Create app.py

app.py

@@ -0,0 +1,63 @@
+import subprocess
+import time
+import os
+import requests
+from fastapi import FastAPI, Request
+from fastapi.responses import StreamingResponse
+from duckduckgo_search import DDGS
+
+app = FastAPI()
+
+# 1. Boot the AI Engine in the background
+subprocess.Popen([
+    "/app/llama-server", 
+    "--model", "/models/model.gguf", 
+    "--host", "127.0.0.1", 
+    "--port", "8080", 
+    "--nopreload"
+])
+
+def fetch_live_data(query):
+    """Fetches real-time web info for news, scores, and dates"""
+    try:
+        with DDGS() as ddgs:
+            results = [f"{r['title']}: {r['body']}" for r in ddgs.text(query, max_results=3)]
+            return "\n".join(results)
+    except Exception:
+        return "No live data found, using internal 2026 knowledge."
+
+@app.post("/v1/chat/completions")
+async def chat_proxy(request: Request):
+    body = await request.json()
+    user_query = body["messages"][-1]["content"]
+
+    # Detect if user needs live info
+    live_triggers = ["news", "score", "price", "weather", "today", "who won"]
+    context = ""
+    if any(word in user_query.lower() for word in live_triggers):
+        context = f"\n[LIVE 2026 DATA]: {fetch_live_data(user_query)}"
+
+    # Inject the "Positive Power" personality and live context
+    system_instr = (
+        "You are a high-speed, witty AI with Positive Power. "
+        "Use the provided live data to give accurate news/scores. "
+        "Be extremely brief (max 3 sentences) and highly motivational! "
+        "Current Date: Feb 2026."
+    )
+    
+    body["messages"].insert(0, {"role": "system", "content": system_instr})
+    body["messages"][-1]["content"] += context
+
+    # Forward to local llama-server
+    def stream_response():
+        with requests.post("http://127.0.0.1:8080/v1/chat/completions", json=body, stream=True) as r:
+            for chunk in r.iter_content(chunk_size=128):
+                yield chunk
+
+    return StreamingResponse(stream_response(), media_type="application/json")
+
+if __name__ == "__main__":
+    # Wait for the model to load into memory
+    time.sleep(10)
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=7860)