feat: Add comprehensive debug information for workflow nodes

Backend changes:
- workflow_models.py: Add debug_prompt, debug_raw_response, debug_node_type, debug_prompt_slug, metadata fields to NodeExecutionState
- workflow_executor.py: Capture and store debug info for analysis, logic, and join nodes when enable_debug=True
  - Analysis nodes: store full prompt + raw AI response
  - Logic nodes: store expression + evaluation result
  - Join nodes: store strategy + path statistics

Frontend changes:
- Analysis.jsx: Enable debug mode by default (debug=true) for all workflow executions

This allows developers to see exactly what prompt was sent to the AI, what response was received, and how each node was processed - essential for debugging workflow issues.

backend/workflow_executor.py

@@ -314,11 +314,11 @@ async def execute_node(
 
         # Logic Nodes (Phase 3)
         if node.type == "logic":
-            return execute_logic_node(node, context, graph)
+            return execute_logic_node(node, context, graph, enable_debug)
 
         # Join Nodes (Phase 4)
         if node.type == "join":
-            return execute_join_node(node, context, graph)
+            return execute_join_node(node, context, graph, enable_debug)
 
         # Analysis Nodes
         if node.type == "analysis":
@@ -398,6 +398,11 @@ async def execute_node(
                 decision_signals=parsed["decision_signals"],
                 normalized_signals=normalized_signals,
                 reasoning_anchors=parsed.get("reasoning_anchors"),
+                # Debug information (nur wenn enable_debug=True)
+                debug_prompt=augmented_prompt if enable_debug else None,
+                debug_raw_response=llm_response if enable_debug else None,
+                debug_node_type="analysis",
+                debug_prompt_slug=node.prompt_slug if enable_debug else None,
                 started_at=started_at,
                 completed_at=datetime.utcnow().isoformat()
             )
@@ -419,7 +424,8 @@ async def execute_node(
 def execute_logic_node(
     node,
     context: Dict[str, Any],
-    graph: WorkflowGraph
+    graph: WorkflowGraph,
+    enable_debug: bool = False
 ) -> NodeExecutionState:
     """
     Führt Logic Node aus (Phase 3).
@@ -519,7 +525,15 @@ def execute_logic_node(
                 "evaluation_result": result if not error else None,
                 "error": error,
                 "activated_edges": activated_edges
-            })
+            }),
+            # Debug information
+            debug_prompt=json.dumps(expression.model_dump() if hasattr(expression, 'model_dump') else str(expression), ensure_ascii=False) if enable_debug and expression else None,
+            debug_raw_response=json.dumps({
+                "result": result,
+                "error": error,
+                "activated_edges": activated_edges
+            }, ensure_ascii=False) if enable_debug else None,
+            debug_node_type="logic"
         )
 
     except Exception as e:
@@ -536,7 +550,8 @@ def execute_logic_node(
 def execute_join_node(
     node,
     context: Dict[str, Any],
-    graph: WorkflowGraph
+    graph: WorkflowGraph,
+    enable_debug: bool = False
 ) -> NodeExecutionState:
     """
     Führt Join Node aus (Phase 4).
@@ -608,7 +623,16 @@ def execute_join_node(
             normalized_signals=consolidated_signals_list,
             metadata=join_result.metadata,
             started_at=started_at,
-            completed_at=datetime.utcnow().isoformat()
+            completed_at=datetime.utcnow().isoformat(),
+            # Debug information
+            debug_prompt=f"Join Strategy: {node.join_strategy.value if node.join_strategy else 'wait_all'}\nPaths: {total_count}\nMinimum Required: {node.min_paths if node.min_paths else 'all'}" if enable_debug else None,
+            debug_raw_response=json.dumps({
+                "executed_paths": executed_count,
+                "total_paths": total_count,
+                "strategy": node.join_strategy.value if node.join_strategy else 'wait_all',
+                "consolidated_nodes": list(join_result.consolidated_analysis_core.keys())
+            }, ensure_ascii=False) if enable_debug else None,
+            debug_node_type="join"
         )
 
     except Exception as e:

backend/workflow_models.py

@@ -349,6 +349,15 @@ class NodeExecutionState(BaseModel):
     normalized_signals: List[NormalizedSignal] = Field(default_factory=list, description="Normalisierte Signale (Phase 2)")
     reasoning_anchors: Optional[str] = Field(None, description="Begründungsanker aus ## Begründung")
 
+    # Debug Information (nur wenn enable_debug=True)
+    debug_prompt: Optional[str] = Field(None, description="Vollständiger Prompt der an die KI gesendet wurde")
+    debug_raw_response: Optional[str] = Field(None, description="Rohe KI-Antwort (ungeparst)")
+    debug_node_type: Optional[str] = Field(None, description="Node-Typ (analysis, logic, join, etc.)")
+    debug_prompt_slug: Optional[str] = Field(None, description="Verwendeter Prompt-Slug (bei ANALYSIS nodes)")
+
+    # Metadata (für Join Nodes und andere Zusatzinfos)
+    metadata: Optional[Dict[str, Any]] = Field(None, description="Zusätzliche Metadaten (z.B. Join-Statistiken)")
+
     # Error & Timing
     error: Optional[str] = Field(None, description="Fehlermeldung bei failed")
     started_at: Optional[str] = Field(None, description="Start-Timestamp (ISO)")

frontend/src/pages/Analysis.jsx

@@ -382,7 +382,7 @@ export default function Analysis() {
     setLoading(slug); setError(null); setNewResult(null); setProgress(null)
     try {
       // Use SSE-based executor for long-running workflows
-      const result = await api.executeUnifiedPromptStream(slug, null, null, false, true, (event) => {
+      const result = await api.executeUnifiedPromptStream(slug, null, null, true, true, (event) => {
         // Progress callback: update UI in real-time
         if (event.type === 'execution_started') {
           setProgress({ total_nodes: 0, completed_nodes: 0, current_node_label: 'Starte...' })