Disabled phi3 model

app.py

@@ -18,6 +18,14 @@ import networkx as nx
 
 from llm_graph import LLMGraph, MODEL_LIST
 
+# Helper function to check if a model is the Phi-3 model (disabled or not)
+def is_phi3_model(model_name):
+    return "Phi-3-mini-128k-instruct-graph" in model_name
+
+# Helper function to check if a model is the OpenAI model
+def is_openai_model(model_name):
+    return "OpenAI" in model_name or "GPT" in model_name
+
 from pyvis.network import Network
 from spacy import displacy
 from spacy.tokens import Span
@@ -107,7 +115,7 @@ def extract_kg(text="", model_name=MODEL_LIST[0], model=None, graph_file=""):
     
     try:
         start_time = time.time()
-        if model_name == MODEL_LIST[1] and os.path.exists(graph_file):
+        if is_openai_model(model_name) and os.path.exists(graph_file):
             # Load the graph directly from cache
             logging.info(f"Loading graph from cache: {graph_file}")
             G = nx.read_graphml(graph_file)
@@ -116,7 +124,7 @@ def extract_kg(text="", model_name=MODEL_LIST[0], model=None, graph_file=""):
             result = nx.node_link_data(G, edges="edges")
         else:
             result = model.extract(text, model_name, graph_file)
-        
+
         end_time = time.time()
         duration = end_time - start_time
         logging.info(f"Response time: {duration:.4f} seconds")
@@ -212,7 +220,7 @@ def create_graph(json_data, model_name=MODEL_LIST[0], graph_file=""):
     Create interactive knowledge graph using Pyvis.
     """
 
-    if model_name == MODEL_LIST[0]:
+    if is_phi3_model(model_name):
         G = nx.Graph()
 
         # Add nodes with tooltips and error handling for missing keys
@@ -222,7 +230,7 @@ def create_graph(json_data, model_name=MODEL_LIST[0], graph_file=""):
 
             # Get detailed type with fallback
             detailed_type = node.get("detailed_type", type)
-            
+
             # Use node ID and type info for the tooltip
             G.add_node(node['id'], title=f"{type}: {detailed_type}")
 
@@ -247,7 +255,7 @@ def create_graph(json_data, model_name=MODEL_LIST[0], graph_file=""):
     
     # Configure network display
     network.from_nx(G)
-    if model_name == MODEL_LIST[0]:
+    if is_phi3_model(model_name):
         network.barnes_hut(
             gravity=-3000,
             central_gravity=0.3,
@@ -339,12 +347,12 @@ def process_and_visualize(text, model_name, progress=gr.Progress()):
     is_first_example = text == EXAMPLES[0][0]
 
     # Try to load from cache if it's the first example
-    if is_first_example and model_name == MODEL_LIST[0] and os.path.exists(EXAMPLE_CACHE_FILE):
+    if is_first_example and is_phi3_model(model_name) and os.path.exists(EXAMPLE_CACHE_FILE):
         try:
             progress(0.3, desc="Loading from cache...")
             with open(EXAMPLE_CACHE_FILE, 'rb') as f:
                 cached_data = pickle.load(f)
-                
+
             progress(1.0, desc="Loaded from cache!")
             return cached_data["graph_html"], cached_data["entities_viz"], cached_data["json_data"], cached_data["stats"]
         except Exception as e:
@@ -379,20 +387,20 @@ def process_and_visualize(text, model_name, progress=gr.Progress()):
     json_data = extract_kg(text, model_name, model, graph_file)
     
     progress(0.5, desc="Creating entity visualization...")
-    if model_name == MODEL_LIST[0]:
+    if is_phi3_model(model_name):
         entities_viz = create_custom_entity_viz(json_data, text, type_col="type")
     else:
         entities_viz = create_custom_entity_viz(json_data, text, type_col="entity_type")
-    
+
     progress(0.8, desc="Building knowledge graph...")
     graph_html = create_graph(json_data, model_name, graph_file)
-    
+
     node_count = len(json_data["nodes"])
     edge_count = len(json_data["edges"])
     stats = f"📊 Extracted {node_count} entities and {edge_count} relationships"
-    
+
     # Save to cache if it's the first example
-    if is_first_example and model_name == MODEL_LIST[0]:
+    if is_first_example and is_phi3_model(model_name):
         try:
             cached_data = {
                 "graph_html": graph_html,
@@ -421,6 +429,7 @@ EXAMPLES = [
 def generate_first_example():
     """
     Generate cache for the first example if it doesn't exist when the app starts.
+    Note: Since the first model (Phi-3) is disabled, we use the second model (OpenAI) for caching.
     """
 
     if not os.path.exists(EXAMPLE_CACHE_FILE):
@@ -428,21 +437,22 @@ def generate_first_example():
 
         try:
             text = EXAMPLES[0][0]
-            model_name = MODEL_LIST[0] if MODEL_LIST else None
+            # Use the second model (OpenAI) since the first is disabled
+            model_name = MODEL_LIST[1] if len(MODEL_LIST) > 1 else None
 
             # Initialize the LLMGraph model
             model = LLMGraph(working_dir=WORKING_DIR)
-            asyncio.run(model.initialize_rag()) 
+            asyncio.run(model.initialize_rag())
 
             # Extract data
             json_data = extract_kg(text, model_name, model)
-            entities_viz = create_custom_entity_viz(json_data, text)
-            graph_html = create_graph(json_data)
-            
+            entities_viz = create_custom_entity_viz(json_data, text, type_col="entity_type")
+            graph_html = create_graph(json_data, model_name)
+
             node_count = len(json_data["nodes"])
             edge_count = len(json_data["edges"])
             stats = f"📊 Extracted {node_count} entities and {edge_count} relationships"
-            
+
             # Save to cache
             cached_data = {
                 "graph_html": graph_html,
@@ -467,7 +477,7 @@ def generate_first_example():
                 return pickle.load(f)
         except Exception as e:
             logging.error(f"Error loading existing cache: {str(e)}")
-    
+
     return None
 
 def create_ui():

llm_graph.py

@@ -28,7 +28,7 @@ AZURE_EMBEDDING_DEPLOYMENT = os.environ["AZURE_EMBEDDING_DEPLOYMENT"]
 AZURE_EMBEDDING_API_VERSION = os.environ["AZURE_EMBEDDING_API_VERSION"]
 
 MODEL_LIST = [
-  "EmergentMethods/Phi-3-mini-128k-instruct-graph",
+  "EmergentMethods/Phi-3-mini-128k-instruct-graph (Disabled)",
   "OpenAI/GPT-4.1-mini",
 ]
 
@@ -135,7 +135,11 @@ class LLMGraph:
         Extract knowledge graph in structured format from text.
         """
 
-        if model_name == MODEL_LIST[0]:
+        # Check if the model is the disabled Phi-3 model
+        if "Phi-3-mini-128k-instruct-graph" in model_name and "Disabled" in model_name:
+            raise ValueError("This model is currently disabled. Please select another model.")
+
+        if model_name == MODEL_LIST[0] or "Phi-3-mini-128k-instruct-graph" in model_name:
             # Use Hugging Face Inference API with Phi-3-mini-128k-instruct-graph
             messages = self._get_messages(text)