Remove hardcoded debug.log file writes - fix file not found errors

src/gepa_optimizer/core/optimizer.py

@@ -139,12 +139,7 @@ class GepaOptimizer:
             self.logger.info(f"Starting optimization session: {session_id}")
             self.logger.info(f"Using model: {self.config.model.model_name} (provider: {self.config.model.provider})")
             
-            # #region agent log
-            import json as _json_debug
-            _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-            with open(_debug_log_path, "a") as _f:
-                _f.write(_json_debug.dumps({"hypothesisId": "E", "location": "optimizer.py:train_start", "message": "Optimization train() started", "data": {"session_id": session_id, "max_iterations": self.config.max_iterations}, "timestamp": int(time.time() * 1000), "sessionId": "debug-session"}) + "\n")
-            # #endregion
+            # Debug logging removed - not needed in production
             
             # 🔥 FIX E: Reset Pareto logger at start of each optimization run
             from ..utils.pareto_logger import reset_pareto_logger
@@ -906,12 +901,7 @@ Output the improved prompt directly and only the prompt."""
             if reflection_lm_passed:
                 self.logger.debug(f"reflection_lm_callable passed to GEPA (may be ignored in adapter mode)")
             
-            # #region agent log
-            import json as _json_debug
-            _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-            with open(_debug_log_path, "a") as _f:
-                _f.write(_json_debug.dumps({"hypothesisId": "A", "location": "optimizer.py:gepa_params", "message": "GEPA params construction", "data": {"max_iterations_from_config": max_iterations, "max_metric_calls": max_metric_calls, "batch_size": batch_size}, "timestamp": int(time.time() * 1000), "sessionId": "debug-session"}) + "\n")
-            # #endregion
+            # Debug logging removed - not needed in production
             
             gepa_params = {
                 'adapter': adapter,  # Use the adapter created above (with or without LLEGO)
@@ -958,10 +948,7 @@ Output the improved prompt directly and only the prompt."""
                 elif key not in VALID_GEPA_PARAMS:
                     self.logger.debug(f"⚠️  Filtering out invalid GEPA parameter: {key}")
             
-            # #region agent log
-            with open(_debug_log_path, "a") as _f:
-                _f.write(_json_debug.dumps({"hypothesisId": "A", "location": "optimizer.py:gepa_params_final", "message": "Final GEPA params keys", "data": {"params_keys": list(gepa_params.keys()), "max_metric_calls": gepa_params.get('max_metric_calls', 'NOT_PASSED')}, "timestamp": int(time.time() * 1000), "sessionId": "debug-session"}) + "\n")
-            # #endregion
+            # Debug logging removed - not needed in production
             
             # 🎯 NEW: Capture GEPA's internal logging for pareto front information
             gepa_output = io.StringIO()
@@ -1060,12 +1047,7 @@ Output the improved prompt directly and only the prompt."""
     
         self.logger.info(f"GEPA Optimization Complete: {max_iteration} iterations")
         
-        # #region agent log
-        import json as _json_debug
-        _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-        with open(_debug_log_path, "a") as _f:
-            _f.write(_json_debug.dumps({"hypothesisId": "F", "location": "optimizer.py:gepa_complete", "message": "GEPA optimization complete - iteration count", "data": {"max_iteration_from_logs": max_iteration, "expected_iterations": self.config.max_iterations, "off_by_one": max_iteration != self.config.max_iterations, "gepa_logs_length": len(gepa_logs)}, "timestamp": int(time.time() * 1000), "sessionId": "debug-session"}) + "\n")
-        # #endregion
+        # Debug logging removed - not needed in production
         
         return max_iteration  # Return the max iteration count
     

src/gepa_optimizer/core/universal_adapter.py

@@ -200,27 +200,11 @@ Output the improved prompt directly and only the prompt."""
                 self.logger.info(f"📐 FORMAT DETECTED: {self._detected_format['format_type']}")
                 self.logger.info(f"   Spec: {self._detected_format['format_spec'][:100]}...")
                 self.logger.info(f"   Avg length: {self._detected_format['avg_length']} chars")
-                # #region agent log
-                import json as _json_debug
-                import time as _time_debug
-                import os as _os_debug
-                _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-                _os_debug.makedirs(_os_debug.path.dirname(_debug_log_path), exist_ok=True)
-                with open(_debug_log_path, "a") as _f:
-                    _f.write(_json_debug.dumps({"hypothesisId": "FORMAT_DETECT", "location": "universal_adapter.py:format_detected", "message": "Format detection successful", "data": {"format_type": self._detected_format['format_type'], "num_outputs": len(expected_outputs), "avg_length": self._detected_format['avg_length'], "has_constraint": bool(self._detected_format.get('format_constraint'))}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-                # #endregion
+            # Debug logging removed - not needed in production
             else:
                 self.logger.warning("⚠️ No expected outputs found for format detection")
                 self._detected_format = None
-                # #region agent log
-                import json as _json_debug
-                import time as _time_debug
-                import os as _os_debug
-                _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-                _os_debug.makedirs(_os_debug.path.dirname(_debug_log_path), exist_ok=True)
-                with open(_debug_log_path, "a") as _f:
-                    _f.write(_json_debug.dumps({"hypothesisId": "FORMAT_DETECT", "location": "universal_adapter.py:format_detected", "message": "Format detection failed - no outputs", "data": {"batch_size": len(batch)}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-                # #endregion
+            # Debug logging removed - not needed in production
                 
         except Exception as e:
             self.logger.warning(f"⚠️ Format detection failed: {e}")
@@ -323,14 +307,7 @@ Output the improved prompt directly and only the prompt."""
                 self.logger.debug("🌱 Detected seed prompt (S₀)")
             else:
                 candidate_type = 'unknown'  # Truly unknown
-        
-        # #region agent log
-        import json as _json_debug
-        import time as _time_debug
-        _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-        with open(_debug_log_path, "a") as _f:
-            _f.write(_json_debug.dumps({"hypothesisId": "C", "location": "universal_adapter.py:candidate_type_detect", "message": "Candidate type detection", "data": {"candidate_type": candidate_type, "evaluation_count": self._evaluation_count, "from_candidate_dict": candidate.get('source'), "from_sources_mapping": self._candidate_sources.get(system_prompt), "from_current_type": self._current_evaluation_type}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-        # #endregion
+            # Debug logging removed - not needed in production
         
         # Store source for future lookups (always update if we found a valid type)
         if candidate_type and candidate_type != 'unknown' and system_prompt not in self._candidate_sources:
@@ -338,19 +315,7 @@ Output the improved prompt directly and only the prompt."""
             self.logger.debug(f"   📝 Stored candidate type: {candidate_type} for prompt (length: {len(system_prompt)})")
         
         # Dataset type already determined above for cache check - reuse it
-        
-        # #region agent log
-        try:
-            import json as _json_debug
-            import time as _time_debug
-            import os as _os_debug
-            _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-            _os_debug.makedirs(_os_debug.path.dirname(_debug_log_path), exist_ok=True)
-            with open(_debug_log_path, "a") as _f:
-                _f.write(_json_debug.dumps({"hypothesisId": "H", "location": "universal_adapter.py:dataset_type_detect", "message": "Dataset type detection", "data": {"batch_size": len(batch), "valset_size": getattr(self, '_valset_size', None), "batch_size_threshold": batch_size_threshold, "detected_type": dataset_type, "evaluation_count": self._evaluation_count}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-        except Exception:
-            pass
-        # #endregion
+            # Debug logging removed - not needed in production
         
         # Check if this is a new candidate (different from last one)
         if self._last_candidate != system_prompt:
@@ -383,14 +348,7 @@ Output the improved prompt directly and only the prompt."""
             )
         
         avg_score = sum(scores) / len(scores) if scores else 0.0
-        
-        # #region agent log
-        import json as _json_debug
-        import time as _time_debug
-        _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-        with open(_debug_log_path, "a") as _f:
-            _f.write(_json_debug.dumps({"hypothesisId": "B,C", "location": "universal_adapter.py:baseline_check", "message": "Baseline check conditions", "data": {"baseline_score_is_none": self._baseline_score is None, "current_dataset_type": self._current_dataset_type, "current_evaluation_type": self._current_evaluation_type, "is_baseline_evaluation": self._is_baseline_evaluation, "batch_size": len(batch), "avg_score": avg_score}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-        # #endregion
+            # Debug logging removed - not needed in production
         
         # 🔥 CRITICAL FIX: Baseline MUST be set from seed's first Dpareto evaluation ONLY
         # This ensures FAIR comparison: seed and candidates evaluated on SAME dataset (Dpareto) with SAME number of datapoints
@@ -416,10 +374,7 @@ Output the improved prompt directly and only the prompt."""
                 pareto_log.set_baseline(avg_score)
                 self.logger.info(f"📊 Baseline score (Dpareto, {len(batch)} samples): {avg_score:.4f}")
                 self.logger.info(f"   ✅ Baseline set from {'baseline evaluation' if is_baseline_eval else 'first Dpareto'} (type: {self._current_evaluation_type})")
-                # #region agent log
-                with open(_debug_log_path, "a") as _f:
-                    _f.write(_json_debug.dumps({"hypothesisId": "B", "location": "universal_adapter.py:baseline_set", "message": "Baseline score SET", "data": {"baseline_score": avg_score, "candidate_type": self._current_evaluation_type, "dataset_type": self._current_dataset_type, "is_baseline_eval": is_baseline_eval}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-                # #endregion
+            # Debug logging removed - not needed in production
             # Note: Test set evaluations are ignored for baseline - baseline comes from Dpareto
         else:
             # 🔥 SAFETY CHECK: Ensure Pareto logger also has baseline if adapter has it
@@ -595,19 +550,7 @@ Output the improved prompt directly and only the prompt."""
             
             composite_score = evaluation_results.get("composite_score", 0.0)
             scores.append(composite_score)
-            
-            # #region agent log
-            try:
-                import json as _json_debug
-                import time as _time_debug
-                import os as _os_debug
-                _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-                _os_debug.makedirs(_os_debug.path.dirname(_debug_log_path), exist_ok=True)
-                with open(_debug_log_path, "a") as _f:
-                    _f.write(_json_debug.dumps({"hypothesisId": "G", "location": "universal_adapter.py:evaluation_result", "message": "Individual evaluation result", "data": {"sample_idx": i, "composite_score": composite_score, "semantic_sim": evaluation_results.get("semantic_similarity", -1), "structural_sim": evaluation_results.get("structural_similarity", -1), "format_mismatch": evaluation_results.get("analysis", {}).get("format_mismatch", False), "predicted_len": len(predicted_output) if predicted_output else 0, "expected_len": len(standardized_item.get('output', ''))}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-            except Exception:
-                pass
-            # #endregion
+            # Debug logging removed - not needed in production
             
             # Update performance tracking
             self._evaluation_count += 1

src/gepa_optimizer/evaluation/universal_evaluator.py

@@ -200,19 +200,7 @@ class UniversalSemanticEvaluator(BaseEvaluator):
         logger.debug(f"📊 Evaluation: composite={result['composite_score']:.3f}, "
                     f"semantic={result['semantic_similarity']:.3f}, "
                     f"structural={result['structural_similarity']:.3f}")
-        
-        # #region agent log
-        try:
-            import json as _json_debug
-            import time as _time_debug
-            import os as _os_debug
-            _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-            _os_debug.makedirs(_os_debug.path.dirname(_debug_log_path), exist_ok=True)
-            with open(_debug_log_path, "a") as _f:
-                _f.write(_json_debug.dumps({"hypothesisId": "G", "location": "universal_evaluator.py:final_score", "message": "Final evaluation score breakdown", "data": {"composite": result["composite_score"], "semantic": result["semantic_similarity"], "structural": result["structural_similarity"], "exact_match": result["exact_match"], "format_mismatch": result.get("analysis", {}).get("format_mismatch", False), "predicted_preview": predicted_str[:150] if predicted_str else "EMPTY", "expected_preview": expected_str[:150] if expected_str else "EMPTY"}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-        except Exception as _e:
-            pass  # Silent fail for instrumentation
-        # #endregion
+            # Debug logging removed - not needed in production
         
         return result
     

src/gepa_optimizer/llms/llego_enhanced_llm.py

@@ -253,15 +253,7 @@ class LLEGOEnhancedLLMClient(BaseLLMClient):
                     json_str = json.dumps(python_dict)
                     data = json.loads(json_str)
                     if 'variations' in data:
-                        # #region agent log
-                        import json as _json_debug
-                        import time as _time_debug
-                        import os as _os_debug
-                        _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-                        _os_debug.makedirs(_os_debug.path.dirname(_debug_log_path), exist_ok=True)
-                        with open(_debug_log_path, "a") as _f:
-                            _f.write(_json_debug.dumps({"hypothesisId": "JSON_FIX", "location": "llego_enhanced_llm.py:python_dict_parse", "message": "Successfully parsed Python dict syntax", "data": {"num_expected": num_expected, "parsed_variations": len(data.get('variations', []))}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-                        # #endregion
+            # Debug logging removed - not needed in production
                         return self._extract_variations_from_json(data, num_expected)
             except (ValueError, SyntaxError, TypeError) as e:
                 # If ast.literal_eval fails, try string replacement as fallback
@@ -271,15 +263,7 @@ class LLEGOEnhancedLLMClient(BaseLLMClient):
                     converted = cleaned.replace("'", '"')
                     data = json.loads(converted)
                     if 'variations' in data:
-                        # #region agent log
-                        import json as _json_debug
-                        import time as _time_debug
-                        import os as _os_debug
-                        _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-                        _os_debug.makedirs(_os_debug.path.dirname(_debug_log_path), exist_ok=True)
-                        with open(_debug_log_path, "a") as _f:
-                            _f.write(_json_debug.dumps({"hypothesisId": "JSON_FIX", "location": "llego_enhanced_llm.py:python_dict_string_replace", "message": "Parsed Python dict via string replacement", "data": {"num_expected": num_expected, "parsed_variations": len(data.get('variations', []))}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-                        # #endregion
+            # Debug logging removed - not needed in production
                         return self._extract_variations_from_json(data, num_expected)
                 except json.JSONDecodeError:
                     pass
@@ -360,14 +344,7 @@ class LLEGOEnhancedLLMClient(BaseLLMClient):
             return self._parse_numbered_section_variations(response_text, num_expected)
         except ValueError:
             pass
-        
-        # #region agent log
-        import json as _json_debug
-        import time as _time_debug
-        _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-        with open(_debug_log_path, "a") as _f:
-            _f.write(_json_debug.dumps({"hypothesisId": "D", "location": "llego_enhanced_llm.py:json_parse_fail", "message": "JSON parsing failed completely", "data": {"num_expected": num_expected, "response_preview": response_text[:500] if response_text else "EMPTY", "response_length": len(response_text) if response_text else 0}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-        # #endregion
+            # Debug logging removed - not needed in production
         
         raise ValueError(f"Could not parse {num_expected} variations from response")
     
@@ -610,19 +587,7 @@ class LLEGOEnhancedLLMClient(BaseLLMClient):
             logger.debug(f"   user_prompt length: {len(user_prompt)} chars")
             logger.debug(f"   in_reflection: {self.reflection_context['in_reflection']}")
             logger.debug(f"   has_image: {bool(image_base64)}")
-        
-        # #region agent log
-        try:
-            import json as _json_debug
-            import time as _time_debug
-            import os as _os_debug
-            _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-            _os_debug.makedirs(_os_debug.path.dirname(_debug_log_path), exist_ok=True)
-            with open(_debug_log_path, "a") as _f:
-                _f.write(_json_debug.dumps({"hypothesisId": "INTERCEPTION", "location": "llego_enhanced_llm.py:generate", "message": "Generate called", "data": {"system_prompt_len": len(system_prompt), "user_prompt_len": len(user_prompt), "has_image": bool(image_base64), "has_candidates": len(getattr(self, '_adapter_generated_candidates', [])), "in_reflection": self.reflection_context.get('in_reflection', False)}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-        except Exception:
-            pass
-        # #endregion
+            # Debug logging removed - not needed in production
         
         # 🔥 CRITICAL: Check if we have pre-generated candidates from adapter-level generation
         # This happens when GEPA calls adapter.llm_client to generate candidates
@@ -999,26 +964,10 @@ class LLEGOEnhancedLLMClient(BaseLLMClient):
             if self._detected_format and self._detected_format.get('format_constraint'):
                 format_constraint = self._detected_format['format_constraint']
                 logger.info(f"📐 Injecting format constraint into candidate generation")
-                # #region agent log
-                import json as _json_debug
-                import time as _time_debug
-                import os as _os_debug
-                _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-                _os_debug.makedirs(_os_debug.path.dirname(_debug_log_path), exist_ok=True)
-                with open(_debug_log_path, "a") as _f:
-                    _f.write(_json_debug.dumps({"hypothesisId": "FORMAT_CONSTRAINT", "location": "llego_enhanced_llm.py:format_injection", "message": "Format constraint injected", "data": {"format_type": self._detected_format.get('format_type', 'unknown'), "constraint_length": len(format_constraint), "avg_length": self._detected_format.get('avg_length', 0)}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-                # #endregion
+            # Debug logging removed - not needed in production
             else:
                 format_constraint = "No specific format detected - ensure output is CONCISE and matches expected examples."
-                # #region agent log
-                import json as _json_debug
-                import time as _time_debug
-                import os as _os_debug
-                _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-                _os_debug.makedirs(_os_debug.path.dirname(_debug_log_path), exist_ok=True)
-                with open(_debug_log_path, "a") as _f:
-                    _f.write(_json_debug.dumps({"hypothesisId": "FORMAT_CONSTRAINT", "location": "llego_enhanced_llm.py:format_injection", "message": "No format constraint available", "data": {"has_detected_format": bool(self._detected_format)}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-                # #endregion
+            # Debug logging removed - not needed in production
             
             # 🔥 EVOLUTIONARY PROMPT ENGINEER: Forces radically different mutations
             # Each variation MUST use a distinct genetic strategy to maximize search space

src/gepa_optimizer/llms/vision_llm.py

@@ -395,31 +395,7 @@ class VisionLLMClient(BaseLLMClient):
         ]
         
         if image_base64:
-            # #region agent log
-            import json as _json_debug
-            import time as _time_debug
-            _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-            try:
-                with open(_debug_log_path, "a") as _f:
-                    _f.write(_json_debug.dumps({
-                        "id": f"log_{int(_time_debug.time() * 1000)}",
-                        "timestamp": int(_time_debug.time() * 1000),
-                        "location": "vision_llm.py:_generate_openai",
-                        "message": "Image base64 BEFORE processing",
-                        "data": {
-                            "image_base64_length": len(image_base64) if image_base64 else 0,
-                            "has_data_uri_prefix": image_base64.startswith("data:image") if image_base64 else False,
-                            "prefix": image_base64[:50] if image_base64 and len(image_base64) > 50 else image_base64,
-                            "is_none": image_base64 is None,
-                            "is_empty": image_base64 == "" if image_base64 else True
-                        },
-                        "sessionId": "debug-session",
-                        "runId": "run1",
-                        "hypothesisId": "A,C,D"
-                    }) + "\n")
-            except Exception:
-                pass
-            # #endregion
+            # Debug logging removed - not needed in production
             
             # Detect and extract image format
             detected_format = "jpeg"  # Default fallback
@@ -473,29 +449,7 @@ class VisionLLMClient(BaseLLMClient):
             final_format = format_map.get(detected_format, "jpeg")
             
             final_url = f"data:image/{final_format};base64,{clean_base64}"
-            
-            # #region agent log
-            try:
-                with open(_debug_log_path, "a") as _f:
-                    _f.write(_json_debug.dumps({
-                        "id": f"log_{int(_time_debug.time() * 1000)}",
-                        "timestamp": int(_time_debug.time() * 1000),
-                        "location": "vision_llm.py:_generate_openai",
-                        "message": "Image URL AFTER processing",
-                        "data": {
-                            "detected_format": detected_format,
-                            "final_format": final_format,
-                            "clean_base64_length": len(clean_base64),
-                            "final_url_length": len(final_url),
-                            "final_url_prefix": final_url[:60]
-                        },
-                        "sessionId": "debug-session",
-                        "runId": "run1",
-                        "hypothesisId": "A,B"
-                    }) + "\n")
-            except Exception:
-                pass
-            # #endregion
+            # Debug logging removed - not needed in production
             
             messages[1]["content"].append({
                 "type": "image_url",
@@ -675,31 +629,7 @@ class VisionLLMClient(BaseLLMClient):
         
         # Add image if provided
         if image_base64:
-            # #region agent log
-            import json as _json_debug
-            import time as _time_debug
-            _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-            try:
-                with open(_debug_log_path, "a") as _f:
-                    _f.write(_json_debug.dumps({
-                        "id": f"log_{int(_time_debug.time() * 1000)}",
-                        "timestamp": int(_time_debug.time() * 1000),
-                        "location": "vision_llm.py:_generate_google",
-                        "message": "Image base64 BEFORE processing (Google)",
-                        "data": {
-                            "image_base64_length": len(image_base64) if image_base64 else 0,
-                            "has_data_uri_prefix": image_base64.startswith("data:image") if image_base64 else False,
-                            "prefix": image_base64[:50] if image_base64 and len(image_base64) > 50 else image_base64,
-                            "is_none": image_base64 is None,
-                            "is_empty": image_base64 == "" if image_base64 else True
-                        },
-                        "sessionId": "debug-session",
-                        "runId": "run1",
-                        "hypothesisId": "A,C,D"
-                    }) + "\n")
-            except Exception:
-                pass
-            # #endregion
+            # Debug logging removed - not needed in production
             
             try:
                 # Strip data URI prefix if present (hypothesis A fix)

src/gepa_optimizer/utils/pareto_logger.py

@@ -106,13 +106,7 @@ class ParetoLogger:
                 logger.error(f"      Cannot evaluate {cand_notation} without baseline f(S₀)")
                 logger.error(f"      💡 Seed prompt must be evaluated on Dpareto first")
                 logger.error(f"      💡 Rejecting candidate to maintain correctness")
-                # #region agent log
-                import json as _json_debug
-                import time as _time_debug
-                _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-                with open(_debug_log_path, "a") as _f:
-                    _f.write(_json_debug.dumps({"hypothesisId": "B", "location": "pareto_logger.py:baseline_not_set", "message": "CRITICAL: Baseline not set when checking Pareto", "data": {"candidate_type": candidate_type, "candidate_notation": cand_notation, "score": score}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-                # #endregion
+            # Debug logging removed - not needed in production
                 return  # Reject candidate - baseline is required
         
         # Check if this candidate dominates any in current front
@@ -450,12 +444,6 @@ def reset_pareto_logger() -> ParetoLogger:
     """Reset global Pareto logger instance (for new runs)"""
     global _pareto_logger
     _pareto_logger = ParetoLogger()
-    # #region agent log
-    import json as _json_debug
-    import time as _time_debug
-    _debug_log_path = "/Users/suhas/Desktop/Projects/Prompt-Optimizer/.cursor/debug.log"
-    with open(_debug_log_path, "a") as _f:
-        _f.write(_json_debug.dumps({"hypothesisId": "E", "location": "pareto_logger.py:reset", "message": "Pareto logger reset", "data": {"baseline_score": _pareto_logger.baseline_score, "pareto_front_size": len(_pareto_logger.pareto_front)}, "timestamp": int(_time_debug.time() * 1000), "sessionId": "debug-session"}) + "\n")
-    # #endregion
+            # Debug logging removed - not needed in production
     return _pareto_logger