@uipath/skills 1.195.0

package/skills/uipath-rpa/references/legacy/activity-docs/_COMMON-PITFALLS.md

@@ -0,0 +1,296 @@
+# UiPath Legacy Activities - Common Real-World Pitfalls
+
+## Purpose
+Gotchas reported by the community and discovered in source code that go beyond typical documentation. These are the issues developers hit in production.
+
+---
+
+## Excel (Classic - ExcelApplicationScope)
+
+### Zombie Excel Processes
+- **Problem**: Invisible `EXCEL.EXE` processes accumulate, especially after exceptions mid-workflow
+- **Root cause**: COM objects not released when workflow crashes before Dispose runs
+- **Fix**: Use Kill Process for `EXCEL.EXE` in a Finally block. The codebase has 207+ `DisposeWithReleaseComObject` calls showing this is a known hard problem
+
+### Dates Read as Serial Numbers (45678 instead of date)
+- **Problem**: Read Range returns OLE Automation doubles instead of dates
+- **Root cause**: Excel stores dates internally as doubles since 1900. `PreserveFormat=false` returns raw numbers
+- **Fix**: Set `PreserveFormat=true`, or convert with `DateTime.FromOADate(doubleValue)`
+
+### Read Range Returns Empty DataTable
+- **Problem**: Activity succeeds but DataTable is empty
+- **Root cause**: Sheet name mismatch (case/whitespace), malformed range, or file locked by open Excel
+- **Fix**: Verify sheet name with Get Workbook Sheets first. Use `""` for range to read entire used range. Close Excel before running
+
+### Write Range Overwrites Formatting
+- **Problem**: Cell formatting (colors, fonts, borders) lost after Write Range
+- **Root cause**: Classic Write Range strips formatting by default
+- **Fix**: Use Write Cell in loop for small updates, or use Interop-based writing with Visible=true
+
+### ClosedXML Formula Evaluation is Limited (Portable)
+- **Problem**: Formulas return stale/wrong values in Portable mode
+- **Root cause**: ClosedXML cannot evaluate all Excel formulas; falls back to cached values via GetRichText()
+- **Fix**: Use Interop mode (ExcelApplicationScope) for formula-dependent workflows
+
+---
+
+## UIAutomation (Classic - Click, TypeInto, Selectors)
+
+### TypeInto Missing or Wrong Characters
+- **Problem**: Characters dropped, wrong, or special chars interpreted as modifiers
+- **Root cause**: `{`, `}`, `[`, `]`, `+`, `^`, `%`, `~` are SendKeys modifier chars. SimulateType may be too fast.
+- **Fix**: Escape special chars: `{{}`, `{}}`, `{+}`. Add `DelayBetweenKeys` (50ms). For passwords use Type Secure Text
+
+### TypeInto at Wrong Cursor Position
+- **Problem**: Text inserted in middle of existing content instead of replacing
+- **Root cause**: EmptyField may not work on all controls. Cursor position unpredictable.
+- **Fix**: Set `EmptyField=true` AND prepend `[k(ctrl+a)]` to select all before typing
+
+### EmptyField Silently Ignored with SimulateType
+- **Problem**: Field not cleared despite EmptyField=true
+- **Root cause**: Source code only runs EmptyField sequence for hardware events and SendWindowMessages, NOT for SimulateType/API mode
+- **Fix**: Use hardware events (default) or SendWindowMessages when EmptyField is needed
+
+### Selector Works in Studio, Fails on Robot/Orchestrator
+- **Problem**: Element not found in production despite working in development
+- **Root cause**: Different resolution, DPI scaling, user session (interactive vs service), window state. RDP disconnect loses UI.
+- **Fix**: Use SimulateClick/SimulateType (don't depend on screen). Keep RDP alive. Use consistent resolution. Avoid `idx` attribute.
+
+### Dynamic Selectors Break Over Time
+- **Problem**: Selectors with timestamps, session IDs, or changing indices fail
+- **Root cause**: Dynamic attributes like `aaname` containing runtime values, `idx` being position-dependent
+- **Fix**: Use wildcards (`*`) for dynamic parts. Prefer `AutomationId` or `name`. Use Anchor Base for relative positioning. Avoid `idx`.
+
+### OpenBrowser Defaults to Internet Explorer
+- **Problem**: Workflows default to IE which is deprecated
+- **Root cause**: `BrowserType` default is `BrowserType.IE` in source code
+- **Fix**: Always explicitly set BrowserType to Chrome, Firefox, or Edge
+
+---
+
+## Mail (Classic - SMTP/IMAP/POP3/Exchange/Outlook)
+
+### SMTP Authentication Fails with Gmail/Microsoft 365
+- **Problem**: Login rejected despite correct credentials
+- **Root cause**: Google/Microsoft disabled "Less Secure Apps". OAuth2 now required.
+- **Fix**: Use App Passwords (Gmail) or configure Azure AD App Registration with SMTP.Send permission (M365). Set UseOAuth=true and pass OAuth token as Password.
+
+### SMTP SSL/TLS Port Mismatch
+- **Problem**: Connection fails with SSL errors
+- **Root cause**: Port 587 requires STARTTLS (`SecureConnection=StartTls`). Port 465 requires implicit SSL. Port 25 is unencrypted.
+- **Fix**: Use Port 587 + StartTls for most SMTP servers. Classic SMTP activity may not support implicit SSL on port 465.
+
+### Attachment Paths Not Found on Robot
+- **Problem**: Attachment path valid in Studio but not on production Robot
+- **Root cause**: Relative paths resolve differently on Robot vs Studio
+- **Fix**: Always use absolute paths. Use `Path.Combine(Environment.CurrentDirectory, filename)`
+
+### Addresses Must Use Semicolons (Not Commas)
+- **Problem**: Multiple recipients not receiving emails
+- **Root cause**: Address separator is `;` not `,` for all protocols EXCEPT Lotus Notes (which uses commas)
+- **Fix**: `"user1@mail.com;user2@mail.com"` - semicolons only
+
+---
+
+## Web (Classic - HTTP Request, SOAP)
+
+### HTTP Request Returns Error Despite Correct Setup
+- **Problem**: 403/401 errors with valid credentials
+- **Root cause**: Missing required headers (`User-Agent`, `Accept`), expired token, or wrong `Content-Type`
+- **Fix**: Add `User-Agent` header. Match `Content-Type` exactly. Verify with Postman first, replicate exact headers.
+
+### HTTP Request ContinueOnError Defaults to TRUE
+- **Problem**: HTTP errors silently swallowed; workflow continues with null/empty response
+- **Root cause**: `NetHttpRequest.ContinueOnError = true` by default in source code
+- **Fix**: Explicitly set `ContinueOnError=false` to get exceptions on 4xx/5xx responses
+
+### Legacy HttpClient Very Short Timeout (6 seconds)
+- **Problem**: Requests timeout on slow APIs
+- **Root cause**: Default timeout is only 6,000ms. NetHttpRequest default is 10,000ms. Both often too low.
+- **Fix**: Increase TimeoutMS to 30,000-60,000ms for production APIs
+
+### SOAP Client Timeout Not Configurable
+- **Problem**: SOAP requests timeout at 60 seconds with no way to increase
+- **Root cause**: Hardcoded `60,000ms` constant in HttpSoapRequestLogic, no exposed property
+- **Fix**: No workaround via activity properties. For long-running SOAP calls, consider using HTTP Request with manual SOAP envelope construction.
+
+---
+
+## PDF (Classic)
+
+### Read PDF Text Returns Empty String
+- **Problem**: Activity returns `""` for a PDF that clearly has content
+- **Root cause**: PDF contains scanned images, not text layer. `ReadPDFText` only extracts native text.
+- **Fix**: Use `Read PDF With OCR` with an OCR engine (Tesseract, UiPath Document OCR). Check if text is selectable in Adobe Reader.
+
+### Read PDF Text Returns Garbled/Out-of-Order Text
+- **Problem**: Text is present but columns are mixed, lines out of order
+- **Root cause**: PDF internal text stream doesn't match visual layout. Multi-column layouts especially problematic.
+- **Fix**: Set `PreserveFormatting=true` (off by default!). Or use Read PDF With OCR for better layout handling.
+
+---
+
+## Database (Community)
+
+### "Unable to find requested .NET Framework Data Provider"
+- **Problem**: Connection fails with provider not found error
+- **Root cause**: Missing database driver, or 32-bit vs 64-bit mismatch with UiPath Robot
+- **Fix**: Install correct bitness driver (64-bit for modern UiPath). For Oracle use `Oracle.ManagedDataAccess` NuGet. Check `machine.config` for provider registration.
+
+### Connection Works in Studio, Fails on Robot
+- **Problem**: Same connection string fails in production
+- **Root cause**: Different ODBC/OLE DB drivers installed, Windows credentials differ (Integrated Security)
+- **Fix**: Use explicit credentials, not `Integrated Security=true`. Ensure drivers installed on Robot machine.
+
+### Execute Query vs Execute Non Query Confusion
+- **Problem**: SELECT returns no data, or INSERT returns DataTable
+- **Root cause**: Using wrong activity for the SQL operation type
+- **Fix**: `Execute Query` for SELECT (returns DataTable). `Execute Non Query` for INSERT/UPDATE/DELETE (returns row count).
+
+---
+
+## Credentials
+
+### Get Credential Fails on Robot Server
+- **Problem**: Works on developer machine but fails on production Robot
+- **Root cause**: Windows Credential Manager is per-user. Robot service account doesn't have the stored credential.
+- **Fix**: Use Orchestrator Assets (Credential type) instead. If must use Windows CM, log in as Robot service account and store credential there.
+
+---
+
+## Python (Community - PythonScope)
+
+### Python Scope Fails to Find Installation
+- **Problem**: "Python not found" error
+- **Root cause**: Wrong Path or Version. 32-bit vs 64-bit mismatch. TargetPlatform must match Python installation.
+- **Fix**: Set full path to `python.exe`. Match Version exactly. Set TargetPlatform to x64 for 64-bit Python.
+
+### Python Script Works Locally but Fails in UiPath
+- **Problem**: Script runs fine in command line but errors in PythonScope
+- **Root cause**: Virtual environment not activated, working directory different, packages not installed globally
+- **Fix**: Point Path to virtualenv's `python.exe`. Use absolute paths in scripts. Install packages globally or in the specific env.
+
+---
+
+## FTP (Community)
+
+### Download/Upload Hangs or Timeouts
+- **Problem**: Connection established but transfers hang
+- **Root cause**: Passive vs Active mode mismatch; firewall blocking data channel ports
+- **Fix**: Toggle UsePassiveMode. Ensure firewall allows passive port range (1024-65535). Use SFTP (port 22) when possible.
+
+---
+
+## GenericValue Traps (System)
+
+### String Comparison Instead of Numeric
+- **Problem**: `"10" > "9"` returns False
+- **Root cause**: GenericValue does string comparison (lexicographic), not numeric
+- **Fix**: Explicitly convert: `CInt(genericValue)`, `CDbl(genericValue)`. Avoid GenericValue in new projects; use strongly-typed variables.
+
+### DataTable Column Name Issues
+- **Problem**: "Column X does not belong to table" after reading data
+- **Root cause**: Column names have trailing/leading whitespace, or case mismatch in filter expressions
+- **Fix**: Trim column names: `col.ColumnName = col.ColumnName.Trim()`. Use column index `row(0)` when names unreliable. Use bracket notation for spaces: `[Column Name]`.
+
+### GenericValue Boolean Conversion Trap (DANGEROUS)
+- **Problem**: `new GenericValue("hello")` converted to bool returns `True`. ANY non-null, non-empty, non-boolean string → `True`
+- **Root cause**: Source code `GenericValue.cs` line 89: if `Convert.ToBoolean` fails, the catch block returns `true` (not false!)
+- **Fix**: NEVER rely on implicit GenericValue-to-bool conversion. Always explicitly compare: `If myVar.ToString() = "True"`. Better yet, avoid GenericValue entirely and use strongly-typed variables.
+
+### GenericValue Null Conversion Traps
+- **Problem**: Silent data loss on null values
+- **Root cause**: `GenericValue(null)` → int returns `0`, → DateTime returns `DateTime.MinValue` (Jan 1, 0001), → double returns `0.0`. No errors thrown.
+- **Fix**: Always check for null/empty before converting: `If Not String.IsNullOrEmpty(myGenericValue.ToString())`
+
+---
+
+## HIDDEN DANGEROUS DEFAULTS (Source Code Verified)
+
+### Activities That Default ContinueOnError=TRUE (silent failures)
+These activities SWALLOW all errors by default. Your workflow will continue as if they succeeded:
+
+| Activity | Package | Impact |
+|----------|---------|--------|
+| `NetHttpRequest` (HTTP Request) | Web | HTTP 500/timeout → empty response, no error |
+| `Data Scraping` wizard output | UIAutomation | Extraction failure → empty DataTable |
+| `ShowCallout` | Forms | Callout display failure → silently ignored |
+
+**Fix**: Always explicitly set `ContinueOnError=False` on HTTP Request activities unless you're checking status codes manually.
+
+### Excel AutoSave=TRUE Causes Performance Disasters in Loops
+- **Problem**: 1000 Write Cell operations = 1000 disk writes
+- **Root cause**: `AutoSave=true` by default on ExcelApplicationScope. Every single activity triggers `Document.Save()`
+- **Fix**: Set `AutoSave=false` on ExcelApplicationScope, then add a single Save Workbook at the end. The `ExcelForEachRowX` activity internally disables AutoSave during iteration for this exact reason.
+
+### Excel DisplayAlerts/ScreenUpdating Left in Bad State After Crash
+- **Problem**: After workflow crash, Excel stays in "frozen" state or silently overwrites files
+- **Root cause**: Operations temporarily set `DisplayAlerts=false` and `ScreenUpdating=false` on the COM app. Crash before restore leaves Excel broken.
+- **Fix**: Use Try-Finally to ensure Excel state is restored. Consider Kill Process as cleanup.
+
+### CSV Encoding Varies by Platform (Encoding.Default trap)
+- **Problem**: Same CSV file produces different results on different machines
+- **Root cause**: `Encoding.Default` returns system ANSI code page on .NET Framework but UTF-8 on .NET Core. Locale-dependent.
+- **Fix**: Always explicitly specify encoding: `"UTF-8"` in CSV activities. Never rely on defaults.
+
+### CSV Delimiter Depends on System Culture
+- **Problem**: CSV activities produce different delimiters on different machines
+- **Root cause**: `CultureInfo.CurrentUICulture` used for CSV configuration internally
+- **Fix**: Always explicitly set the `Delimitator` (note spelling) property. Don't rely on system defaults.
+
+---
+
+## DATE/LOCALE TRAPS
+
+### Date Formats Cached from Windows Registry for Entire Process
+- **Problem**: Date parsing/formatting uses stale format if Windows settings change
+- **Root cause**: Short/long date formats read from `HKCU\Control Panel\International` and cached in static fields for process lifetime (never refreshed)
+- **Fix**: Don't change date format settings during Robot execution. If needed, restart Robot process.
+
+### Outlook Mail Filter Depends on Windows Date Format
+- **Problem**: Outlook date filters silently return no results
+- **Root cause**: Filter reads `sShortDate` from registry. If your filter date format doesn't match Windows-configured short date format, Outlook filtering fails silently.
+- **Fix**: Format dates in Outlook filter expressions to match the system's short date format.
+
+---
+
+## PROCESS/MEMORY SAFETY
+
+### COM Singleton Reference Counting (Excel, Word, PowerPoint)
+- **Problem**: One crashed activity corrupts COM state for all subsequent activities in same process
+- **Root cause**: All three Office apps use process-wide singleton `ComAppReferenceCountManager`. Delayed cleanup via `Task.Factory.StartNew` can race with new scope creation. Falls back to `Process.Kill()` as last resort.
+- **Fix**: Wrap all Office scope activities in Try-Finally. Use Kill Process as safety net. Don't run multiple Excel/Word/PowerPoint scopes concurrently.
+
+### Mail Session Server Orphan Processes
+- **Problem**: `UiPath.Mail.MainSessionServer.exe` runs forever after Robot disconnects
+- **Root cause**: Server uses `await Task.Delay(int.MaxValue)` in infinite loop. Only terminates if parent PID argument is passed correctly.
+- **Fix**: Monitor for orphaned MainSessionServer.exe processes. Kill after workflow completion if needed.
+
+### Regex with Infinite Timeout
+- **Problem**: Complex regex patterns can hang workflow indefinitely
+- **Root cause**: `System.Activities` Regex Replace uses `Regex.InfiniteMatchTimeout` when TimeoutMS not specified
+- **Fix**: Always set TimeoutMS on Regex activities. Test regex patterns for catastrophic backtracking.
+
+### Terminal Infinite Wait on Unresponsive Server
+- **Problem**: Workflow hangs forever waiting for terminal response
+- **Root cause**: Terminal operations use `Timeout.Infinite` when no timeout specified in 3 code paths
+- **Fix**: Always set explicit TimeoutMS on all terminal activities. Never rely on defaults.
+
+---
+
+## SILENT ERROR SWALLOWING (Design-Time)
+
+### HTTP Request Designer Hides Configuration Errors
+- **Problem**: Incorrectly configured HTTP requests show no warnings at design time
+- **Root cause**: Four separate empty `catch { }` blocks in HttpRequestWindow.xaml.cs silently eat configuration errors
+- **Fix**: Always test HTTP requests manually (e.g., in Postman) before relying on the designer configuration.
+
+### Connection Service Resolution Silently Fails
+- **Problem**: Mail connection falls back to null ConnectionId without warning
+- **Root cause**: Empty `catch (Exception) { }` in BaseMailConnectionServiceActivities when resolving connection bindings
+- **Fix**: Test connection resolution explicitly in a separate workflow step before using in production.
+
+### Office365 Token Storage Silently Fails
+- **Problem**: Token refresh/storage operations fail without any indication
+- **Root cause**: OrchestratorGlobalAssetStore catches and ignores HttpClientFramework.ApiServiceException on delete, get, list, and folder operations
+- **Fix**: Monitor Orchestrator logs for token-related errors. Test auth flow end-to-end.

package/skills/uipath-rpa/references/legacy/activity-docs/_DU-PROCESS.md

@@ -0,0 +1,340 @@
+# UiPath Document Understanding Process Template - Complete Reference
+
+## Overview
+The Document Understanding Process is a **fully functional Studio template** for processing documents through a complete pipeline: digitize, classify, validate classification, extract data, validate extraction, train models, and export results. It comes with built-in retry logic, error handling, multi-robot synchronization, and Action Center integration.
+
+**Source**: [UiPath Docs](https://docs.uipath.com/document-understanding/automation-suite/2024.10/classic-user-guide/du-process) | [Forum Announcement](https://forum.uipath.com/t/document-understanding-process-new-studio-template/330234)
+
+**Key Design**: One job per document (not bulk). Requires a **dispatcher** to feed queue items.
+
+---
+
+## 1. File Structure (from Desktop Template)
+
+```
+DocumentUnderstandingProcess/
+├── Main-ActionCenter.xaml              # Entry point for UNATTENDED (Action Center validation)
+├── Main-Attended.xaml                  # Entry point for ATTENDED (local Validation Station)
+├── project.json                        # Project config (supportsPersistence=true!)
+│
+├── Data/
+│   ├── Config.xlsx                     # Configuration (Settings/Constants/Assets)
+│   └── ExampleDocuments/
+│       └── MergedDocuments.pdf         # Sample multi-page document
+│
+├── DocumentProcessing/
+│   ├── taxonomy.json                   # Document taxonomy definition
+│   └── IntelligentKeywordLearningFile.json  # Classifier training data
+│
+├── Framework/                          # Processing pipeline (numbered for execution order)
+│   ├── 00_ReadConfigFile.xaml          # Read Config.xlsx → Dictionary
+│   ├── 10_InitializeProcess.xaml       # Load taxonomy + Orchestrator assets
+│   ├── 15_GetTransactionItem.xaml      # Get queue item + extract TargetFile
+│   ├── 20_Digitize.xaml               # OCR → Document Object Model + Text
+│   ├── 30_Classify.xaml               # Auto-classify document type
+│   ├── 35_ClassificationBussinessRuleValidation.xaml  # Validate classification
+│   ├── 40_TrainClassifiers.xaml        # Train classifier with validated data
+│   ├── 50_Extract.xaml                # Extract field values
+│   ├── 55_ExtractionBussinessRuleValidation.xaml      # Validate extraction
+│   ├── 60_TrainExtractors.xaml         # Train extractor with validated data
+│   ├── 70_Export.xaml                 # Export results (Excel example)
+│   ├── 80_EndProcess.xaml             # Mark transaction successful
+│   ├── ERR_AbortProcess.xaml           # Handle fatal process errors
+│   ├── ERR_HandleDocumentError.xaml    # Handle per-document errors
+│   └── ReusableWorkflows/
+│       ├── SetTransactionStatus.xaml   # Set queue item Success/Failed
+│       ├── SetTransactionProgress.xaml # Update progress in Orchestrator
+│       ├── InvoicePostProcessing.xaml  # Invoice-specific validation rules
+│       ├── LockFile.xaml              # File lock for multi-robot sync
+│       ├── UnlockFile.xaml            # File unlock
+│       ├── GetWritePermission.xaml     # Queue-based semaphore acquire
+│       └── GiveUpWritePermission.xaml  # Queue-based semaphore release
+│
+├── UserGuide/
+│   └── Document Understanding Process - User Guide.pdf
+│
+└── Exceptions_Screenshots/             # (optional) Error screenshots
+```
+
+---
+
+## 2. Two Entry Points: Attended vs Action Center
+
+| | Main-Attended.xaml | Main-ActionCenter.xaml |
+|---|-------------------|----------------------|
+| **`in_UseQueue`** | `False` (default) | `True` (default) |
+| **Validation** | Local Validation Station (desktop UI) | Action Center (Orchestrator web UI) |
+| **Persistence** | Not required | `supportsPersistence=true` required |
+| **Robot Type** | Attended or unattended | Unattended (long-running) |
+| **Human-in-loop** | User sees validation dialog | User completes tasks in Action Center |
+| **Queue** | Optional | Required |
+
+### Main.xaml Arguments
+| Argument | Type | Default | Description |
+|----------|------|---------|-------------|
+| `in_TargetFile` | InArgument\<String\> | | Path to document to process |
+| `in_UseQueue` | InArgument\<Boolean\> | True/False | Whether using Orchestrator queues |
+
+---
+
+## 3. Processing Pipeline (Step by Step)
+
+```
+00_ReadConfigFile → 10_Initialize → 15_GetTransactionItem
+                                          ↓
+    ┌─────── FOR EACH CLASSIFIED DOCUMENT ────────┐
+    │                                               │
+    │  20_Digitize → 30_Classify                   │
+    │       ↓                                       │
+    │  35_ClassificationValidation                  │
+    │       ↓ (if manual needed)                    │
+    │  [Classification Station / Action Center]     │
+    │       ↓                                       │
+    │  40_TrainClassifiers (with validated data)    │
+    │       ↓                                       │
+    │  50_Extract → 55_ExtractionValidation         │
+    │       ↓ (if manual needed)                    │
+    │  [Validation Station / Action Center]         │
+    │       ↓                                       │
+    │  60_TrainExtractors (with validated data)     │
+    │       ↓                                       │
+    │  70_Export → 80_EndProcess                    │
+    │                                               │
+    └───────────────────────────────────────────────┘
+           ↓ (on error)
+    ERR_HandleDocumentError (per-document)
+    ERR_AbortProcess (fatal/process-level)
+```
+
+---
+
+## 4. Workflow Details
+
+### 00_ReadConfigFile.xaml
+- Reads Config.xlsx Settings + Constants sheets into Dictionary\<String, String\>
+- Calculates `MaxAttempts` (minimum 1) and `RetryInterval` from config
+- **No retry mechanism** — config file must be accessible
+
+### 10_InitializeProcess.xaml
+- **LoadTaxonomy** from `taxonomy.json`
+- Reads Assets sheet → fetches each from Orchestrator via **Get Robot Asset**
+- If asset fetch fails but key exists in config: silently continues (fallback value)
+- If asset fetch fails and key missing: throws (asset is required)
+- Wrapped in **RetryScope** for Orchestrator connectivity issues
+
+### 15_GetTransactionItem.xaml
+- **Get Transaction Item** from Orchestrator queue
+- Extracts `TargetFile` from `TransactionItem.SpecificContent["TargetFile"]`
+- **Copies ALL SpecificContent into Config dictionary** for process-wide access
+- Wrapped in RetryScope
+
+### 20_Digitize.xaml
+- **DigitizeDocument** activity with embedded OCR engine
+- Default: **UiPath Document OCR** (cloud API)
+- DegreeOfParallelism=-1 (unlimited parallel page processing)
+- Has config-specific `MaxExecutionAttemptsDigitize` override
+- **Warning**: OCR is API-based and incurs costs on retry
+
+### 30_Classify.xaml
+- **ClassifyDocumentScope** with **IntelligentKeywordClassifier**
+- Uses learning file from `DocumentProcessing/IntelligentKeywordLearningFile.json`
+- Pre-configured for: Receipt, Invoice, W-9, Certificate-of-filing
+- Returns ClassificationResult[] with confidence scores
+
+### 35_ClassificationBussinessRuleValidation.xaml (Flowchart)
+- Checks `AlwaysValidateClassification` config flag
+- FlowSwitch on DocumentTypeId for type-specific rules
+- Sets `AutoClassificationSuccess` = False to trigger manual validation
+- **Customization point**: Add your classification business rules here
+
+### 40_TrainClassifiers.xaml
+- **File locking** mechanism for multi-robot synchronization:
+  1. LockFile.xaml copies classifier file with `.lock` suffix
+  2. TrainClassifiersScope writes to lock copy
+  3. UnlockFile.xaml moves lock copy back (Overwrite=True)
+- **ContinueOnError=True** on training (prefer losing training data over blocking extraction)
+- **ContinueOnError=False** on unlock (critical — failure blocks ALL future training)
+
+### 50_Extract.xaml
+- **DataExtractionScope** with configured extractors
+- Extractor choice depends on document type (Form Extractor, ML Extractor, Regex)
+- Has config-specific `MaxExecutionAttemptsExtract` override
+
+### 55_ExtractionBussinessRuleValidation.xaml (Flowchart)
+- Checks `AlwaysValidateExtraction` config flag
+- FlowSwitch on DocumentTypeId
+- **Invoice case**: Calls InvoicePostProcessing.xaml with validation rules:
+  1. Verify mandatory fields extracted
+  2. Verify Quantity × Unit Price = Line Amount per row
+  3. Verify sum of Line Amounts = Net Amount
+  4. Verify Net Amount + charges = Total
+  5. Verify field confidence thresholds
+- Sets `AutoExtractionSuccess` = False for manual validation
+
+### 60_TrainExtractors.xaml
+- **TrainExtractorsScope** with human-validated data
+- ContinueOnError=True (prefer export over training loss)
+
+### 70_Export.xaml
+- **ExportExtractionResults** → DataSet (IncludeConfidence=True)
+- Default: Writes each DataTable to Excel sheet
+- Output path: `ExportsFolder/DocumentId_StartPage-EndPage.xlsx`
+- **Example only** — real implementations should use Data Service or database
+
+### 80_EndProcess.xaml
+- Calls SetTransactionStatus with Success
+- Customization point for post-export logic
+
+---
+
+## 5. Error Handling
+
+### Two Error Handlers
+| Handler | Scope | Triggered By | Default Action |
+|---------|-------|-------------|----------------|
+| `ERR_HandleDocumentError.xaml` | Per-document | Exception during document processing | Log warning, continue to next document |
+| `ERR_AbortProcess.xaml` | Process-level | Fatal exception (init failure, queue error) | Log error, terminate process |
+
+### Exception Types (same as REFramework)
+- **BusinessRuleException**: Data problem → skip document, don't retry
+- **System.Exception**: Technical problem → can retry with reinitialization
+
+### SetTransactionStatus (Flowchart)
+- **Success**: Both exceptions null → Set status "Successful"
+- **Business Exception**: BusinessRuleException or DocumentRejectedByUserException → Status "Failed", ErrorType "Business"
+- **System Exception**: Any other exception → Status "Failed", ErrorType "Application"
+
+---
+
+## 6. Multi-Robot Synchronization
+
+Two mechanisms for concurrent access to training files:
+
+### File Locking (used by default for classifier training)
+1. `LockFile.xaml`: Copy file with `.lock` suffix → work on copy
+2. `UnlockFile.xaml`: Move copy back → overwrite original
+- **Risk**: Two robots writing simultaneously → one's training data lost
+
+### Queue-Based Semaphore (alternative)
+1. Create Orchestrator queue with SINGLE item, no auto-retry
+2. `GetWritePermission.xaml`: Loop until queue item obtained (= semaphore acquired)
+3. `GiveUpWritePermission.xaml`: Mark item successful + recreate it (= semaphore released)
+- **Stronger guarantee** but requires Orchestrator queue setup
+
+---
+
+## 7. Dependencies (Current Template)
+
+```json
+{
+  "UiPath.DocumentUnderstanding.ML.Activities": "[1.9.1]",
+  "UiPath.Excel.Activities": "[2.11.4]",
+  "UiPath.IntelligentOCR.Activities": "[5.0.2]",
+  "UiPath.System.Activities": "[21.10.2]",
+  "UiPath.UIAutomation.Activities": "[21.10.3]"
+}
+```
+
+Additional assembly references: `UiPath.Persistence.Activities`, `UiPath.OmniPage.Activities`, `UiPath.PDF.Activities`, `UiPath.DocumentProcessing.Contracts`, `UiPath.OCR.Contracts`
+
+**Key**: `supportsPersistence: true` in project.json (required for Action Center mode)
+
+---
+
+## 8. Taxonomy Structure (taxonomy.json)
+
+Pre-configured document types:
+
+| Document Type ID | Category |
+|-----------------|----------|
+| `Semi-StructuredDocuments.Financial.Receipt` | Receipts |
+| `Semi-StructuredDocuments.Financial.Invoice` | Invoices (with line items table) |
+| `StructuredDocuments.LendingForms.W-9` | US tax form |
+| `Documents.Other.Certificate-of-filing` | Filing certificates |
+
+### Invoice Fields (example)
+- **Header**: Vendor Name, Address, VAT Number, Invoice No, PO No, Invoice Date, Due Date
+- **Amounts**: Net Amount, Tax Amount, Total, Discount, Shipping Charges
+- **Line Items Table**: Line Number, Description, Quantity, Unit Price, Line Amount, Part Number
+- **Payment**: Terms, Currency
+
+Supported languages: 26 (EN, DE, FR, ES, RU, ZH, JA, etc.)
+
+---
+
+## 9. Config.xlsx Key Settings
+
+| Setting | Purpose | Default |
+|---------|---------|---------|
+| `MaxExecutionAttempts` | Default retry count | 3 |
+| `RetryInterval` | Seconds between retries | 5 |
+| `MaxExecutionAttemptsDigitize` | Digitization-specific retries | (overrides default) |
+| `MaxExecutionAttemptsClassify` | Classification-specific retries | (overrides default) |
+| `MaxExecutionAttemptsExtract` | Extraction-specific retries | (overrides default) |
+| `AlwaysValidateClassification` | Force manual classification review | True/False |
+| `AlwaysValidateExtraction` | Force manual extraction review | True/False |
+| `DocumentUnderstandingQueueName` | Orchestrator queue name | |
+| `DocumentUnderstandingQueuePath` | Orchestrator queue folder | |
+| `ClassifierLearningFilePath` | Path to classifier training file | |
+| `ExportsFolder` | Output folder for exported results | |
+| `logKey` | Auto-generated GUID+timestamp for log tracking | |
+
+---
+
+## 10. Critical Gotchas
+
+### Architecture
+1. **One job per document** — NOT bulk processing. Requires dispatcher to populate queue.
+2. **supportsPersistence=true required** for Action Center mode — set in project.json
+3. **Process.xaml does NOT exist at root** — business logic is inside the numbered Framework workflows
+4. **NOT a State Machine** — unlike REFramework, this is a Sequence/Flowchart pipeline
+
+### OCR & Digitization
+5. **OCR is API-based (cloud)** — incurs costs on every call, including retries
+6. **DegreeOfParallelism=-1** means unlimited — can hit API rate limits
+7. **"Input document not found"** — [common error](https://forum.uipath.com/t/document-understanding-process-download-document/738008) when path resolution fails
+8. **.NET framework missing** — [reported issue](https://forum.uipath.com/t/document-understanding-error-install-missing-frameworks-for-net/727634) with template setup
+
+### Classification & Extraction
+9. **Classifiers and extractors are process-dependent** — must configure for YOUR document types
+10. **IntelligentKeywordClassifier needs training data** — `IntelligentKeywordLearningFile.json` must be populated
+11. **ML Extractor failures** — [common](https://forum.uipath.com/t/machine-learning-extractor-failed-in-50-extract-document-understanding-template/539547) with wrong API keys or model configuration
+12. **Taxonomy must match exactly** — DocumentTypeId in taxonomy.json must match classifier configuration
+
+### Training & Synchronization
+13. **Training data CAN be lost** — ContinueOnError=True on training; two robots writing simultaneously overwrites
+14. **UnlockFile ContinueOnError=False** — if unlock fails, ALL future training blocked
+15. **Queue semaphore requires setup** — single-item queue with NO auto-retry
+
+### Validation
+16. **InvoicePostProcessing uses EN-US culture** — decimal ".", thousand "," — will fail for other locales
+17. **"Do NOT use InvoicePostProcessing as-is"** — template annotation warns it's for demos only
+18. **AlwaysValidateClassification/Extraction** — set True during development, tune confidence thresholds for production
+
+### Action Center Mode
+19. **Requires Orchestrator 2021.10+** for full Action Center support
+20. **Persistence activities serialize workflow state** — all variables at suspension point must be serializable
+21. **Cannot have open UI scopes** across Action Center suspension points
+
+---
+
+## 11. Customization Checklist
+
+- [ ] **Config.xlsx**: Set queue name, asset names, retry parameters
+- [ ] **taxonomy.json**: Define YOUR document types and fields via Taxonomy Manager
+- [ ] **30_Classify.xaml**: Configure classifiers for your document types
+- [ ] **50_Extract.xaml**: Configure extractors for your document types
+- [ ] **35_ClassificationBussinessRuleValidation.xaml**: Add classification rules
+- [ ] **55_ExtractionBussinessRuleValidation.xaml**: Add extraction validation rules
+- [ ] **70_Export.xaml**: Replace Excel export with your target system (DB, API, Data Service)
+- [ ] **Dispatcher workflow**: Create separate workflow to populate queue with TargetFile paths
+- [ ] Choose entry point: Main-Attended.xaml (local) or Main-ActionCenter.xaml (Orchestrator)
+- [ ] Train classifiers with sample documents before production
+
+---
+
+## Sources
+- [UiPath Docs - Document Understanding Process Template](https://docs.uipath.com/document-understanding/automation-suite/2024.10/classic-user-guide/du-process)
+- [UiPath Forum - DU Process Template Announcement](https://forum.uipath.com/t/document-understanding-process-new-studio-template/330234)
+- [UiPath Docs - About DU Process](https://docs.uipath.com/activities/other/latest/document-understanding/about-document-understanding-process-studio-template)
+- Desktop template: `C:\Users\alexandru.roman\Documents\UiPath\legacy_tests\DocumentUnderstandingProcess\`