update all (#530)

rdagent/components/coder/data_science/ensemble/eval.py

@@ -64,12 +64,22 @@ class EnsembleCoSTEEREvaluator(CoSTEEREvaluator):
         implementation.inject_files(**{fname: test_code})
         stdout = implementation.execute(env=de, entry=f"python {fname}")
 
+        if "main.py" in implementation.file_dict:
+            workflow_stdout = implementation.execute(env=de, entry="python main.py")
+        else:
+            workflow_stdout = None
+
         system_prompt = T(".prompts:ensemble_eval.system").r(
             task_desc=target_task_information,
             test_code=test_code,
             code=implementation.file_dict["ensemble.py"],
+            workflow_stdout=workflow_stdout,
+            workflow_code=implementation.all_codes,
+        )
+        user_prompt = T(".prompts:ensemble_eval.user").r(
+            stdout=stdout,
+            workflow_stdout=workflow_stdout,
         )
-        user_prompt = T(".prompts:ensemble_eval.user").r(stdout=stdout)
 
         resp = APIBackend().build_messages_and_create_chat_completion(user_prompt, system_prompt, json_mode=True)
         return EnsembleEvalFeedback(**json.loads(resp))

rdagent/components/coder/data_science/ensemble/prompts.yaml

@@ -67,6 +67,13 @@ ensemble_eval:
     {{test_code}}
     ```
     
+    {% if workflow_stdout is not none %}
+    Your ensemble code is also part of the whole workflow, the user also tested the whole workflow and provided you the stdout.
+    The whole workflow code is:
+    {{workflow_code}}
+    Please consider both stdout and approve the code when both the ensemble test and the whole workflow test pass.
+    {% endif %}
+    
     You'll be given the stdout of your testing scripts.
     Please respond with your feedback in the following JSON format: 
     {
@@ -76,4 +83,9 @@ ensemble_eval:
         "final_decision": <true/false>
     }    
   user: |-    
+    Ensemble test stdout:
     {{stdout}}   
+    {% if workflow_stdout is not none %}
+    Whole workflow test stdout:
+    {{workflow_stdout}}
+    {% endif %}