fix

Author: prakash100198

devtron-docs-rag-server/Dockerfile

@@ -6,6 +6,12 @@
 # Stage 1: Builder - Install dependencies
 FROM python:3.12-slim AS builder
 
+# Set cache directories BEFORE downloading anything
+ENV TRANSFORMERS_CACHE=/root/.cache/huggingface \
+    HF_HOME=/root/.cache/huggingface \
+    TORCH_HOME=/root/.cache/torch \
+    SENTENCE_TRANSFORMERS_HOME=/root/.cache/torch/sentence_transformers
+
 # Install build dependencies
 RUN apt-get update && \
     apt-get install -y --no-install-recommends \
@@ -38,7 +44,10 @@ RUN pip install --user --no-cache-dir -r requirements.txt && \
 # This prevents downloading ~1.34GB on every container startup
 COPY download_model.py .
 RUN python download_model.py BAAI/bge-large-en-v1.5 && \
-    rm download_model.py
+    rm download_model.py && \
+    echo "Verifying model cache..." && \
+    ls -lah /root/.cache/torch/sentence_transformers/ && \
+    echo "Model cache verified!"
 
 # ============================================================================
 # Stage 2: Runtime - Minimal production image
@@ -82,6 +91,8 @@ ENV PYTHONUNBUFFERED=1 \
     HF_HOME=/root/.cache/huggingface \
     TORCH_HOME=/root/.cache/torch \
     SENTENCE_TRANSFORMERS_HOME=/root/.cache/torch/sentence_transformers \
+    HF_HUB_OFFLINE=1 \
+    TRANSFORMERS_OFFLINE=1 \
     PIP_NO_CACHE_DIR=1
 
 # Expose port

devtron-docs-rag-server/download_model.py

@@ -5,6 +5,7 @@
 """
 
 import logging
+import os
 import sys
 from sentence_transformers import SentenceTransformer
 
@@ -16,21 +17,35 @@
 
 def download_model(model_name: str = "BAAI/bge-large-en-v1.5"):
     """Download and cache the embedding model."""
+
+    # Verify cache directories are set
+    cache_dir = os.getenv('SENTENCE_TRANSFORMERS_HOME')
+    logger.info(f"Cache directory: {cache_dir}")
     logger.info(f"Downloading embedding model: {model_name}")
     logger.info("This will download ~1.34GB and may take several minutes...")
-    
+
     try:
+        # Download model - it will use SENTENCE_TRANSFORMERS_HOME env var automatically
         model = SentenceTransformer(model_name)
         dimension = model.get_sentence_embedding_dimension()
-        
+
         logger.info(f"✓ Model downloaded successfully!")
         logger.info(f"  Model: {model_name}")
         logger.info(f"  Embedding dimension: {dimension}")
-        logger.info(f"  Model is now cached and ready to use")
-        
+        logger.info(f"  Cache location: {cache_dir}")
+
+        # Verify the cache exists
+        if cache_dir and os.path.exists(cache_dir):
+            logger.info(f"  Cache verified at: {cache_dir}")
+            # List contents
+            for root, dirs, files in os.walk(cache_dir):
+                logger.info(f"    {root}: {len(files)} files")
+
         return True
     except Exception as e:
         logger.error(f"✗ Failed to download model: {str(e)}")
+        import traceback
+        traceback.print_exc()
         return False
 
 if __name__ == "__main__":

devtron-docs-rag-server/vector_store.py

@@ -4,6 +4,7 @@
 
 import logging
 import json
+import os
 from typing import List, Dict, Any, Optional
 from pathlib import Path
 import hashlib
@@ -27,12 +28,23 @@ def __init__(self, model_name: str = "BAAI/bge-large-en-v1.5"):
             model_name: HuggingFace model name
         """
         logger.info(f"Loading embedding model: {model_name}")
+
+        # Verify cache directory exists
+        cache_dir = os.getenv('SENTENCE_TRANSFORMERS_HOME')
+        if cache_dir and os.path.exists(cache_dir):
+            logger.info(f"Using cached model from: {cache_dir}")
+        else:
+            logger.warning(f"Cache directory not found: {cache_dir}")
+
         try:
+            # Load model - it will use SENTENCE_TRANSFORMERS_HOME env var automatically
             self.model = SentenceTransformer(model_name)
             self.dimension = self.model.get_sentence_embedding_dimension()
             logger.info(f"✓ Embedding model loaded (dimension: {self.dimension})")
         except Exception as e:
             logger.error(f"✗ Failed to load embedding model: {str(e)}")
+            logger.error(f"Cache directory: {cache_dir}")
+            logger.error(f"Cache exists: {os.path.exists(cache_dir) if cache_dir else 'N/A'}")
             raise
 
     def embed_documents(self, texts: List[str]) -> List[List[float]]: