Update app.py

app.py

@@ -1,10 +1,8 @@
 import gradio as gr
 import os
-from huggingface_hub import InferenceClient
 
-# Use HF Inference API instead of loading model locally
-# This is FREE and doesn't require GPU/high memory
-client = InferenceClient()
+# Try multiple ways to get the token
+HF_TOKEN = os.getenv("HF_TOKEN") or os.getenv("HF_API_TOKEN") or os.getenv("HUGGING_FACE_HUB_TOKEN")
 
 def generate_output(project_name, user_input, file_path=None):
     """Generates specialized output using the LLM based on the project context."""
@@ -51,31 +49,58 @@ def generate_output(project_name, user_input, file_path=None):
         system_prompt = "You are a general-purpose AI assistant. Respond helpfully."
         max_tokens = 250
 
-    # Call the HF Inference API
+    # Call the HF Inference API using requests
     try:
-        messages = [
-            {"role": "system", "content": system_prompt},
-            {"role": "user", "content": user_input}
-        ]
+        import requests
         
-        response = client.chat_completion(
-            messages=messages,
-            model="mistralai/Mistral-7B-Instruct-v0.3",
-            max_tokens=max_tokens,
-            temperature=0.7
-        )
+        API_URL = "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.2"
+        
+        headers = {}
+        if HF_TOKEN:
+            headers["Authorization"] = f"Bearer {HF_TOKEN}"
+        
+        # Format the prompt for Mistral
+        prompt = f"<s>[INST] {system_prompt}\n\n{user_input} [/INST]"
+        
+        payload = {
+            "inputs": prompt,
+            "parameters": {
+                "max_new_tokens": max_tokens,
+                "temperature": 0.7,
+                "top_p": 0.9,
+                "return_full_text": False
+            }
+        }
         
-        generated_text = response.choices[0].message.content.strip()
-        return generated_text
+        response = requests.post(API_URL, headers=headers, json=payload, timeout=60)
+        
+        if response.status_code == 200:
+            result = response.json()
+            if isinstance(result, list) and len(result) > 0:
+                generated_text = result[0].get('generated_text', '').strip()
+                return generated_text
+            else:
+                return "Error: Unexpected response format from API"
+        elif response.status_code == 503:
+            return "⏳ Model is loading... Please wait 20-30 seconds and try again."
+        elif response.status_code == 401:
+            return "⚠️ Authentication Error: Your HF_TOKEN might be invalid. Please check it in Space Settings > Variables and secrets."
+        else:
+            return f"⚠️ API Error (Code {response.status_code}): {response.text}"
 
+    except ImportError:
+        return "ERROR: 'requests' library not found. Please add 'requests' to requirements.txt"
     except Exception as e:
-        return f"ERROR: Inference failed. Detail: {e}"
+        return f"ERROR: {str(e)}"
 
 
 # Build the Gradio Interface
 with gr.Blocks(title="VYZ Interstellar Multi-Tool AI", theme=gr.themes.Soft()) as demo:
     gr.Markdown("# 🚀 VYZ Interstellar AI Assistant")
     gr.Markdown("A specialized, multi-tool assistant for VYZVERSE, ReVYZGen, and VYZENVO projects. **(Powered by Mistral 7B via HF Inference API)**")
+    
+    if not HF_TOKEN:
+        gr.Markdown("### ⚠️ Warning: No HF_TOKEN found. Add it in Settings > Variables and secrets > New secret (Name: HF_TOKEN)")
 
     with gr.Tabs() as tabs:
         # TAB 1: VYZVERSE (Animation/Lore Generation)