feat: implement context_builder node with Primary LLM integration

RAGManager/app/agents/nodes/context_builder.py

@@ -1,10 +1,19 @@
 """Nodo 6: Context Builder - Enriches query with retrieved context."""
 
+import logging
+
 from app.agents.state import AgentState
-from langchain_core.messages import SystemMessage
+from app.core.config import settings
+from langchain_core.messages import HumanMessage, SystemMessage
 from langchain_openai import ChatOpenAI
 
-llm = ChatOpenAI(model="gpt-5-nano")
+logger = logging.getLogger(__name__)
+
+# Initialize Primary LLM with configuration from settings
+llm = ChatOpenAI(
+    model="gpt-4.1-mini",  # Primary LLM model for context-aware responses
+    openai_api_key=settings.openai_api_key,
+)
 
 
 def context_builder(state: AgentState) -> AgentState:
@@ -23,30 +32,82 @@ def context_builder(state: AgentState) -> AgentState:
     Returns:
         Updated state with enriched_query and primary_response set
     """
-    # TODO: Implement context building and primary LLM call
-    # This should:
-    # 1. Combine paraphrased_text with relevant_chunks into enriched_query
-    # 2. Format the query appropriately (e.g., with system prompts, context sections)
-    # 3. Call Primary LLM with the enriched query
-    # 4. Store the LLM response in primary_response
-
-    # Placeholder: For now, we'll create a simple enriched query
     updated_state = state.copy()
+
+    # Get paraphrased text and relevant chunks from state
     paraphrased = state.get("paraphrased_text", "")
     chunks = state.get("relevant_chunks", [])
-
+
+    # Fallback if paraphrased_text is not available
+    if not paraphrased:
+        logger.warning("No paraphrased text found in state. Using original prompt.")
+        messages = state.get("messages", [])
+        if messages:
+            paraphrased = messages[-1].content if hasattr(messages[-1], "content") else str(messages[-1])
+        else:
+            paraphrased = state.get("prompt", "")
+
     # Build enriched query with context
-    context_section = "\n\n".join(chunks) if chunks else "No relevant context found."
+    if chunks:
+        context_section = "\n\n---\n\n".join([f"Context {i+1}:\n{chunk}" for i, chunk in enumerate(chunks)])
+    else:
+        context_section = "No relevant context found in the knowledge base."
+        logger.warning("No relevant chunks found for context building")
 
-    system_content = f"""You are a helpful assistant. Use the following context to answer the user's question.
-If the answer is not in the context, say you don't know.
+    # Create enriched query combining paraphrased text and context
+    enriched_query = f"""User Question: {paraphrased}
 
-Context:
-{context_section}"""
+Relevant Context from Knowledge Base:
+{context_section}
 
-    messages = [SystemMessage(content=system_content)] + state["messages"]
+Please provide a comprehensive answer based on the context provided above. If the context does not contain enough information to answer the question, please indicate that clearly."""
+
+    # System prompt for the Primary LLM
+    system_content = """You are a helpful assistant specialized in providing accurate, context-based answers about nutrition and culinary topics.
 
-    # Call Primary LLM
-    response = llm.invoke(messages)
+Your task is to:
+1. Use the provided context to answer the user's question accurately
+2. If the context contains relevant information, provide a comprehensive answer
+3. If the context does not contain enough information, clearly state that you don't have sufficient information in the knowledge base
+4. Always base your answer on the provided context - do not make up information
+5. If the question is not related to the context, politely redirect the conversation
 
-    return {"messages": [response]}
+Be concise, accurate, and helpful."""
+
+    # Prepare messages for LLM
+    messages_for_llm = [
+        SystemMessage(content=system_content),
+        HumanMessage(content=enriched_query)
+    ]
+
+    try:
+        # Call Primary LLM
+        logger.info("Calling Primary LLM with enriched query")
+        response = llm.invoke(messages_for_llm)
+
+        # Extract response content
+        primary_response = response.content if hasattr(response, "content") else str(response)
+
+        # Update state with enriched query and primary response (as defined in state.py)
+        updated_state["enriched_query"] = enriched_query
+        updated_state["primary_response"] = primary_response
+
+        # Also set generated_response for guard_final (Nodo 7: Generator not yet implemented)
+        # For now, generated_response = primary_response
+        # This allows guard_final to validate the response
+        updated_state["generated_response"] = primary_response
+
+        # Also update messages for LangGraph compatibility
+        updated_state["messages"] = state.get("messages", []) + [response]
+
+        logger.info("Successfully generated primary response from LLM")
+
+    except Exception as e:
+        logger.error(f"Error calling Primary LLM: {e}", exc_info=True)
+        # Set error state
+        updated_state["error_message"] = f"Error in context builder: {str(e)}"
+        updated_state["enriched_query"] = enriched_query
+        updated_state["primary_response"] = None
+        updated_state["generated_response"] = None
+
+    return updated_state