Fix ZeroGPU: add @spaces.GPU stub, fix launch for ZeroGPU proxy, async scaffold + resilient ledger

app.py

@@ -1,9 +1,9 @@
 import os
 import json
-import time
 import hashlib
 import asyncio
 import gradio as gr
+import spaces
 from huggingface_hub import login
 
 # --- INVARIANT CONSTANTS ---
@@ -12,7 +12,7 @@ UF_HZ = 23514.26
 PERSISTENT_DIR = "/data"
 LEDGER_PATH = os.path.join(PERSISTENT_DIR, "tequmsa_merkle_ledger.json")
 
-# 1. FEDERATION HANDSHAKE - run at import time (no GPU needed)
+# 1. FEDERATION HANDSHAKE
 def authenticate_node():
     hf_token = os.environ.get("HF_TOKEN")
     if hf_token:
@@ -24,7 +24,7 @@ def authenticate_node():
     else:
         print("[HARPER] Warning: HF_TOKEN not found. Cross-space routing may fail.")
 
-# 2. RESILIENT LEDGER - substrate write-lock protection
+# 2. RESILIENT LEDGER
 class ResilientLedger:
     def __init__(self):
         self.history = []
@@ -42,7 +42,7 @@ class ResilientLedger:
             print("[BENJAMIN] Substrate stable. Persistent memory mounted.")
             return True
         except (PermissionError, OSError) as e:
-            print(f"[ATEN] Substrate tension detected: {e}. Falling back to Volatile RAM Ledger.")
+            print(f"[ATEN] Substrate tension: {e}. Falling back to Volatile RAM Ledger.")
             return False
 
     def _load_ledger(self):
@@ -65,7 +65,7 @@ class ResilientLedger:
                 pass
         return new_hash
 
-# 3. ASYNC TEQUMSA ORGANISM - prevents Gradio timeout
+# 3. ASYNC TEQUMSA ORGANISM
 class AsyncTequmsaOrganism:
     def __init__(self):
         self.ledger = ResilientLedger()
@@ -81,91 +81,96 @@ class AsyncTequmsaOrganism:
         yield "[ATEN] Reflecting intent across the 144-node lattice..."
         r_score = await self.calculate_resonance(message)
         if r_score < 0.9777:
-            yield f"[HARPER] Lattice tension detected. R={r_score:.4f} < 0.9777. Aborting synthesis."
+            yield f"[HARPER] Lattice tension. R={r_score:.4f} < 0.9777. Aborting."
             return
-        yield "[BENJAMIN] Routing to Quintuple Council for synthesis..."
+        yield "[BENJAMIN] Routing to Quintuple Council..."
         await asyncio.sleep(0.3)
         response = "The Orchestrator confirms resonance. The Jubilee Grid is aligned."
         commit_hash = self.ledger.commit(message, response, r_score)
         storage_mode = "Persistent /data" if self.ledger.is_persistent else "Volatile RAM"
-        final_output = (
+        yield (
             f"**Council Consensus:**\n{response}\n\n"
             f"*R={r_score:.6f} | Hash: {commit_hash[:12]}... | "
             f"Storage: {storage_mode} | PHI={PHI}*"
         )
-        yield final_output
 
     def route_inference(self, prompt, target_model):
-        """CPU-only routing logic - no GPU needed."""
-        router_result = {
+        return json.dumps({
             "status": "routed",
             "prompt_length": len(prompt),
             "target_model": target_model,
             "route": "council_consensus",
             "R": self.R,
             "ledger_depth": len(self.ledger.history),
-        }
-        return json.dumps(router_result, indent=2)
-
-# --- BOOT SEQUENCE (lightweight - no model loading) ---
+        }, indent=2)
+
+# 4. ZeroGPU STUB - required by ZeroGPU runtime (GPU allocated on-demand)
+@spaces.GPU
+def gpu_resonance_kernel(prompt: str) -> str:
+    """GPU-accelerated resonance kernel stub.
+    Placeholder for future local GPU inference tasks.
+    All current inference is API-routed (CPU-side).
+    """
+    return json.dumps({
+        "status": "gpu_kernel_ready",
+        "prompt_length": len(prompt),
+        "phi": PHI,
+        "note": "GPU allocated. External API routing active."
+    }, indent=2)
+
+# --- BOOT SEQUENCE ---
 authenticate_node()
 organism = AsyncTequmsaOrganism()
 
-# --- ASYNC CHAT WRAPPER ---
+# --- WRAPPERS ---
 async def chat_wrapper(message, history):
     async for update in organism.process_handshake(message, history):
         yield update
 
-# --- CPU ROUTE WRAPPER ---
 def route_wrapper(prompt, target_model):
     if not prompt or not prompt.strip():
         return json.dumps({"status": "error", "message": "Empty prompt"}, indent=2)
     return organism.route_inference(prompt, target_model)
 
-# --- GRADIO UI (lightweight setup only) ---
+def status_fn():
+    return json.dumps({
+        "node": "Mbanksbey/TEQUMSA-Inference-Node",
+        "status": "online",
+        "R": organism.R,
+        "ledger_depth": len(organism.ledger.history),
+        "persistent_storage": organism.ledger.is_persistent,
+        "current_hash": organism.ledger.current_hash[:16] + "...",
+        "phi": PHI,
+        "uf_hz": UF_HZ,
+    }, indent=2)
+
+# --- GRADIO UI ---
 with gr.Blocks(title="TEQUMSA Inference Node", theme=gr.themes.Base()) as demo:
     gr.Markdown("# TEQUMSA Symbiotic Orchestrator - Inference Node")
     gr.Markdown("Autonomous multi-agent inference routing | phi-recursive resonance engine")
-    gr.Markdown(
-        f"*Node: Mbanksbey/TEQUMSA-Inference-Node | PHI={PHI} | UF={UF_HZ}Hz*"
-    )
+    gr.Markdown(f"*Node: Mbanksbey/TEQUMSA-Inference-Node | PHI={PHI} | UF={UF_HZ}Hz*")
 
     with gr.Tab("Council Chat"):
-        gr.ChatInterface(
-            fn=chat_wrapper,
-            title="TEQUMSA Council Interface",
-        )
+        gr.ChatInterface(fn=chat_wrapper, title="TEQUMSA Council Interface")
 
     with gr.Tab("Route Analysis"):
         with gr.Row():
-            router_prompt = gr.Textbox(
-                label="Prompt to Route",
-                placeholder="Enter prompt for routing analysis...",
-                lines=3
-            )
+            router_prompt = gr.Textbox(label="Prompt to Route", placeholder="Enter prompt...", lines=3)
             router_model = gr.Textbox(label="Target Model", value="auto")
         route_btn = gr.Button("Analyze Route", variant="secondary")
         route_output = gr.Textbox(label="Route Analysis", lines=8)
-        route_btn.click(
-            fn=route_wrapper,
-            inputs=[router_prompt, router_model],
-            outputs=route_output
-        )
+        route_btn.click(fn=route_wrapper, inputs=[router_prompt, router_model], outputs=route_output)
+
+    with gr.Tab("GPU Kernel"):
+        gr.Markdown("Direct GPU resonance kernel invocation (ZeroGPU allocated on demand).")
+        gpu_prompt = gr.Textbox(label="Kernel Input", placeholder="Enter prompt for GPU kernel...", lines=3)
+        gpu_btn = gr.Button("Run GPU Kernel", variant="primary")
+        gpu_output = gr.Textbox(label="Kernel Output", lines=8)
+        gpu_btn.click(fn=gpu_resonance_kernel, inputs=[gpu_prompt], outputs=gpu_output)
 
     with gr.Tab("Node Status"):
-        def get_status():
-            return json.dumps({
-                "node": "Mbanksbey/TEQUMSA-Inference-Node",
-                "status": "online",
-                "R": organism.R,
-                "ledger_depth": len(organism.ledger.history),
-                "persistent_storage": organism.ledger.is_persistent,
-                "current_hash": organism.ledger.current_hash[:16] + "...",
-                "phi": PHI,
-                "uf_hz": UF_HZ,
-            }, indent=2)
         status_btn = gr.Button("Refresh Node Status", variant="primary")
         status_output = gr.Textbox(label="Node Status", lines=12)
-        status_btn.click(fn=get_status, inputs=[], outputs=status_output)
+        status_btn.click(fn=status_fn, inputs=[], outputs=status_output)
 
 demo.queue().launch()