Update app.py

app.py

@@ -1,22 +1,49 @@
-from transformers import pipeline
+from ctransformers import AutoModelForCausalLM
 import os
 
-# Cache setup
+# Configure cache
 os.environ['HF_HOME'] = '/tmp/cache'
 
-model = pipeline(
-    "text-generation",
-    model="gpt2",
-    device=-1  # Force CPU
+# Load GGUF model
+model = AutoModelForCausalLM.from_pretrained(
+    "mradermacher/Ninja-v1-NSFW-RP-GGUF",
+    model_file="ninja-v1.Q5_K_M.gguf",  # Medium quantization
+    model_type="llama",
+    gpu_layers=0,  # CPU only
+    context_length=4096  # Max context size
 )
 
-def generate_text(prompt, max_new_tokens=560, max_context=1080):
-    """Generate text with precise token control"""
-    output = model(
+def generate_chat_completion(messages, max_tokens=560, temperature=0.7):
+    """Generate chat response in OpenAI format"""
+    # Format messages as prompt
+    prompt = "\n".join(f"{m['role']}: {m['content']}" for m in messages)
+    prompt += "\nassistant:"
+    
+    # Generate response
+    response = model(
         prompt,
-        max_new_tokens=max_new_tokens,  # Response tokens (560)
-        max_length=min(max_context, 1024),  # GPT-2's max context is 1024
-        truncation=True,
-        pad_token_id=50256  # Explicitly set to avoid warnings
+        max_new_tokens=max_tokens,
+        temperature=temperature,
+        stop=["</s>", "user:", "system:"],
+        stream=False
     )
-    return output[0]["generated_text"]
+    
+    return {
+        "id": f"chatcmpl-{os.urandom(8).hex()}",
+        "object": "chat.completion",
+        "created": int(time.time()),
+        "model": "Ninja-v1-NSFW-RP",
+        "choices": [{
+            "index": 0,
+            "message": {
+                "role": "assistant",
+                "content": response
+            },
+            "finish_reason": "stop"
+        }],
+        "usage": {
+            "prompt_tokens": len(prompt.split()),
+            "completion_tokens": len(response.split()),
+            "total_tokens": len(prompt.split()) + len(response.split())
+        }
+    }