fix: use google/flan-t5-large model for HF Inference API

config.py

@@ -37,7 +37,8 @@ CHROMA_COLLECTION_NAME = "document_embeddings"
 CHROMA_DISTANCE_METRIC = "cosine"
 
 # LLM Configuration - Hugging Face Inference API
-HF_MODEL = "meta-llama/Llama-3.2-3B-Instruct"
+# Using a reliable model that works with HF Inference API free tier
+HF_MODEL = "google/flan-t5-large"
 HF_TOKEN = os.getenv("HF_TOKEN", "")
 
 # Retrieval Configuration