Update app.py

app.py

@@ -1,31 +1,50 @@
 from fastapi import FastAPI, HTTPException
 from fastapi.middleware.cors import CORSMiddleware
 from pydantic import BaseModel, Field
-from llama_cpp import Llama
 from contextlib import asynccontextmanager
 import re
+import os
+
+try:
+    from llama_cpp import Llama
+except ImportError:
+    raise ImportError("Install llama-cpp-python: pip install llama-cpp-python")
 
 MODEL_REPO = "bartowski/Phi-3.5-mini-instruct-GGUF"
 MODEL_FILE = "Phi-3.5-mini-instruct-Q4_K_M.gguf"
 
-llm = None  # global reference
+llm = None
 
 @asynccontextmanager
 async def lifespan(app: FastAPI):
     global llm
-    print("🚀 Loading Phi-3.5 Mini (Fast Summarizer)...")
-    llm = Llama.from_pretrained(
-        repo_id=MODEL_REPO,
-        filename=MODEL_FILE,
-        n_threads=4,
-        n_ctx=2048,
-        n_batch=256,
-        n_gpu_layers=0,
-        verbose=False,
-    )
-    print("✅ Model loaded")
+    try:
+        print("🚀 Loading Phi-3.5 Mini (Fast Summarizer)...")
+        
+        # Try to load model with error handling
+        llm = Llama.from_pretrained(
+            repo_id=MODEL_REPO,
+            filename=MODEL_FILE,
+            n_threads=4,
+            n_ctx=2048,
+            n_batch=256,
+            n_gpu_layers=0,
+            verbose=False,
+        )
+        print("✅ Model loaded successfully")
+    except Exception as e:
+        print(f"❌ Error loading model: {e}")
+        print("Make sure you have:")
+        print("1. Installed llama-cpp-python")
+        print("2. Have internet connection for model download")
+        print("3. Have sufficient disk space (~2GB)")
+        llm = None
+    
     yield
+    
     print("🛑 Shutting down...")
+    if llm:
+        del llm
 
 app = FastAPI(
     title="AI Summarizer",
@@ -43,7 +62,7 @@ app.add_middleware(
 
 class SummarizeRequest(BaseModel):
     text: str = Field(..., min_length=1, max_length=2000)
-    length: str = "short"
+    length: str = Field(default="short", pattern="^(short|medium|long)$")
 
 LENGTH_INSTRUCTIONS = {
     "short": "Summarize in 2–3 concise sentences.",
@@ -52,6 +71,7 @@ LENGTH_INSTRUCTIONS = {
 }
 
 def clean_output(text: str) -> str:
+    """Clean model output from special tokens"""
     text = re.sub(r"<\|.*?\|>", "", text)
     text = re.sub(r"\s+", " ", text)
     return text.strip()
@@ -59,36 +79,81 @@ def clean_output(text: str) -> str:
 @app.post("/api/summarize")
 async def summarize(req: SummarizeRequest):
     if llm is None:
-        raise HTTPException(status_code=503, detail="Model not loaded")
-
-    text = req.text.strip()
-    length_instruction = LENGTH_INSTRUCTIONS.get(req.length, LENGTH_INSTRUCTIONS["short"])
-
-    prompt = f"""
-<|user|>
+        raise HTTPException(
+            status_code=503, 
+            detail="Model not loaded. Check server logs for errors."
+        )
+    
+    try:
+        text = req.text.strip()
+        length_instruction = LENGTH_INSTRUCTIONS.get(
+            req.length, 
+            LENGTH_INSTRUCTIONS["short"]
+        )
+        
+        prompt = f"""<|user|>
 You are an expert text summarizer.
 {length_instruction}
 
 Text:
 {text}
 <|end|>
-<|assistant|>
-"""
-
-    output = llm(
-        prompt,
-        max_tokens=140 if req.length == "short" else 220,
-        temperature=0.3,
-        top_p=0.9,
-        top_k=40,
-        repeat_penalty=1.05,
-        stop=["<|end|>", "<|user|>"],
-    )
-
-    summary = clean_output(output["choices"][0]["text"])
-    return {"summary": summary, "success": True}
+<|assistant|>"""
+        
+        max_tokens_map = {
+            "short": 140,
+            "medium": 220,
+            "long": 300
+        }
+        
+        output = llm(
+            prompt,
+            max_tokens=max_tokens_map.get(req.length, 140),
+            temperature=0.3,
+            top_p=0.9,
+            top_k=40,
+            repeat_penalty=1.05,
+            stop=["<|end|>", "<|user|>"],
+            echo=False
+        )
+        
+        summary = clean_output(output["choices"][0]["text"])
+        
+        if not summary:
+            raise HTTPException(
+                status_code=500,
+                detail="Model produced empty output"
+            )
+        
+        return {
+            "summary": summary,
+            "success": True,
+            "length": req.length
+        }
+    
+    except Exception as e:
+        raise HTTPException(
+            status_code=500,
+            detail=f"Summarization error: {str(e)}"
+        )
 
 @app.get("/")
 def health():
-    return {"status": "ok", "model": MODEL_FILE}
-
+    return {
+        "status": "ok" if llm else "model_not_loaded",
+        "model": MODEL_FILE,
+        "ready": llm is not None
+    }
+
+@app.get("/health")
+def detailed_health():
+    return {
+        "status": "healthy" if llm else "unhealthy",
+        "model_loaded": llm is not None,
+        "model_name": MODEL_FILE,
+        "repo": MODEL_REPO
+    }
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=8000)