Update app.py

app.py

@@ -4,12 +4,12 @@ from array import array
 from functools import lru_cache
 import os, re, time, asyncio
 
-# 1. API Configuration - Switched to 1.5B for 4x faster cloud responses
+# 1. API Configuration
 HF_TOKEN = os.getenv("HF_TOKEN")
-MODEL_ID = "Qwen/Qwen2.5-1.5B-Instruct" # Faster, leaner, demo-optimized
+MODEL_ID = "Qwen/Qwen2.5-1.5B-Instruct" 
 client = AsyncInferenceClient(MODEL_ID, token=HF_TOKEN)
 
-# 2. T3 High-Speed Logic Kernel (Local CPU Execution)
+# 2. T3 High-Speed Logic Kernel
 class StateController:
     __slots__ = ("_state", "_rom60", "_symbols", "_rendered")
     def __init__(self):
@@ -18,7 +18,7 @@ class StateController:
         self._symbols    = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz01234567"
         self._rendered   = "".join(" [NODE_120] " if i == 120 else ("<" if i % 10 == 0 else ".") for i in range(121))
 
-    @lru_cache(maxsize=128) # Optimization from your refactor
+    @lru_cache(maxsize=128)
     def compute_distribution(self, total, nodes) -> str:
         if nodes <= 0: return "Error: Node count must be positive."
         base, rem = divmod(total, nodes)
@@ -36,20 +36,21 @@ def format_telemetry(seconds: float) -> str:
     if seconds < 0.001: return f"{seconds * 1_000_000:.2f} \u03BCs"
     return f"{seconds * 1_000:.2f} ms" if seconds < 1 else f"{seconds:.2f} s"
 
-# 3. Hardened Dual-Response Logic (Async Optimized)
+# 3. Core Response Logic (Async Corrected)
 async def generate_responses(user_message, p_hist, c_hist):
     msg = user_message.strip()
     if not msg: yield p_hist or [], c_hist or [], ""; return
 
     p_hist, c_hist = p_hist or [], c_hist or []
-    p_hist.append({"role": "user", "content": msg}); p_hist.append({"role": "assistant", "content": ""})
-    c_hist.append({"role": "user", "content": msg}); c_hist.append({"role": "assistant", "content": ""})
+    p_hist.append({"role": "user", "content": msg})
+    p_hist.append({"role": "assistant", "content": ""})
+    c_hist.append({"role": "user", "content": msg})
+    c_hist.append({"role": "assistant", "content": ""})
     yield p_hist, c_hist, ""
     
     start_time = time.perf_counter()
 
-    # --- HARDENED LOCAL INTERCEPTORS ---
-    # Using the refined regex from your refactor
+    # --- LOCAL INTERCEPTORS ---
     dist_match = re.search(r"(?P<units>\d{1,9})\s+units\s+across\s+(?P<nodes>\d{1,4})\s+nodes", msg, re.I)
     diag_match = any(kw in msg.lower() for kw in ["diagnostic", "grid"])
 
@@ -63,30 +64,46 @@ async def generate_responses(user_message, p_hist, c_hist):
         p_hist[-1]["content"] = f"{res}\n\n---\n*Telemetry: {format_telemetry(elapsed)} | Source: LOCAL T3 KERNEL*"
         yield p_hist, c_hist, ""
     else:
-        # ASYNC PRIMARY STREAM
+        # ASYNC PRIMARY STREAM (Corrected with 'await' for the stream object)
         try:
             res_text = ""
-            async for chunk in client.chat_completion(messages=[{"role":"system","content":"Logic Engine"}] + p_hist[:-1], max_tokens=512, stream=True, temperature=0.1):
+            # FIX: We must await the chat_completion call to get the AsyncIterable
+            stream = await client.chat_completion(
+                messages=[{"role":"system","content":"Logic Engine"}] + p_hist[:-1], 
+                max_tokens=512, 
+                stream=True, 
+                temperature=0.1
+            )
+            async for chunk in stream:
                 res_text += (chunk.choices[0].delta.content or "")
                 p_hist[-1]["content"] = res_text
                 yield p_hist, c_hist, ""
             p_hist[-1]["content"] += f"\n\n---\n*Telemetry: {format_telemetry(time.perf_counter()-start_time)} | Source: AUGMENTED CLOUD*"
+            yield p_hist, c_hist, ""
         except Exception as e:
             p_hist[-1]["content"] = f"Primary Error: {str(e)}"
-        yield p_hist, c_hist, ""
+            yield p_hist, c_hist, ""
 
-    # ASYNC VANILLA STREAM
+    # ASYNC VANILLA STREAM (Corrected with 'await' for the stream object)
     comp_start = time.perf_counter()
     c_hist[-1]["content"] = "*Routing...*"
     yield p_hist, c_hist, ""
     
     try:
         res_text = ""
-        async for chunk in client.chat_completion(messages=[{"role":"system","content":"Standard AI"}] + c_hist[:-1], max_tokens=512, stream=True, temperature=0.7):
+        # FIX: We must await the chat_completion call to get the AsyncIterable
+        stream = await client.chat_completion(
+            messages=[{"role":"system","content":"Standard AI"}] + c_hist[:-1], 
+            max_tokens=512, 
+            stream=True, 
+            temperature=0.7
+        )
+        async for chunk in stream:
             res_text += (chunk.choices[0].delta.content or "")
             c_hist[-1]["content"] = res_text
             yield p_hist, c_hist, ""
         c_hist[-1]["content"] += f"\n\n---\n*Telemetry: {format_telemetry(time.perf_counter()-comp_start)} | Source: VANILLA CLOUD*"
+        yield p_hist, c_hist, ""
     except Exception as e:
         c_hist[-1]["content"] = f"Competitor Error: {str(e)}"
     yield p_hist, c_hist, ""
@@ -99,12 +116,14 @@ with gr.Blocks() as demo:
     p_chat = gr.Chatbot(label="Augmented Logic Kernel (T3 Architecture)", height=350)
     with gr.Row():
         msg_in = gr.Textbox(label="Message", placeholder="Test P vs NP or Logistics Distribution...", scale=8)
-        btn = gr.Button("Execute", scale=1, variant="primary")
+        submit_btn = gr.Button("Execute", scale=1, variant="primary")
+    
     gr.Examples(examples=["Define P vs. NP. Then validate a 120-unit distribution across 3 nodes.", "Run grid diagnostic"], inputs=msg_in)
+    
     c_chat = gr.Chatbot(label="Vanilla Qwen 2.5 (Standard Infrastructure)", height=350)
 
     msg_in.submit(generate_responses, [msg_in, p_chat, c_chat], [p_chat, c_chat, msg_in])
-    btn.click(generate_responses, [msg_in, p_chat, c_chat], [p_chat, c_chat, msg_in])
+    submit_btn.click(generate_responses, [msg_in, p_chat, c_chat], [p_chat, c_chat, msg_in])
 
 if __name__ == "__main__":
     demo.queue().launch(theme=gr.themes.Soft(primary_hue="orange"), css=custom_css)