fix: match verbatim notebook configuration and extract raw keywords with UI formatting

app.py

@@ -9,7 +9,7 @@ from langchain_core.prompts import ChatPromptTemplate
 from langchain_openai import ChatOpenAI
 
 # ==========================================================
-# 1. LOAD DATA & EXACT NOTEBOOK PARSING INDEX GENERATOR
+# 1. DATA INGESTION & NOTEBOOK INDEX GENERATOR
 # ==========================================================
 with open("gcmd_hierarchy.json", "r") as f:
     gcmd_data = json.load(f)
@@ -17,7 +17,6 @@ with open("gcmd_hierarchy.json", "r") as f:
 def build_gcmd_indices(gcmd_json):
     """
     Parses the GCMD hierarchy exactly as done in the working notebook.
-    Omits high-level 'EARTH SCIENCE' and 'Topic' prefixes from the path string.
     """
     topic_list = []
     sub_tree_indices = {}
@@ -61,8 +60,9 @@ def build_gcmd_indices(gcmd_json):
 VALID_TOPICS, SUB_TREE_LOOKUP = build_gcmd_indices(gcmd_data)
 
 # ==========================================================
-# 2. LANGGRAPH ASYNC MULTI-TOPIC ARCHITECTURE
+# 2. VERBATIM WORKFLOW FROM YOUR NOTEBOOK
 # ==========================================================
+
 def merge_lists(left: list, right: list) -> list:
     """A reducer function that merges list contents across parallel branches."""
     return list(set((left or []) + (right or [])))
@@ -77,44 +77,39 @@ class MultiTopicState(TypedDict):
 class TopicsChoice(BaseModel):
     topics: List[str] = Field(description="List of matching topic areas from the allowed dataset.")
 
-async def route_multi_topic(state: MultiTopicState):
-    """Step 1: Identify ALL relevant high-level topics using async execution."""
-    llm = ChatOpenAI(model="gpt-4o-mini", temperature=0) 
+def route_multi_topic(state: MultiTopicState):
+    """Step 1: Identify ALL relevant high-level topics for the paper."""
+    llm = ChatOpenAI(model="gpt-4o", temperature=0)
     structured_llm = llm.with_structured_output(TopicsChoice)
     
     prompt = ChatPromptTemplate.from_messages([
-        ("system", f"You are an expert science cataloger. Identify ALL relevant major topic areas that apply to this paper. Choose ONLY from the allowed list: {', '.join(VALID_TOPICS)}"),
-        ("user", "Title: {title}\nAbstract: {abstract}\n\nSelect all relevant Topics.")
+        ("system", f"You are an expert science cataloger. Identify ALL relevant major topic areas that apply to this paper. Choose ONLY from: {', '.join(VALID_TOPICS)}"),
+        ("user", "Title: {title}\nAbstract: {abstract}\n\nSelect all relevant Topics as a structured list.")
     ])
     
-    result = await structured_llm.ainvoke(prompt.format(title=state["title"], abstract=state["abstract"]))
+    result = structured_llm.invoke(prompt.format(title=state["title"], abstract=state["abstract"]))
     valid_selected = [t.upper().strip() for t in result.topics if t.upper().strip() in VALID_TOPICS]
     
     if not valid_selected:
         valid_selected = ["FALLBACK"]
+        
     return {"chosen_topics": valid_selected}
 
 def classify_individual_topic(topic_name: str):
-    """Factory generating ASYNC node runners to avoid cloud thread-thrashing."""
+    """A dynamic factory function that returns a custom node runner for a specific topic."""
     
-    async def node_runner(state: MultiTopicState):
-        llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
+    def node_runner(state: MultiTopicState):
+        llm = ChatOpenAI(model="gpt-4o", temperature=0)
         target_sub_tree = SUB_TREE_LOOKUP.get(topic_name, "")
         
         prompt = ChatPromptTemplate.from_messages([
-            ("system", (
-                f"You are a specialist in {topic_name} data mapping. "
-                f"Extract exact keyword pathways present in the provided list. "
-                f"CRITICAL: Do NOT prepend '{topic_name} >' or 'EARTH SCIENCE >' to your answers. "
-                f"Your predictions must match the entries in the valid path list exactly, character-for-character."
-            )),
-            ("user", "Title: {title}\nAbstract: {abstract}\n\nValid Path List:\n{sub_tree}\n\nReturn exact matching entries as a clean, comma-separated list.")
+            ("system", f"You are a specialist in {topic_name} data mapping. Extract exact keyword pathways present in the provided list."),
+            ("user", "Title: {title}\nAbstract: {abstract}\n\nValid Paths:\n{sub_tree}\n\nReturn exact matching entries as a comma-separated list.")
         ])
         
-        response = await llm.ainvoke(prompt.format(title=state["title"], abstract=state["abstract"], sub_tree=target_sub_tree))
+        response = llm.invoke(prompt.format(title=state["title"], abstract=state["abstract"], sub_tree=target_sub_tree))
         raw_keywords = [k.strip() for k in response.content.split(",") if k.strip()]
         
-        # Reconstruct the exact validator set matching your notebook layout
         valid_set = set()
         for line in target_sub_tree.split("\n"):
             if line.strip():
@@ -122,10 +117,9 @@ def classify_individual_topic(topic_name: str):
                 path = re.sub(r"\s*\(Definition:.*?\)$", "", path).strip()
                 valid_set.add(path)
                 
-        # Filter generated terms deterministically 
-        # Prepend the topic string to the final outputs so they present nicely
-        valid_kws = [f"{topic_name} > {kw}" for kw in raw_keywords if kw in valid_set]
+        valid_kws = [kw for kw in raw_keywords if kw in valid_set]
         invalid_kws = [kw for kw in raw_keywords if kw not in valid_set]
+        
         return {"predicted_keywords": valid_kws, "invalid_keywords": invalid_kws}
         
     return node_runner
@@ -134,7 +128,7 @@ def parallel_router(state: MultiTopicState) -> List[str]:
     """Tells LangGraph to trigger multiple sub-nodes simultaneously based on state."""
     return [f"classify_{topic.lower()}" for topic in state["chosen_topics"]]
 
-# Assemble the workflow routing map
+# Assemble the Graph
 workflow = StateGraph(MultiTopicState)
 workflow.add_node("top_router", route_multi_topic)
 
@@ -158,30 +152,45 @@ workflow.add_edge("classify_fallback", END)
 app = workflow.compile()
 
 # ==========================================================
-# 3. GRADIO USER INTERFACE FOR DEMONSTRATION
+# 3. GRADIO USER INTERFACE (WITH PRESENTATION FORMATTING)
 # ==========================================================
-async def run_agent_classifier(title, abstract):
+def run_agent_classifier(title, abstract):
     if not title or not abstract:
         return "Please fill out both Title and Abstract fields.", "N/A", "N/A"
         
     inputs = {"title": title, "abstract": abstract}
-    output = await app.ainvoke(inputs)
     
-    topics_str = ", ".join(output.get("chosen_topics", []))
-    keywords_list = output.get("predicted_keywords", [])
-    invalid_list = output.get("invalid_keywords", [])
+    # Execute your exact notebook graph synchronously
+    output = app.invoke(inputs)
     
-    # Format valid output display strings
-    if not keywords_list:
+    chosen_topics = output.get("chosen_topics", [])
+    predicted_kws = output.get("predicted_keywords", [])
+    invalid_kws = output.get("invalid_keywords", [])
+    
+    topics_str = ", ".join(chosen_topics)
+    
+    # Requirement 1: Format output keywords to prepend their corresponding parent Topic
+    formatted_keywords = []
+    for kw in predicted_kws:
+        matched_topic = "UNKNOWN"
+        # Find which topic sub-tree this validated keyword originally belonged to
+        for topic in chosen_topics:
+            sub_tree_text = SUB_TREE_LOOKUP.get(topic, "")
+            if kw in sub_tree_text:
+                matched_topic = topic
+                break
+        formatted_keywords.append(f"• {matched_topic} > {kw}")
+        
+    if not formatted_keywords:
         keywords_str = "No explicit keywords mapped."
     else:
-        keywords_str = "\n".join([f"• {kw}" for kw in keywords_list])
+        keywords_str = "\n".join(formatted_keywords)
         
-    # Format invalid output display strings
-    if not invalid_list:
-        invalid_str = "None! The agent validation loop passed with 100% data integrity."
+    # Requirement 2: Format separate display for caught and filtered hallucinations
+    if not invalid_kws:
+        invalid_str = "None! The agent validation pass achieved 100% data integrity."
     else:
-        invalid_str = "\n".join([f"⚠ Caught & Removed: {ikw}" for ikw in invalid_list])
+        invalid_str = "\n".join([f"⚠ Caught & Removed: {ikw}" for ikw in invalid_kws])
         
     return topics_str, keywords_str, invalid_str
 
@@ -193,7 +202,7 @@ demo = gr.Interface(
     ],
     outputs=[
         gr.Textbox(label="Routed Multi-Topic Domains"),
-        gr.Textbox(label="Verified GCMD Keywords Extracted (with Topics)"),
+        gr.Textbox(label="Verified GCMD Keywords Extracted (Formatted with Topics)"),
         gr.Textbox(label="Hallucinated/Invalid Keywords Caught and Removed")
     ],
     title="GCMD Science Keyword Classifier Agent",