s

app.py

@@ -6,6 +6,12 @@ import asyncio
 from contextlib import asynccontextmanager
 import tempfile
 
+# Set environment variables for Hugging Face Spaces
+os.environ["TRANSFORMERS_CACHE"] = "/tmp/transformers_cache"
+os.environ["HF_HOME"] = "/tmp/hf_home"
+os.environ["HUGGINGFACE_HUB_CACHE"] = "/tmp/hf_hub_cache"
+os.environ["SENTENCE_TRANSFORMERS_HOME"] = "/tmp/sentence_transformers"
+
 import requests
 from fastapi import FastAPI, HTTPException, BackgroundTasks
 from fastapi.middleware.cors import CORSMiddleware
@@ -39,11 +45,15 @@ from qdrant_client.http.models import Distance, VectorParams, PointStruct
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
-# Environment variables
+# Environment variables - Use environment variables or fallback to defaults
 GROQ_API_KEY = os.getenv("GROQ_API_KEY")
 QDRANT_URL = os.getenv("QDRANT_URL")
 QDRANT_API_KEY = os.getenv("QDRANT_API_KEY")
 
+# Create cache directories if they don't exist
+for cache_dir in ["/tmp/transformers_cache", "/tmp/hf_home", "/tmp/hf_hub_cache", "/tmp/sentence_transformers"]:
+    os.makedirs(cache_dir, exist_ok=True)
+
 if not GROQ_API_KEY:
     raise ValueError("GROQ_API_KEY environment variable is required")
 
@@ -79,16 +89,38 @@ async def lifespan(app: FastAPI):
             verify=False
         )
     
-    # Initialize embeddings model
-    embeddings_model = HuggingFaceEmbeddings(
-        model_name="sentence-transformers/all-MiniLM-L6-v2",
-        model_kwargs={'device': 'cpu'}
-    )
+    # Initialize embeddings model with error handling
+    try:
+        # Try to initialize with custom cache directory
+        embeddings_model = HuggingFaceEmbeddings(
+            model_name="sentence-transformers/all-MiniLM-L6-v2",
+            model_kwargs={
+                'device': 'cpu',
+                'cache_folder': '/tmp/sentence_transformers'
+            },
+            cache_folder='/tmp/sentence_transformers'
+        )
+        logger.info("HuggingFace embeddings initialized successfully")
+    except Exception as e:
+        logger.error(f"Failed to initialize HuggingFace embeddings: {e}")
+        # Try with a smaller model that might work better
+        try:
+            embeddings_model = HuggingFaceEmbeddings(
+                model_name="all-MiniLM-L6-v2",  # Try without the prefix
+                model_kwargs={'device': 'cpu'},
+                cache_folder='/tmp/sentence_transformers'
+            )
+            logger.info("HuggingFace embeddings initialized with fallback model")
+        except Exception as e2:
+            logger.error(f"Failed to initialize fallback embeddings: {e2}")
+            # Use a minimal fallback - this will cause issues but app will start
+            embeddings_model = None
+            logger.warning("No embeddings model available - embedding operations will fail")
     
     # Initialize LLM
     llm = ChatGroq(
         groq_api_key=GROQ_API_KEY,
-        model_name="openai/gpt-oss-20b",
+        model_name="mixtral-8x7b-32768",
         temperature=0
     )
     
@@ -285,6 +317,9 @@ async def extract_pdf_pdfplumber(pdf_path: str) -> str:
 async def store_in_qdrant(documents: List[Document], collection_name: str):
     """Store documents in Qdrant vector database"""
     try:
+        if embeddings_model is None:
+            raise ValueError("Embeddings model not available")
+        
         # Create collection if it doesn't exist
         try:
             qdrant_client.get_collection(collection_name)
@@ -329,6 +364,9 @@ async def store_in_qdrant(documents: List[Document], collection_name: str):
 def retriever_tool(query: str, collection_name: str) -> str:
     """Retrieve relevant documents from Qdrant based on the query."""
     try:
+        if embeddings_model is None:
+            return "Embeddings model not available. Cannot perform retrieval."
+        
         # Generate query embedding
         query_embedding = embeddings_model.embed_query(query)
         

req.txt

@@ -1,18 +1,17 @@
 fastapi
 uvicorn
-langchain
 langchain-core
 langchain-groq
-langchain-community
+langchain-huggingface
 langgraph
-docling
 qdrant-client
+docling
+PyPDF2
+pdfplumber
+requests
+pydantic
 sentence-transformers
 transformers
 torch
-requests
-pydantic
-python-multipart
-numpy
-pandas
-Pillow
+torchvision
+torchaudio