Update core/evolution_engine.py

core/evolution_engine.py

@@ -1,80 +1,57 @@
 # algoforge_prime/core/evolution_engine.py
 print("DEBUG: Importing core.evolution_engine")
 
-# --- Corrected Imports ---
-# Absolute imports for modules outside the 'core' package
-from prompts.system_prompts import get_system_prompt
-
-# Absolute imports for other modules within the 'core' package (or relative for siblings)
 from core.llm_clients import call_huggingface_api, call_gemini_api, LLMResponse
-
-# Relative import for a sibling module within the 'core' package
-# from .safe_executor import ExecutionResult # Not directly used in this module, but evaluation_output_obj might contain it
-# from .evaluation_engine import EvaluationResultOutput # For type hinting the parameter
+from prompts.system_prompts import get_system_prompt
+# from core.evaluation_engine import EvaluationResultOutput # For type hinting if needed
 
 print("DEBUG: core.evolution_engine - Imports successful")
 
 def evolve_solution(
     original_solution_text: str,
-    evaluation_output_obj, # This object comes from evaluation_engine and contains EvaluationResultOutput
-                           # It will have a .get_display_critique() method and .combined_score attribute
+    evaluation_output_obj, # This is an EvaluationResultOutput object
     problem_description: str,
     problem_type: str,
-    llm_client_config: dict # {"type": ..., "model_id": ..., "temp": ..., "max_tokens": ...}
-) -> str: # Returns evolved solution text or an error string
-    """
-    Attempts to evolve a solution based on its comprehensive evaluation details.
-    """
+    llm_client_config: dict
+) -> str:
     print(f"DEBUG: evolution_engine.py - Evolving solution. Problem type: {problem_type}")
-    system_p_evolve = get_system_prompt("evolution_general") # problem_type can be used for specialization here too
+    system_p_evolve = get_system_prompt("evolution_general")
     
-    # Extract necessary info from the evaluation_output_obj
-    # This assumes evaluation_output_obj is an instance of EvaluationResultOutput from evaluation_engine.py
-    # or at least has these attributes/methods.
     try:
+        # Use the method from EvaluationResultOutput to get formatted critique and test results
         critique_and_test_feedback = evaluation_output_obj.get_display_critique() 
         original_score = evaluation_output_obj.combined_score
-    except AttributeError as e:
-        print(f"ERROR: evolution_engine.py - evaluation_output_obj is missing expected attributes/methods: {e}")
-        # Fallback if the object structure is not as expected
-        critique_and_test_feedback = "Critique data was not in the expected format."
-        original_score = 0 # Assign a neutral score if real one can't be found
+    except AttributeError: # Fallback if evaluation_output_obj is not as expected
+        critique_and_test_feedback = "Detailed evaluation feedback was not available or malformed."
+        original_score = 0 # Or try to get it from evaluation_output_obj if it's just a simple dict
+        if hasattr(evaluation_output_obj, 'score'): original_score = evaluation_output_obj.score
+        elif isinstance(evaluation_output_obj, dict) and 'score' in evaluation_output_obj: original_score = evaluation_output_obj['score']
 
     user_p_evolve = (
         f"Original Problem Context: \"{problem_description}\"\n\n"
         f"The solution to be evolved achieved a combined score of {original_score}/10.\n"
         f"Here is the original solution text:\n```python\n{original_solution_text}\n```\n\n"
-        f"Here is the comprehensive evaluation it received (including LLM critique and automated test feedback if run):\n'''\n{critique_and_test_feedback}\n'''\n\n"
+        f"Here is the comprehensive evaluation it received (including LLM critique AND automated test feedback/errors if run):\n'''\n{critique_and_test_feedback}\n'''\n\n"
         f"Your Task: Based on ALL the information above (solution, LLM critique, and crucially any test execution results/errors mentioned in the evaluation), "
         f"evolve the provided solution to make it demonstrably superior. "
-        f"Prioritize fixing any reported execution errors or failed tests. "
+        f"**Your HIGHEST PRIORITY is to fix any reported execution errors or failed tests.** "
         f"Then, address other critique points like efficiency, clarity, or completeness. "
-        f"Output the *complete evolved solution*. "
-        f"Follow this with a brief explanation of the key changes and improvements you implemented, especially how you addressed test failures or execution issues."
+        f"Output ONLY the *complete, raw, evolved Python code block*. Do not include explanations outside the code block unless explicitly part of the solution's comments."
     )
 
-    llm_response_obj = None # type: LLMResponse
+    llm_response_obj = None
     if llm_client_config["type"] == "hf":
-        llm_response_obj = call_huggingface_api(
-            user_p_evolve, llm_client_config["model_id"],
-            temperature=llm_client_config["temp"], max_new_tokens=llm_client_config["max_tokens"],
-            system_prompt_text=system_p_evolve
-        )
+        llm_response_obj = call_huggingface_api(user_p_evolve, llm_client_config["model_id"], llm_client_config["temp"], llm_client_config["max_tokens"], system_p_evolve)
     elif llm_client_config["type"] == "google_gemini":
-        llm_response_obj = call_gemini_api(
-            user_p_evolve, llm_client_config["model_id"],
-            temperature=llm_client_config["temp"], max_new_tokens=llm_client_config["max_tokens"],
-            system_prompt_text=system_p_evolve
-        )
+        llm_response_obj = call_gemini_api(user_p_evolve, llm_client_config["model_id"], llm_client_config["temp"], llm_client_config["max_tokens"], system_p_evolve)
     else:
-        error_msg = f"ERROR (Evolution): Unknown LLM client type '{llm_client_config['type']}'"
-        print(f"ERROR: evolution_engine.py - {error_msg}")
-        return error_msg
+        return f"ERROR (Evolution): Unknown LLM client type '{llm_client_config['type']}'"
 
     if llm_response_obj.success:
-        return llm_response_obj.text
+        # Optional: basic cleanup of the LLM output if it tends to add markdown
+        from core.utils import basic_text_cleanup # Assuming you have this
+        return basic_text_cleanup(llm_response_obj.text)
     else:
-        # Error is already logged by call_..._api functions if it's from there
         return f"ERROR (Evolution with {llm_response_obj.model_id_used}): {llm_response_obj.error}"
 
 print("DEBUG: core.evolution_engine - Module fully defined.")