Update workflow.py

workflow.py

@@ -24,6 +24,8 @@ class AgentState(TypedDict):
 class ResearchWorkflow:
     """
     Defines a multi-step research workflow using a state graph.
+    This workflow now employs a Retrieval-Augmented Generation (RAG) approach.
+    When no documents are retrieved, the model dynamically synthesizes an analysis using its internal knowledge.
     """
     def __init__(self) -> None:
         self.processor = EnhancedCognitiveProcessor()
@@ -58,7 +60,12 @@ class ResearchWorkflow:
             query = state["messages"][-1].content
             # Retrieve the domain from the state's context (defaulting to Biomedical Research)
             domain = state.get("context", {}).get("domain", "Biomedical Research")
-            new_context = {"raw_query": query, "domain": domain, "refine_count": 0, "refinement_history": []}
+            new_context = {
+                "raw_query": query,
+                "domain": domain,
+                "refine_count": 0,
+                "refinement_history": []
+            }
             logger.info(f"Query ingested. Domain: {domain}")
             return {
                 "messages": [AIMessage(content="Query ingested successfully")],
@@ -73,7 +80,7 @@ class ResearchWorkflow:
         try:
             query = state["context"]["raw_query"]
             # For demonstration, we use an empty document list.
-            # Replace this with actual retrieval logic as needed.
+            # In a full RAG system, you would retrieve relevant documents from an external index.
             docs = []
             logger.info(f"Retrieved {len(docs)} documents for query.")
             return {
@@ -93,28 +100,25 @@ class ResearchWorkflow:
     def analyze_content(self, state: Dict) -> Dict:
         try:
             domain = state["context"].get("domain", "Biomedical Research").strip().lower()
-            fallback_analyses = ResearchConfig.DOMAIN_FALLBACKS
-            # If the domain matches a fallback, return it.
-            if domain in fallback_analyses:
-                logger.info(f"Using fallback analysis for domain: {state['context'].get('domain')}")
-                return {
-                    "messages": [AIMessage(content=fallback_analyses[domain].strip())],
-                    "context": state["context"]
-                }
-            else:
-                docs = state["context"].get("documents", [])
+            docs = state["context"].get("documents", [])
+            # Use retrieved documents if available; otherwise, use the raw query.
+            if docs:
                 docs_text = "\n\n".join([d.page_content for d in docs])
-                domain_prompt = ResearchConfig.DOMAIN_PROMPTS.get(domain, "")
-                full_prompt = f"{domain_prompt}\n\n" + ResearchConfig.ANALYSIS_TEMPLATE.format(context=docs_text)
-                response = self.processor.process_query(full_prompt)
-                if "error" in response:
-                    logger.error("Backend response error during analysis.")
-                    return self._error_state(response["error"])
-                logger.info("Content analysis completed.")
-                return {
-                    "messages": [AIMessage(content=response.get('choices', [{}])[0].get('message', {}).get('content', ''))],
-                    "context": state["context"]
-                }
+            else:
+                docs_text = state["context"].get("raw_query", "")
+                logger.info("No documents retrieved; switching to dynamic RAG mode.")
+            domain_prompt = ResearchConfig.DOMAIN_PROMPTS.get(domain, "")
+            # RAG approach: combine domain prompt with retrieved docs or raw query.
+            full_prompt = f"{domain_prompt}\n\n" + ResearchConfig.ANALYSIS_TEMPLATE.format(context=docs_text)
+            response = self.processor.process_query(full_prompt)
+            if "error" in response:
+                logger.error("Backend response error during analysis.")
+                return self._error_state(response["error"])
+            logger.info("Content analysis completed using RAG approach.")
+            return {
+                "messages": [AIMessage(content=response.get('choices', [{}])[0].get('message', {}).get('content', ''))],
+                "context": state["context"]
+            }
         except Exception as e:
             logger.exception("Error during content analysis.")
             return self._error_state(f"Analysis Error: {str(e)}")