Update app.py

app.py

@@ -17,6 +17,61 @@ import re
 import chromadb # ADDED for client check
 from typing import List, Dict, Any, Optional
 
+# --- Constants ---
+MODEL_REPO = "TheBloke/zephyr-7B-beta-GGUF"
+MODEL_FILE = "zephyr-7b-beta.Q4_K_M.gguf"
+# Define a path within the persistent storage for the model
+# Using os.environ.get('HF_HOME', '/data') ensures it uses HF_HOME if set,
+# otherwise defaults to /data. You might want a specific models subdir.
+# Let's create a dedicated model path within /data:
+MODEL_DIR = "/data/models" # Store models in /data/models
+LOCAL_MODEL_PATH = os.path.join(MODEL_DIR, MODEL_FILE)
+
+# --- Function to Download Model (Runtime Check) ---
+def download_model_if_needed():
+    """Checks if model exists in persistent storage, downloads if not."""
+    print(f"Checking for model file at: {LOCAL_MODEL_PATH}")
+    if not os.path.exists(LOCAL_MODEL_PATH):
+        print(f"Model not found locally. Downloading from {MODEL_REPO}...")
+        try:
+            # Create the directory if it doesn't exist
+            os.makedirs(MODEL_DIR, exist_ok=True)
+            # Use hf_hub_download for robust downloading & caching (respects HF_HOME)
+            # We specify local_dir to force it into our /data structure,
+            # and local_dir_use_symlinks=False to avoid symlinks if that causes issues.
+            # If you set HF_HOME=/data in Dockerfile, it *should* cache there by default,
+            # but explicitly downloading to a specific path within /data is safer.
+            hf_hub_download(
+                repo_id=MODEL_REPO,
+                filename=MODEL_FILE,
+                local_dir=MODEL_DIR, # Download directly into this folder
+                local_dir_use_symlinks=False, # Avoid symlinks, copy directly
+                # cache_dir=os.environ.get('HF_HOME') # Optional: force cache dir if needed
+            )
+            # Verify download
+            if os.path.exists(LOCAL_MODEL_PATH):
+                 print(f"Model downloaded successfully to {LOCAL_MODEL_PATH}")
+            else:
+                 print(f"Download attempted but file still not found at {LOCAL_MODEL_PATH}. Check download path and permissions.")
+                 # Consider raising an error or exiting if download fails critically
+                 raise FileNotFoundError("Model download failed.")
+
+        except Exception as e:
+            print(f"Error downloading model: {e}")
+            # Handle error appropriately - maybe exit or try fallback
+            raise # Re-raise the exception to stop execution if model is critical
+    else:
+        print("Model file already exists locally.")
+
+# --- Call the download function at the start ---
+try:
+    download_model_if_needed()
+except Exception as e:
+    print(f"Failed to ensure model availability: {e}")
+    exit() # Exit if model download fails and is required
+
+
+
 # --- Load Structured Resume Data ---
 resume_filename = "resume_corrected.json" # Using the revamped JSON
 resume_data = {}
@@ -395,23 +450,23 @@ if not vectorstore:
 
 # --- Load Fine-tuned CTransformers model ---
 # (This part remains unchanged)
-model_path_gguf = "/data/zephyr-7b-beta.Q4_K_M.gguf" # MAKE SURE THIS PATH IS CORRECT
-print(f"Initializing Fine-Tuned CTransformers LLM from: {model_path_gguf}")
+# model_path_gguf = "/data/zephyr-7b-beta.Q4_K_M.gguf" # MAKE SURE THIS PATH IS CORRECT
+print(f"Initializing Fine-Tuned CTransformers LLM from: {LOCAL_MODEL_PATH}")
 config = {
     'max_new_tokens': 512, 'temperature': 0.1, 'context_length': 2048,
     'gpu_layers': 0, 'stream': False, 'threads': -1, 'top_k': 40,
     'top_p': 0.9, 'repetition_penalty': 1.1
     }
 llm = None
-if not os.path.exists(model_path_gguf):
+if not os.path.exists(LOCAL_MODEL_PATH):
      print(f"!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!")
-     print(f"ERROR: GGUF Model file not found at: {model_path_gguf}")
+     print(f"ERROR: GGUF Model file not found at: {LOCAL_MODEL_PATH}")
      print(f"Please download the model and place it at the correct path, or update model_path_gguf.")
      print(f"!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!")
      print("LLM initialization skipped.")
 else:
     try:
-        llm = CTransformers(model=model_path_gguf, model_type='llama', config=config)
+        llm = CTransformers(model=LOCAL_MODEL_PATH, model_type='llama', config=config)
         print("Fine-Tuned CTransformers LLM initialized.")
     except Exception as e:
         print(f"Error initializing CTransformers: {e}")