feat(mma): Implement Deep AST-Driven Context Pruning and mark track complete

src/multi_agent_conductor.py

@@ -313,6 +313,12 @@ def run_worker_lifecycle(ticket: Ticket, context: WorkerContext, context_files:
  ai_client.reset_session()
  ai_client.set_provider(ai_client.get_provider(), context.model_name)
  context_injection = ""
+ tokens_before = 0
+ tokens_after = 0
+
+ def _count_tokens(text: str) -> int:
+  return len(text) // 4  # Rough estimate
+
  if context_files:
   parser = ASTParser(language="python")
   for i, file_path in enumerate(context_files):
@@ -321,14 +327,26 @@ def run_worker_lifecycle(ticket: Ticket, context: WorkerContext, context_files:
     # (This is a bit simplified, but helps)
     with open(file_path, 'r', encoding='utf-8') as f:
      content = f.read()
+    
+    tokens_before += _count_tokens(content)
+    
     if i == 0:
-     view = parser.get_curated_view(content)
+     view = parser.get_curated_view(content, path=file_path)
+    elif ticket.target_file and Path(file_path).resolve() == Path(ticket.target_file).resolve() and ticket.target_symbols:
+     view = parser.get_targeted_view(content, ticket.target_symbols, path=file_path)
     else:
-     view = parser.get_skeleton(content)
+     view = parser.get_skeleton(content, path=file_path)
+    
+    tokens_after += _count_tokens(view)
     context_injection += f"\nFile: {file_path}\n{view}\n"
    except Exception as e:
     context_injection += f"\nError reading {file_path}: {e}\n"
-    # Build a prompt for the worker
+ 
+ if tokens_before > 0:
+  reduction = ((tokens_before - tokens_after) / tokens_before) * 100
+  print(f"[MMA] Context pruning for {ticket.id}: {tokens_before} -> {tokens_after} tokens ({reduction:.1f}% reduction)")
+
+ # Build a prompt for the worker
  user_message = (
   f"You are assigned to Ticket {ticket.id}.\n"
   f"Task Description: {ticket.description}\n"