Upload folder using huggingface_hub

app.py

@@ -1,4 +1,6 @@
 import os
+import random
+import time
 import gradio as gr
 from config import SanatanConfig
 from drive_downloader import ZipDownloader
@@ -17,10 +19,10 @@ graph = generate_graph()
 def init():
     load_dotenv(override=True)
 
-    # if(os.path.isdir("./chromadb-store")):
-    #     logger.warning("database exists locally. not downloading.")
-    #     return
-    
+    if(os.path.isdir("./chromadb-store")):
+        logger.warning("database exists locally. not downloading.")
+        return
+
     downloader = ZipDownloader(
         service_account_json=os.getenv("GOOGLE_SERVICE_ACCOUNT_JSON")
     )
@@ -58,15 +60,61 @@ def chat(message, history, thread_id):
     )
     return response["messages"][-1].content
 
+
+import uuid
+import asyncio
+from html import escape
+
 async def chat_streaming(message, history, thread_id):
     state = {"messages": (history or []) + [{"role": "user", "content": message}]}
     config = {"configurable": {"thread_id": thread_id}}
-    async for step in graph.astream(state,config=config):
-        # LangGraph yields steps like {"respond": ChatState, "post_process": ChatState}
-        for node_output in step.values():
-            yield node_output["messages"][-1].content
 
-    print("received chat message for thread:", thread_id)
+    buffered_final = None
+    step_counter = 0
+
+    thinking_verbs = ["thinking","processing","crunching data","please wait","just a few more seconds","closing in","racing to the finish line"]
+
+    async for step in graph.astream(state, config=config):
+        for node_name, node_output in step.items():
+            step_counter += 1
+            messages = node_output.get("messages", [])
+            if not messages:
+                continue
+
+            last_message = messages[-1]
+            content = getattr(last_message, "content", "") or ""
+
+            full = escape(content)
+            truncated = (full[:300] + "…") if len(full) > 300 else full
+
+            buffered_final = full  # Save latest response in case it's the final one
+
+            # Unique ID for each step to help frontend match animation + content
+            thinking_id = str(uuid.uuid4())
+
+            # Yield animated thinking bubble BEFORE actual message
+            yield f"<div id='{thinking_id}'><em>💭 {random.choice(thinking_verbs)}...</em></div>"
+            await asyncio.sleep(0.5)  # let the thinking bubble show up slightly before
+
+            # Yield actual message in transparent font with node name
+            # Instead of two yields: one for thinking, one for message
+            # Combine both in one yield so frontend shows them as one stream block
+            html = (
+                f"<div><em>💭 {random.choice(thinking_verbs)} ...</em></div>"
+                f"<div style='opacity: 0.5' title='{full}'>"
+                f"<strong>[{node_name}]</strong> {truncated or '...'}"
+                f"</div>"
+            )
+            yield html
+
+    # ✅ Final answer: normal font, no node label
+    if buffered_final:
+        temp = ""
+        for char in buffered_final:
+            temp = temp + char
+            await asyncio.sleep(0.001)  # ✅ non-blocking!
+            yield temp
+
 
 thread_id = gr.State(init_session)
 supported_scriptures = "\n - ".join(
@@ -108,7 +156,7 @@ chatbot = gr.Chatbot(
     render_markdown=True,
 )
 
-debug_checkbox = gr.Checkbox(label="Debug (Streaming)", value=False)
+debug_checkbox = gr.Checkbox(label="Debug (Streaming)", value=True)
 
 chatInterface = gr.ChatInterface(
     title="Sanatan-AI",