Update app.py

app.py

@@ -1,55 +1,104 @@
 import os
+import sys
 import struct
+import traceback
 import gradio as gr
-from llama_cpp import Llama
-from cryptography.hazmat.primitives.ciphers.aead import AESGCM
 from huggingface_hub import hf_hub_download, login
+from cryptography.hazmat.primitives.ciphers.aead import AESGCM
 from fastapi import FastAPI, Request
 
+# --- GLOBAL ERROR TRACKER ---
+DIAGNOSTIC_LOG = []
+def log_status(msg):
+    print(msg)
+    DIAGNOSTIC_LOG.append(msg)
+
+# --- 1. CRITICAL IMPORT WRAPPER ---
+Llama = None
+try:
+    log_status("📡 [IMPORT] Attempting to load llama_cpp...")
+    from llama_cpp import Llama
+    log_status("✅ [IMPORT] llama_cpp library linked successfully.")
+except Exception as e:
+    log_status(f"❌ [IMPORT ERROR] Library mismatch detected: {e}")
+    log_status(f"DEBUG: System Path: {sys.path}")
+    log_status(traceback.format_exc())
+
 # --- CONFIG ---
-HF_TOKEN = os.environ.get("HF_TOKEN")
-SECRET_KEY_HEX = os.environ.get("DECRYPTION_KEY")
 SOURCE_REPO = "metanthropic/metanthropic-phi3-encrypted"
 SOURCE_FILE = "metanthropic-phi3-v1.mguf"
-TEMP_DECRYPTED = "/tmp/model.gguf"
-
-print("🔄 [BOOT] Metanthropic Node Initiating...")
-
-def boot_engine():
-    if os.path.exists(TEMP_DECRYPTED): return
-    if HF_TOKEN: login(token=HF_TOKEN)
-    
-    print(f"⬇️ Fetching {SOURCE_FILE}...")
-    path = hf_hub_download(repo_id=SOURCE_REPO, filename=SOURCE_FILE, local_dir=".")
-    
-    print("🔓 Decrypting...")
-    key = bytes.fromhex(SECRET_KEY_HEX)
-    aes = AESGCM(key)
-    with open(path, "rb") as f_in, open(TEMP_DECRYPTED, "wb") as f_out:
-        nonce = f_in.read(12)
-        h_len = struct.unpack("<I", f_in.read(4))[0]
-        f_out.write(aes.decrypt(nonce, f_in.read(h_len), None))
-        while chunk := f_in.read(64*1024*1024): f_out.write(chunk)
-    os.remove(path)
-    print("✅ Engine Ready.")
-
-# --- LOAD ---
+TEMP_DECRYPTED = "/tmp/model_stable.gguf"
+HF_TOKEN = os.environ.get("HF_TOKEN")
+SECRET_KEY_HEX = os.environ.get("DECRYPTION_KEY")
+
+def robust_boot():
+    try:
+        if os.path.exists(TEMP_DECRYPTED):
+            log_status("⚡ [CACHE] Decrypted model exists.")
+            return True
+
+        # Check Secrets
+        if not HF_TOKEN or not SECRET_KEY_HEX:
+            log_status("❌ [AUTH ERROR] Missing HF_TOKEN or DECRYPTION_KEY in Secrets.")
+            return False
+
+        # Login
+        log_status("🔐 [AUTH] Authenticating...")
+        login(token=HF_TOKEN)
+
+        # Download
+        log_status(f"⬇️ [NETWORK] Fetching {SOURCE_FILE}...")
+        path = hf_hub_download(repo_id=SOURCE_REPO, filename=SOURCE_FILE, local_dir=".")
+        
+        # Decrypt
+        log_status("🔓 [SECURITY] Decrypting model...")
+        key = bytes.fromhex(SECRET_KEY_HEX)
+        aes = AESGCM(key)
+        with open(path, "rb") as f_in, open(TEMP_DECRYPTED, "wb") as f_out:
+            nonce = f_in.read(12)
+            h_len = struct.unpack("<I", f_in.read(4))[0]
+            f_out.write(aes.decrypt(nonce, f_in.read(h_len), None))
+            while chunk := f_in.read(64*1024*1024):
+                f_out.write(chunk)
+        
+        os.remove(path)
+        log_status("✅ [SUCCESS] Model ready for engine.")
+        return True
+
+    except Exception as e:
+        log_status(f"❌ [BOOT ERROR] {e}")
+        log_status(traceback.format_exc())
+        return False
+
+# --- ENGINE INITIALIZATION ---
 llm = None
-boot_engine()
-llm = Llama(model_path=TEMP_DECRYPTED, n_ctx=2048, n_threads=2)
+if Llama and robust_boot():
+    try:
+        log_status("🧠 [ENGINE] Initializing Llama...")
+        llm = Llama(model_path=TEMP_DECRYPTED, n_ctx=2048, n_threads=2)
+        log_status("🚀 [SYSTEM] Node Online.")
+    except Exception as e:
+        log_status(f"❌ [ENGINE ERROR] Failed to load model file: {e}")
+        log_status(traceback.format_exc())
 
-# --- API ---
+# --- API & INTERFACE ---
 app = FastAPI()
 
 @app.post("/run_inference")
 async def run_inference(request: Request):
+    if not llm:
+        return {"error": "Model offline", "logs": DIAGNOSTIC_LOG}
     data = await request.json()
     prompt = data.get("prompt", "")
-    out = llm(f"<|user|>\n{prompt}<|end|>\n<|assistant|>", max_tokens=512, stop=["<|end|>"])
-    return {"response": out['choices'][0]['text'].strip()}
+    output = llm(f"<|user|>\n{prompt}<|end|>\n<|assistant|>", max_tokens=512, stop=["<|end|>"])
+    return {"response": output['choices'][0]['text'].strip()}
+
+def ui_chat(msg, hist):
+    if not llm:
+        return f"🚨 SYSTEM ERROR\n\nLatest Logs:\n" + "\n".join(DIAGNOSTIC_LOG[-5:])
+    return llm(f"<|user|>\n{msg}<|end|>\n<|assistant|>", max_tokens=512, stop=["<|end|>"])['choices'][0]['text'].strip()
 
-# UI
-demo = gr.ChatInterface(lambda msg, hist: llm(f"<|user|>\n{msg}<|end|>\n<|assistant|>", max_tokens=512, stop=["<|end|>"])['choices'][0]['text'].strip())
+demo = gr.ChatInterface(ui_chat, title="Metanthropic Sovereign Node (Diagnostic Mode)")
 app = gr.mount_gradio_app(app, demo, path="/")
 
 if __name__ == "__main__":