fix: fix a bug in mini case (#389)

* fix: fix a bug in mini case

* fix a bug

rdagent/app/kaggle/conf.py

@@ -66,7 +66,7 @@ class KaggleBasePropSetting(BasePropSetting):
 
     auto_submit: bool = True
 
-    dsagent: bool = False
+    mini_case: bool = False
 
 
 KAGGLE_IMPLEMENT_SETTING = KaggleBasePropSetting()

rdagent/scenarios/kaggle/experiment/scenario.py

@@ -37,7 +37,7 @@ class KGScenario(Scenario):
         self.evaluation_desc = None
         self.evaluation_metric_direction = None
         self.vector_base = None
-        self.dsagent = KAGGLE_IMPLEMENT_SETTING.dsagent
+        self.mini_case = KAGGLE_IMPLEMENT_SETTING.mini_case
         self._analysis_competition_description()
         self.if_action_choosing_based_on_UCB = KAGGLE_IMPLEMENT_SETTING.if_action_choosing_based_on_UCB
         self.if_using_graph_rag = KAGGLE_IMPLEMENT_SETTING.if_using_graph_rag

rdagent/scenarios/kaggle/kaggle_crawler.py

@@ -117,13 +117,13 @@ def notebook_to_knowledge(notebook_text: str) -> str:
 
     sys_prompt = (
         Environment(undefined=StrictUndefined)
-        .from_string(prompt_dict["gen_knowledge_from_code_DSAgent"]["system"])
+        .from_string(prompt_dict["gen_knowledge_from_code_mini_case"]["system"])
         .render()
     )
 
     user_prompt = (
         Environment(undefined=StrictUndefined)
-        .from_string(prompt_dict["gen_knowledge_from_code_DSAgent"]["user"])
+        .from_string(prompt_dict["gen_knowledge_from_code_mini_case"]["user"])
         .render(notebook=notebook_text)
     )
 
@@ -203,7 +203,7 @@ def collect_knowledge_texts(local_path: str = "/data/userdata/share/kaggle") ->
 
 # %%
 if __name__ == "__main__":
-    dsagent_cs = [
+    mini_case_cs = [
         "feedback-prize-english-language-learning",
         "playground-series-s3e11",
         "playground-series-s3e14",
@@ -254,7 +254,7 @@ if __name__ == "__main__":
         "facebook-v-predicting-check-ins",
     ]
 
-    all_cs = dsagent_cs + other_cs
+    all_cs = mini_case_cs + other_cs
     for c in all_cs:
         convert_notebooks_to_text(c)
     exit()

rdagent/scenarios/kaggle/knowledge_management/vector_base.py

@@ -267,12 +267,12 @@ class KaggleExperienceBase(PDVectorBase):
 
 if __name__ == "__main__":
     kaggle_base = KaggleExperienceBase(
-        kaggle_experience_path="git_ignore_folder/data-dsagent/kaggle_experience_results.json"
+        kaggle_experience_path="git_ignore_folder/data_minicase/kaggle_experience_results.json"
     )
 
     kaggle_base.add_experience_to_vector_base()
 
-    kaggle_base.save("git_ignore_folder/experience/tabular_cases/kaggle_vector_base.pkl")
+    kaggle_base.save("git_ignore_folder/vector_base/kaggle_vector_base.pkl")
 
     print(f"There are {kaggle_base.shape()[0]} records in the vector base.")
 

rdagent/scenarios/kaggle/prompts.yaml

@@ -316,7 +316,7 @@ model_feature_selection:
       {{ feature }}
     {% endfor %}
 
-gen_knowledge_from_code_DSAgent:
+gen_knowledge_from_code_mini_case:
   system: |-
     You were a proficient data scientist.
   user: |-

rdagent/scenarios/kaggle/proposal/proposal.py

@@ -91,8 +91,8 @@ class KGHypothesisGen(ModelHypothesisGen):
 
     def generate_RAG_content(self, trace: Trace, hypothesis_and_feedback: str) -> str:
         if self.scen.if_using_vector_rag:
-            if self.scen.dsagent:
-                rag_results, _ = self.scen.dsagent.vector_base.search_experience(hypothesis_and_feedback, topk_k=1)
+            if self.scen.mini_case:
+                rag_results, _ = self.scen.vector_base.search_experience(hypothesis_and_feedback, topk_k=1)
             else:
                 rag_results, _ = self.scen.vector_base.search_experience(hypothesis_and_feedback, topk_k=5)
             return "\n".join([doc.content for doc in rag_results])