = _services.GetRequiredService<WebDriverService>();
var path = webDriverService.GetScreenshotFilePath(message.MessageId);
message.Data = await browser.ScreenshotAsync(msg, path);

View file

@ -22,17 +22,17 @@ public class UtilWebLocateElementFn : IFunctionCallback
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);

View file

@ -23,7 +23,7 @@ public class WebDriverPlugin : IBotSharpPlugin
var settingService = provider.GetRequiredService<ISettingService>();
return settings;
});
services.AddSingleton<WebBrowsingSettings>();
services.AddScoped<PlaywrightWebDriver>();
services.AddSingleton<PlaywrightInstance>();