Merge pull request #28 from Andrei-Constantin-Programmer/IBM-21-Explainer_Agent

Ibm 21 explainer agent

Author: Andrei-Constantin-Programmer

AntiPattern_Remediator/main.py

@@ -1,9 +1,7 @@
-
 """
 Main entry point - Legacy Code Migration Tool
 """
 from config.settings import initialize_settings
-# from scripts import seed_database
 from dotenv import load_dotenv
 load_dotenv()
 from colorama import Fore, Style
@@ -12,11 +10,35 @@
 from full_repo_workflow import run_full_repo_workflow
 
 
-def run_code_snippet_workflow(settings, db_manager, prompt_manager, langgraph):
-    """Run the original workflow with a hardcoded Java code snippet."""
-    print(Fore.BLUE + "\n=== Code Snippet Analysis Workflow ===" + Style.RESET_ALL)
-    print("Analyzing the provided Java code snippet...")
-    
+
+def main():
+    """Main function: Run antipattern analysis"""
+
+    # Let user select provider
+    print("Available providers: 1) ollama  2) ibm  3) vllm")
+    choice = input("Select provider (1-3): ").strip()
+
+    provider_map = {"1": "ollama", "2": "ibm", "3": "vllm"}
+    provider = provider_map.get(choice, "ollama")  # default to ollama
+
+    # Let us choose which DB to interact with
+    print("Choose your trove: 1) ChromaDB (VectorDB) 2) TinyDB (DocumentDB)")
+    db_choice = input("Choose 1 or 2: ").strip()
+
+    # Initialize global settings with selected provider
+    settings = initialize_settings(provider)
+    print(Fore.GREEN + f"Using {settings.LLM_PROVIDER} with model {settings.LLM_MODEL}" + Style.RESET_ALL)
+
+    # Temporary Lazy Imports
+    from src.core.graph import CreateGraph
+    from src.data.database import VectorDBManager, TinyDBManager
+    from src.core.prompt import PromptManager
+    from scripts import seed_database
+
+    # Initialize PromptManager
+    print("Initializing PromptManager...")
+    prompt_manager = PromptManager()
+
     # Example Java code
     legacy_code = """
     public class ApplicationManager {
@@ -55,7 +77,8 @@ def run_code_snippet_workflow(settings, db_manager, prompt_manager, langgraph):
         }
     }
     """
-    
+
+    # Initial workflow state
     initial_state = {
         "code": legacy_code,
         "context": None,
@@ -66,58 +89,12 @@ def run_code_snippet_workflow(settings, db_manager, prompt_manager, langgraph):
         "code_review_results": None,
         "code_review_times": 0,
         "msgs": [],
-        "answer": None
-    }
-
-    final_state = langgraph.invoke(initial_state)
-
-    print(Fore.GREEN + f"\nAnalysis Complete!" + Style.RESET_ALL)
-    print(f"Final state keys: {list(final_state.keys())}")
-    print(f"Context retrieved: {'Yes' if final_state.get('context') else 'No'}")
-    print(f"Analysis completed: {'Yes' if final_state.get('antipatterns_scanner_results') else 'No'}")
-    print(f"Refactored code: {'Yes' if final_state.get('refactored_code') else 'No'}")
-    print(f"Code review results: {final_state.get('code_review_times')}")
-
-
-def main():
-    """Main function: Choose between code snippet analysis or full repository run"""
-
-    print(Fore.BLUE + "=== AntiPattern Remediator Tool ===" + Style.RESET_ALL)
-    print("Choose your analysis mode:")
-    print("1) Code Snippet Analysis - Analyze a sample Java code snippet")
-    print("2) Full Repository Run - Process files with 100% test coverage from JaCoCo results")
-    
-    # Let user choose analysis mode
-    mode_choice = input("\nSelect mode (1-2): ").strip()
-    
-    if mode_choice not in ["1", "2"]:
-        print(Fore.RED + "Invalid choice. Defaulting to Code Snippet Analysis." + Style.RESET_ALL)
-        mode_choice = "1"
-    
-    # Let user select provider
-    print("\nAvailable providers: 1) ollama  2) ibm  3) vllm")
-    choice = input("Select provider (1-3): ").strip()
-
-    provider_map = {"1": "ollama", "2": "ibm", "3": "vllm"}
-    provider = provider_map.get(choice, "ollama")  # default to ollama
-
-    # Let us choose which DB to interact with
-    print("Choose your trove: 1) ChromaDB (VectorDB) 2) TinyDB (DocumentDB)")
-    db_choice = input("Choose 1 or 2: ").strip()
-
-    # Initialize global settings with selected provider
-    settings = initialize_settings(provider)
-    print(Fore.GREEN + f"Using {settings.LLM_PROVIDER} with model {settings.LLM_MODEL}" + Style.RESET_ALL)
+        "answer": None,
 
-    # Temporary Lazy Imports
-    from src.core.graph import CreateGraph
-    from src.data.database import VectorDBManager, TinyDBManager
-    from src.core.prompt import PromptManager
-    from scripts import seed_database
-
-    # Initialize PromptManager
-    print("Initializing PromptManager...")
-    prompt_manager = PromptManager()
+        # ExplainerAgent fields
+        "explanation_response_raw": None,
+        "explanation_json": None,
+    }
 
     # Setup Database
     if db_choice == "2":
@@ -133,11 +110,22 @@ def main():
     retriever = db_manager.as_retriever()
     langgraph = CreateGraph(db_manager, prompt_manager, retriever=retriever).workflow
 
-    # Run the selected workflow
-    if mode_choice == "1":
-        run_code_snippet_workflow(settings, db_manager, prompt_manager, langgraph)
+    # Final results summary
+    print(Fore.GREEN + f"\nAnalysis Complete!" + Style.RESET_ALL)
+    print(f"Final state keys: {list(final_state.keys())}")
+    print(f"Context retrieved: {'Yes' if final_state.get('context') else 'No'}")
+    print(f"Analysis completed: {'Yes' if final_state.get('antipatterns_scanner_results') else 'No'}")
+    print(f"Refactored code: {'Yes' if final_state.get('refactored_code') else 'No'}")
+    print(f"Code review results: {final_state.get('code_review_times')}")
+
+    # Show explanation from ExplainerAgent
+    if final_state.get("explanation_json"):
+        import json
+        print(Fore.CYAN + "\n=== Explanation (JSON) ===" + Style.RESET_ALL)
+        print(json.dumps(final_state["explanation_json"], indent=2, ensure_ascii=False))
     else:
-        run_full_repo_workflow(settings, db_manager, prompt_manager, langgraph)
+        print(Fore.RED + "\nNo explanation was generated." + Style.RESET_ALL)
+
 
 if __name__ == "__main__":
     main()

AntiPattern_Remediator/src/core/agents/__init__.py

@@ -7,10 +7,12 @@
 from .refactor_strategist import RefactorStrategist
 from .code_transformer import CodeTransformer
 from .code_reviewer import CodeReviewerAgent
+from .explainer import ExplainerAgent
 
 __all__ = [
     "AntipatternScanner", 
     "RefactorStrategist",
     "CodeTransformer",
-    "CodeReviewerAgent"
+    "CodeReviewerAgent",
+    "ExplainerAgent"
 ]

AntiPattern_Remediator/src/core/agents/explainer.py

@@ -0,0 +1,131 @@
+"""
+ExplainerAgent — minimal version
+- Delegates state handling to create_graph.py
+- Uses PromptManager if available; otherwise a tiny inline fallback prompt
+- Always passes msgs; always returns a non-empty explanation_json
+"""
+from __future__ import annotations
+
+from typing import Dict, Any
+import json
+
+from langchain_core.language_models import BaseLanguageModel
+from langchain.prompts import ChatPromptTemplate, MessagesPlaceholder
+from ..prompt import PromptManager
+from src.core.utils import extract_first_json
+
+PROMPT_KEY = "explainer"
+
+
+class ExplainerAgent:
+    def __init__(self, llm: BaseLanguageModel, prompt_manager: PromptManager):
+        self.llm = llm
+        self.prompt_manager = prompt_manager
+
+    def explain_antipattern(self, state: Dict[str, Any]) -> Dict[str, Any]:
+        """Generate explanation JSON for detected antipatterns and refactor."""
+        kwargs = dict(
+            code=state.get("code", ""),
+            language=state.get("language", "Java"),
+            context=state.get("context", ""),
+            refactored_code=state.get("refactored_code", ""),
+            refactor_rationale=state.get("refactor_rationale", ""),
+            antipattern_name=state.get("antipattern_name", "Unknown antipattern"),
+            antipattern_description=state.get("antipattern_description", ""),
+            antipatterns_json=json.dumps(state.get("antipatterns_json", []), ensure_ascii=False),
+            msgs=state.get("msgs", []),  # ensure MessagesPlaceholder is satisfied
+        )
+
+        messages = self._build_messages(**kwargs)
+
+        try:
+            response = self.llm.invoke(messages)
+            raw = getattr(response, "content", None) or str(response)
+        except Exception as e:
+            raw = f"LLM error: {e}"
+
+        state["explanation_response_raw"] = raw
+
+        # Robust parse: accept dict, wrap list, or emit a minimal fallback
+        try:
+            parsed = extract_first_json(raw)
+        except Exception:
+            parsed = None
+
+        if isinstance(parsed, dict):
+            state["explanation_json"] = parsed
+        elif isinstance(parsed, list):
+            state["explanation_json"] = {"items": parsed}
+        else:
+            state["explanation_json"] = self._fallback_payload(state)
+
+        return state
+
+    def display_explanation(self, state: Dict[str, Any]) -> Dict[str, Any]:
+        print("\n=== Explanation (raw) ===\n", state.get("explanation_response_raw", "N/A"))
+        if state.get("explanation_json"):
+            print("\n=== Explanation (JSON) ===\n",
+                  json.dumps(state["explanation_json"], indent=2, ensure_ascii=False))
+        return state
+
+    def _build_messages(self, **kwargs) -> Any:
+        # Always ensure msgs exists
+        if "msgs" not in kwargs or kwargs["msgs"] is None:
+            kwargs = {**kwargs, "msgs": []}
+
+        # 1) Try preloaded template from PromptManager
+        prompt = None
+        getp = getattr(self.prompt_manager, "get_prompt", None)
+        if callable(getp):
+            prompt = getp(PROMPT_KEY)
+        if prompt is not None:
+            return prompt.format_messages(**kwargs)
+
+        # 2) Minimal inline fallback
+        schema = {
+            "items": [{
+                "antipattern_name": "",
+                "antipattern_description": "",
+                "impact": "",
+                "why_it_is_bad": "",
+                "how_we_fixed_it": "",
+                "refactored_code": "",
+                "summary": ""
+            }],
+            "what_changed": [],
+            "why_better": [],
+            "principles_applied": [],
+            "trade_offs": [],
+            "closing_summary": ""
+        }
+        content = (
+            "Given inputs (JSON):\n" + json.dumps({k: v for k, v in kwargs.items() if k != "msgs"}, ensure_ascii=False) +
+            "\nRespond with STRICT JSON using exactly this schema:\n" +
+            json.dumps(schema, ensure_ascii=False)
+        )
+        fallback = ChatPromptTemplate.from_messages([
+            ("system", "Return STRICT JSON only. No commentary."),
+            ("user", content),
+            MessagesPlaceholder("msgs"),
+        ])
+        return fallback.format_messages(**kwargs)
+
+    @staticmethod
+    def _fallback_payload(state: Dict[str, Any]) -> Dict[str, Any]:
+        """Tiny fallback so downstream never breaks if parsing fails."""
+        return {
+            "items": [{
+                "antipattern_name": state.get("antipattern_name", "Unknown antipattern"),
+                "antipattern_description": state.get("antipattern_description", ""),
+                "impact": "",
+                "why_it_is_bad": "",
+                "how_we_fixed_it": state.get("refactor_rationale", ""),
+                "refactored_code": state.get("refactored_code", ""),
+                "summary": "Auto-generated minimal explanation (parser fallback)."
+            }],
+            "what_changed": [],
+            "why_better": [],
+            "principles_applied": [],
+            "trade_offs": [],
+            "closing_summary": ""
+        }

AntiPattern_Remediator/src/core/graph/create_graph.py

@@ -14,6 +14,7 @@
 from ..agents import RefactorStrategist
 from ..agents import CodeTransformer
 from ..agents import CodeReviewerAgent
+from ..agents import ExplainerAgent
 
 # Imports for LangSmith tracing
 import os
@@ -62,7 +63,7 @@ def __init__(self, db_manager, prompt_manager: PromptManager, retriever=None, ll
         self.prompt_manager = prompt_manager
         self.conditional_edges = ConditionalEdges()
 
-        # ✅ assign the instance attribute before use
+        # assign the instance attribute before use
         self.retriever = retriever or self.db_manager.as_retriever()
 
         retriever_tool = create_retriever_tool(
@@ -78,13 +79,13 @@ def __init__(self, db_manager, prompt_manager: PromptManager, retriever=None, ll
             "strategist": RefactorStrategist(self.llm, self.prompt_manager, retriever=self.retriever),
             "transformer": CodeTransformer(self.llm, self.prompt_manager),
             "reviewer": CodeReviewerAgent(self.llm, self.prompt_manager),
+            "explainer": ExplainerAgent(self.llm, self.prompt_manager)
         }
 
         # Build the LangGraph workflow
         self.workflow = self._build_graph()
 
     def _build_graph(self):
-        """Build LangGraph workflow"""
         graph = StateGraph(AgentState)
 
         # Scanner: retrieve + analyze
@@ -104,6 +105,10 @@ def _build_graph(self):
         graph.add_node("review_code", self.agents["reviewer"].review_code)
         graph.add_node("display_code_review_results", self.agents["reviewer"].display_code_review_results)
 
+        # Explainer: final storytelling
+        graph.add_node("explain_antipattern", self.agents["explainer"].explain_antipattern)
+        graph.add_node("display_explanation", self.agents["explainer"].display_explanation)
+
         # Topology
         graph.set_entry_point("retrieve_context")
         graph.add_edge("retrieve_context", "analyze_antipatterns")
@@ -123,6 +128,8 @@ def _build_graph(self):
             },
         )
 
-        graph.add_edge("display_code_review_results", END)
+        graph.add_edge("display_code_review_results", "explain_antipattern")
+        graph.add_edge("explain_antipattern", "display_explanation")
+        graph.add_edge("display_explanation", END)
 
         return graph.compile()