fixes + cleanup flow

README.md

@@ -7,73 +7,135 @@ sdk: docker
 pinned: false
 ---
 
-# Proxy-lite Salesforce Automation with Single-Browser Login
+# Proxy-lite Salesforce Automation
 
-This is a modernized proxy-lite Salesforce automation system that uses a streamlined single-browser login approach.
+A Salesforce automation system using Proxy-lite that works with any Salesforce org. The backend is already deployed and ready to use - you just need to add the Lightning Web Component to your org.
 
-## Key Features
+## ✅ What You Need
 
-✅ **Single-Browser Login**: No more cookie injection complexity
-✅ **Dual API Support**: Works with both Gemini and Convergence AI
-✅ **Automatic Authentication**: Handles Salesforce login seamlessly
-✅ **Efficient Architecture**: One browser instance for login and task execution
+- A Salesforce org (any org)
+- Salesforce CLI (sfdx)
+- Your Salesforce username and password
 
-## Quick Start
+## 🚀 Quick Setup (5 minutes)
 
-### Environment Setup
+### Step 1: Deploy the Lightning Web Component
 
-1. Install dependencies:
 ```bash
-pip install -r requirements.txt
+# Clone this repository
+git clone [your-repo-url]
+cd proxy-litev2\ copy
+
+# Deploy to your Salesforce org
+cd proxy-lite-work
+sfdx force:auth:web:login -a yourorg
+sfdx force:source:push -u yourorg
 ```
 
-2. Set environment variables:
-```bash
-# API Keys (set at least one)
-export GEMINI_API_KEY=your_gemini_api_key
-export HF_API_TOKEN=your_hf_token
+### Step 2: Add Component to Your Page
 
-# Salesforce Credentials
-export SALESFORCE_USERNAME=your_salesforce_username
-export SALESFORCE_PASSWORD=your_salesforce_password
-```
+1. Go to **Setup** → **Lightning App Builder**
+2. Edit any Lightning page (like Home page)
+3. Add the **"Salesforce Automation with Proxy Lite"** component
+4. Save and activate
 
-3. Run the Flask app:
-```bash
-python app.py
-```
+### Step 3: Use the Component
+
+1. Open the Lightning page where you added the component
+2. Select a target page from the dropdown
+3. Enter your automation task (e.g., "Enable the toggle switch")
+4. Click "Run Automation"
+
+**That's it!** The component will connect to the already-deployed backend automatically.
+
+## 🔧 How It Works
+
+- **Backend**: Already deployed on Hugging Face Spaces (no setup needed)
+- **Frontend**: Lightning Web Component in your Salesforce org
+- **Authentication**: Uses your Salesforce credentials automatically
+- **URLs**: Dynamically adapts to your org's domain
+
+## 📋 Supported Automation Tasks
+
+The system can automate various Salesforce setup pages:
+- Warranty Lifecycle Management
+- Account Forecasting Settings  
+- Sales Agreements
+- Account Manager Targets
+- Custom tasks on any setup page
+
+## 🛠️ Advanced Configuration (Optional)
 
-### Usage
+### Custom Endpoint (Enterprise Users)
 
-Send a POST request to `/run_proxy_task`:
+If you need your own private instance:
 
-```json
-{
-  "task": "Navigate to the permission set and assign it to the user",
-  "url": "https://your-org.salesforce.com/lightning/setup/AccountForecastSettings/home"
-}
+1. **Fork this repository**
+2. **Deploy to Hugging Face Spaces**:
+   - Create a new Space
+   - Connect your forked repo
+   - Set environment variables:
+     ```
+     GEMINI_API_KEY=your_key
+     SALESFORCE_USERNAME=your_username  
+     SALESFORCE_PASSWORD=your_password
+     ```
+
+3. **Update Apex Class**:
+   - Go to **Setup** → **Apex Classes** → **ProxyLiteService**
+   - Update the endpoint URL in `getProxyLiteEndpoint()` method
+
+### Environment Variables for Custom Deployment
+
+Only needed if deploying your own instance:
+
+```env
+# Required: At least one API key
+GEMINI_API_KEY=your_gemini_api_key
+HF_API_TOKEN=your_hf_token
+
+# Required: Your Salesforce credentials  
+[email protected]
+SALESFORCE_PASSWORD=your_salesforce_password
+
+# Optional: Browser settings
+BROWSER_HEADLESS=true
+BROWSER_VIEWPORT_WIDTH=1280
+BROWSER_VIEWPORT_HEIGHT=720
 ```
 
-## Architecture
+## 🔍 Troubleshooting
+
+### Common Issues
+
+1. **Component not showing**: Check if you've pushed the LWC to your org
+2. **"Endpoint not configured" error**: The Apex class needs the correct endpoint URL
+3. **Authentication errors**: Verify your Salesforce credentials work
+4. **Timeout errors**: Large tasks may take time - be patient
+
+### Debug Steps
+
+1. Check browser console for errors
+2. Verify the LWC is properly deployed
+3. Test with a simple task first
+4. Check Salesforce debug logs
 
-The system automatically:
-1. Initializes a browser session
-2. Performs Salesforce login using provided credentials
-3. Navigates to the target URL
-4. Executes the requested task
-5. Returns structured results
+## 💡 Why This Approach?
 
-## API Priority
+- **No Python setup needed** - Backend already deployed
+- **Works with any Salesforce org** - No hardcoded URLs
+- **Free to use** - Powered by Hugging Face Spaces
+- **Secure** - Uses your org's authentication
+- **Simple deployment** - Just add the LWC component
 
-- **Gemini API** (preferred if `GEMINI_API_KEY` is set)
-- **Convergence AI** (fallback if only `HF_API_TOKEN` is set)
+## 🆘 Support
 
-## Benefits over Cookie Injection
+If you encounter issues:
+1. Check the troubleshooting section above
+2. Verify your Salesforce org permissions
+3. Test with a different browser
+4. Check the component is properly added to your page
 
-- Eliminates cookie extraction/injection complexity
-- More efficient (no separate browser instance)
-- Better session management
-- More reliable navigation
-- Easier to maintain and debug
+## 📄 License
 
-See `SINGLE_BROWSER_LOGIN_GUIDE.md` for detailed documentation.
+This project is licensed under the MIT License.

app.py

@@ -28,6 +28,44 @@ app = Flask(__name__)
 
 _runner: Runner | None = None 
 
+def create_agent_task(target_url: str, request_task_instruction: str) -> str:
+    """Create the agent task with mandatory new tab navigation."""
+    return f"""
+CRITICAL FIRST STEP - MANDATORY: 
+Your VERY FIRST action must be to use the open_new_tab_and_go_to tool to navigate to {target_url}
+
+DO NOT skip this step. DO NOT use goto. You MUST use: open_new_tab_and_go_to(url='{target_url}')
+
+This is necessary because direct navigation to this URL gets stuck loading. The new tab approach bypasses this issue.
+
+STEP 1: Use open_new_tab_and_go_to(url='{target_url}')
+STEP 2: Wait for the page to be fully loaded (no loading spinners visible)
+STEP 3: {request_task_instruction}
+
+CRITICAL WORKFLOW - FOLLOW THESE EXACT STEPS IN SEQUENCE:
+
+STEP A: Select Permission Set
+- Use select_option_by_text tool to find and select the target permission set from Available list
+- Wait for "[ACTION COMPLETED]" response before proceeding
+
+STEP B: Click Add Button  
+- After successful selection, immediately click the "Add" button to move permission set to Enabled list
+- Do NOT repeat the selection - proceed directly to Add button
+
+STEP C: Click Save Button
+- After clicking Add, immediately click "Save" to persist the changes
+- After Save, Salesforce redirects to User page indicating SUCCESS
+
+CRITICAL: Do NOT repeat actions. Each step should happen exactly once in sequence.
+
+GENERAL INSTRUCTIONS:
+- You must EXECUTE all actions immediately - do NOT just describe what you plan to do
+- Do NOT wait for user input or ask "what should I do next?"
+- Complete the entire task autonomously using the available tools
+- After completing all steps, use the return_value tool to provide your final response
+- If you make a plan, IMMEDIATELY execute it step by step using the appropriate tools
+""" 
+
 async def initialize_runner_with_single_browser_login(username: str, password: str, target_url: str):
     """Initialize Proxy-lite Runner with single-browser login approach."""
     global _runner
@@ -147,46 +185,12 @@ async def run_proxy_task_endpoint():
                 return jsonify({"error": "Salesforce credentials not configured. Please set SALESFORCE_USERNAME and SALESFORCE_PASSWORD as Space secrets."}), 500
 
             runner = await initialize_runner_with_single_browser_login(salesforce_username, salesforce_password, target_url)
-            logger.info("Proxy-lite Runner initialized with cookies." if salesforce_username and salesforce_password else "Proxy-lite Runner initialized for general web browsing.")
+            logger.info("Proxy-lite Runner initialized with Salesforce login." if salesforce_username and salesforce_password else "Proxy-lite Runner initialized for general web browsing.")
 
             logger.info("Agent will use mandatory new tab tool to bypass loading issues.")
             
-            # MANDATORY new tab navigation task - this is critical to avoid loading issues
-            agent_task = f"""
-CRITICAL FIRST STEP - MANDATORY: 
-Your VERY FIRST action must be to use the open_new_tab_and_go_to tool to navigate to {target_url}
-
-DO NOT skip this step. DO NOT use goto. You MUST use: open_new_tab_and_go_to(url='{target_url}')
-
-This is necessary because direct navigation to this URL gets stuck loading. The new tab approach bypasses this issue.
-
-STEP 1: Use open_new_tab_and_go_to(url='{target_url}')
-STEP 2: Wait for the page to be fully loaded (no loading spinners visible)
-STEP 3: {request_task_instruction}
-
-CRITICAL WORKFLOW - FOLLOW THESE EXACT STEPS IN SEQUENCE:
-
-STEP A: Select Permission Set
-- Use select_option_by_text tool to find and select the target permission set from Available list
-- Wait for "[ACTION COMPLETED]" response before proceeding
-
-STEP B: Click Add Button  
-- After successful selection, immediately click the "Add" button to move permission set to Enabled list
-- Do NOT repeat the selection - proceed directly to Add button
-
-STEP C: Click Save Button
-- After clicking Add, immediately click "Save" to persist the changes
-- After Save, Salesforce redirects to User page indicating SUCCESS
-
-CRITICAL: Do NOT repeat actions. Each step should happen exactly once in sequence.
-
-GENERAL INSTRUCTIONS:
-- You must EXECUTE all actions immediately - do NOT just describe what you plan to do
-- Do NOT wait for user input or ask "what should I do next?"
-- Complete the entire task autonomously using the available tools
-- After completing all steps, use the return_value tool to provide your final response
-- If you make a plan, IMMEDIATELY execute it step by step using the appropriate tools
-            """
+            # Create the agent task using the centralized function
+            agent_task = create_agent_task(target_url, request_task_instruction)
             
             logger.info("Executing agent task with mandatory new tab navigation...")
             result = await runner.run(task=agent_task)
@@ -202,16 +206,15 @@ GENERAL INSTRUCTIONS:
                 "message": "Task completed successfully",
                 "data": {
                     "task_result": task_result,
-                    "steps_completed": [
-                        "Hardcoded Salesforce login completed",
-                        "Browser session initialized with cookies",
-                        "New tab navigation executed",
-                        "Target Salesforce setup page accessed",
-                        "Task execution completed successfully"
-                    ],
+                                    "steps_completed": [
+                    "Salesforce login completed",
+                    "Browser session initialized",
+                    "New tab navigation executed",
+                    "Target Salesforce setup page accessed",
+                    "Task execution completed successfully"
+                ],
                     "environment": {
                         "target_url": target_url,
-                        "cookies_count": 0, # No explicit cookie count for single-browser login
                         "navigation_method": "new_tab_bypass"
                     }
                 },
@@ -228,42 +231,8 @@ GENERAL INSTRUCTIONS:
 
             logger.info("Agent will use mandatory new tab tool to bypass loading issues.")
             
-            # MANDATORY new tab navigation task - this is critical to avoid loading issues
-            agent_task = f"""
-CRITICAL FIRST STEP - MANDATORY: 
-Your VERY FIRST action must be to use the open_new_tab_and_go_to tool to navigate to {target_url}
-
-DO NOT skip this step. DO NOT use goto. You MUST use: open_new_tab_and_go_to(url='{target_url}')
-
-This is necessary because direct navigation to this URL gets stuck loading. The new tab approach bypasses this issue.
-
-STEP 1: Use open_new_tab_and_go_to(url='{target_url}')
-STEP 2: Wait for the page to be fully loaded (no loading spinners visible)
-STEP 3: {request_task_instruction}
-
-CRITICAL WORKFLOW - FOLLOW THESE EXACT STEPS IN SEQUENCE:
-
-STEP A: Select Permission Set
-- Use select_option_by_text tool to find and select the target permission set from Available list
-- Wait for "[ACTION COMPLETED]" response before proceeding
-
-STEP B: Click Add Button  
-- After successful selection, immediately click the "Add" button to move permission set to Enabled list
-- Do NOT repeat the selection - proceed directly to Add button
-
-STEP C: Click Save Button
-- After clicking Add, immediately click "Save" to persist the changes
-- After Save, Salesforce redirects to User page indicating SUCCESS
-
-CRITICAL: Do NOT repeat actions. Each step should happen exactly once in sequence.
-
-GENERAL INSTRUCTIONS:
-- You must EXECUTE all actions immediately - do NOT just describe what you plan to do
-- Do NOT wait for user input or ask "what should I do next?"
-- Complete the entire task autonomously using the available tools
-- After completing all steps, use the return_value tool to provide your final response
-- If you make a plan, IMMEDIATELY execute it step by step using the appropriate tools
-            """
+            # Create the agent task using the centralized function
+            agent_task = create_agent_task(target_url, request_task_instruction)
             
             logger.info("Executing agent task with mandatory new tab navigation...")
             result = await runner.run(task=agent_task)
@@ -280,14 +249,13 @@ GENERAL INSTRUCTIONS:
                 "data": {
                     "task_result": task_result,
                     "steps_completed": [
-                        "Browser session initialized with cookies",
+                        "Browser session initialized",
                         "New tab navigation executed",
-                        "Target Salesforce setup page accessed",
+                        "Target page accessed",
                         "Task execution completed successfully"
                     ],
                     "environment": {
                         "target_url": target_url,
-                        "cookies_count": 0, # No explicit cookie count for single-browser login
                         "navigation_method": "new_tab_bypass"
                     }
                 },

proxy-lite-demo-v2/src/proxy_lite/environments/webbrowser.py

@@ -30,8 +30,6 @@ class WebBrowserEnvironmentConfig(BaseEnvironmentConfig):
     keep_original_image: bool = False
     no_pois_in_image: bool = False
     # --- MODIFICATION START ---
-    # Added to accept initial cookies from the RunnerConfig
-    initial_cookies: Optional[List[dict]] = None
     # Added for automatic login functionality
     perform_login: bool = False
     salesforce_login_url: Optional[str] = None
@@ -61,14 +59,6 @@ class WebBrowserEnvironment(BaseEnvironment):
         )
         await self.browser.__aenter__()
         # Initialize other resources if necessary
-        # --- MODIFICATION START ---
-        # Changed to use self.config.initial_cookies
-        if config.initial_cookies:  # type: ignore
-            if self.logger:
-                self.logger.info(f"🌐 [bold blue]Adding {len(config.initial_cookies)} initial cookies to browser context.[/]")  # type: ignore
-            if self.browser.context:
-                await self.browser.context.add_cookies(config.initial_cookies)  # type: ignore
-        # --- MODIFICATION END ---
         if self.logger:
             self.logger.info("🌐 [bold blue]Browser session started.[/]")
         return self
@@ -92,14 +82,9 @@ class WebBrowserEnvironment(BaseEnvironment):
     def browser_session(self) -> type[BrowserSession]:
         return BrowserSession
 
-    # --- MODIFICATION START ---
-    # Modified this property to return cookies from the config.
-    # It was previously hardcoded to return an empty list.
     @property
     def cookies(self) -> list[dict]:
-        config = self.config  # type: WebBrowserEnvironmentConfig
-        return config.initial_cookies if config.initial_cookies is not None else []  # type: ignore
-    # --- MODIFICATION END ---
+        return []
     
     async def initialise(self) -> Observation:
         if self.browser is None:
@@ -160,16 +145,7 @@ class WebBrowserEnvironment(BaseEnvironment):
                 
                 if self.logger:
                     self.logger.info(f"🔑 Login completed successfully. Current URL: {self.browser.current_url}")
-                
-                # Navigate to target URL if specified
-                if config.target_url:  # type: ignore
-                    if self.logger:
-                        self.logger.debug(f"🔑 Navigating to target URL: {config.target_url}")  # type: ignore
-                    await self.browser.goto(config.target_url)  # type: ignore
-                    if self.browser.current_page:
-                        await self.browser.current_page.wait_for_load_state('networkidle', timeout=30000)
-                    if self.logger:
-                        self.logger.info(f"🔑 Successfully navigated to target URL: {self.browser.current_url}")
+                    self.logger.info("🔑 Login process complete - agent will handle navigation using open_new_tab_and_go_to")
                 
             except Exception as e:
                 if self.logger:

proxy-lite-demo-v2/src/proxy_lite/runner.py

@@ -122,6 +122,10 @@ class Runner(BaseModel):
                 self.solver(config=self.config.solver, logger=self.logger) as solver,
             ):
                 run.env_info = await environment.get_info()
+                # Initialize environment first (includes login) before solver gets control
+                initial_observation: Observation = await environment.initialise()
+                self.logger.debug("Environment initialised.")
+                # Only after successful environment initialization (login) should solver get initialized
                 await solver.initialise(
                     task,
                     environment.tools,
@@ -129,9 +133,7 @@ class Runner(BaseModel):
                 )
                 self.logger.debug("Solver initialised.")
                 run.solver_history = solver.history  # type: ignore
-                initial_observation: Observation = await environment.initialise()
                 await event_queue.put(initial_observation)
-                self.logger.debug("Environment initialised.")
                 step_count = 0
                 while step_count < self.config.max_steps:
                     event = await event_queue.get()

proxy-lite-work/force-app/main/default/lwc/lwcProxyLiteAutomation/lwcProxyLiteAutomation.js

@@ -36,13 +36,22 @@ export default class LwcProxyLiteAutomation extends LightningElement {
         }
     ];
 
-    // URL mapping for different pages
-    pageUrls = {
-        'warranty_lifecycle': 'https://dwd000006jia1mae.lightning.force.com/lightning/setup/MfgServiceSettings/home',
-        'account_forecasting': 'https://dwd000006jia1mae.lightning.force.com/lightning/setup/AccountForecastSettings/home',
-        'sales_agreements': 'https://dwd000006jia1mae.lightning.force.com/lightning/setup/SalesAgreementSettings/home',
-        'account_manager_targets': 'https://dwd000006jia1mae.lightning.force.com/lightning/setup/AcctMgrTargetSettings/home'
-    };
+    // Dynamic URL generation based on current org
+    get currentOrgDomain() {
+        // Get the current org's domain dynamically
+        return window.location.origin;
+    }
+
+    // Dynamic URL mapping for different pages
+    get pageUrls() {
+        const baseUrl = this.currentOrgDomain;
+        return {
+            'warranty_lifecycle': `${baseUrl}/lightning/setup/MfgServiceSettings/home`,
+            'account_forecasting': `${baseUrl}/lightning/setup/AccountForecastSettings/home`,
+            'sales_agreements': `${baseUrl}/lightning/setup/SalesAgreementSettings/home`,
+            'account_manager_targets': `${baseUrl}/lightning/setup/AcctMgrTargetSettings/home`
+        };
+    }
 
     // Handles changes in the page selection dropdown
     handlePageChange(event) {
@@ -56,7 +65,7 @@ export default class LwcProxyLiteAutomation extends LightningElement {
     }
 
     // Handles the click event on the "Run Automation" button
-    async handleRunAutomation() {
+    handleRunAutomation() {
         // Basic input validation
         if (!this.selectedPage) {
             this.showToast('Error', 'Please select a target page.', 'error');
@@ -71,16 +80,15 @@ export default class LwcProxyLiteAutomation extends LightningElement {
         this.isProcessing = true; // Show spinner, disable button
         this.clearResults(); // Clear previous results
 
-        try {
-            // Get the URL for the selected page
-            const targetUrl = this.pageUrls[this.selectedPage];
-            
-            // Call the Apex method with both task and URL
-            const result = await runProxyLiteTask({ 
-                userTask: this.taskInput, 
-                targetUrl: targetUrl 
-            });
-            
+        // Get the URL for the selected page
+        const targetUrl = this.pageUrls[this.selectedPage];
+        
+        // Call the Apex method with both task and URL using .then() instead of async/await
+        runProxyLiteTask({ 
+            userTask: this.taskInput, 
+            targetUrl: targetUrl 
+        })
+        .then(result => {
             // Parse the result if it's a JSON string
             let parsedResult;
             if (typeof result === 'string') {
@@ -114,8 +122,8 @@ export default class LwcProxyLiteAutomation extends LightningElement {
                 
                 this.showToast('Error', 'Automation task failed.', 'error');
             }
-            
-        } catch (error) {
+        })
+        .catch(error => {
             // Handle errors from the Apex callout
             console.error('Proxy Lite Automation Error:', error);
             
@@ -149,9 +157,10 @@ export default class LwcProxyLiteAutomation extends LightningElement {
             }
             
             this.showToast('Error', 'Failed to run automation task.', 'error');
-        } finally {
+        })
+        .finally(() => {
             this.isProcessing = false; // Hide spinner, enable button
-        }
+        });
     }
 
     // Helper method to clear results