clean final: HF Inference API embeddings (no disk)

rag.py

@@ -1,4 +1,3 @@
-# rag.py – bullet-proof: online fetch with fallback on any error
 from __future__ import annotations
 import os, re, json, requests
 from functools import lru_cache
@@ -94,28 +93,22 @@ def get_texts() -> List[str]:
 @lru_cache(maxsize=1)
 def get_vectorstore() -> FAISS:
     texts = get_texts()
-    embeddings = None
 
     try:
-        # Try new API first
         from langchain_huggingface import HuggingFaceInferenceAPIEmbeddings
         embeddings = HuggingFaceInferenceAPIEmbeddings(
             api_key=HF_TOKEN,
             model_name="sentence-transformers/all-MiniLM-L6-v2"
         )
     except ImportError:
-        try:
-            # Fallback for updated version
-            from langchain_huggingface import HuggingFaceEmbeddings
-            embeddings = HuggingFaceEmbeddings(
-                model_name="sentence-transformers/all-MiniLM-L6-v2"
-            )
-            print("⚙️ Using HuggingFaceEmbeddings fallback")
-        except Exception as e:
-            raise RuntimeError(f"❌ Failed to load embeddings: {e}")
+        from langchain.embeddings import HuggingFaceEmbeddings
+        embeddings = HuggingFaceEmbeddings(
+            model_name="sentence-transformers/all-MiniLM-L6-v2"
+        )
+        print("⚠️ Falling back to HuggingFaceEmbeddings (local) – inference API not available")
 
     if not texts:
-        return FAISS.from_texts([""], embeddings)  # dummy FAISS instance
+        return FAISS.from_texts([""], embeddings)
 
     splitter = RecursiveCharacterTextSplitter(chunk_size=600, chunk_overlap=50)
     docs = splitter.create_documents(texts, metadatas=[{"source": DATASET}] * len(texts))
@@ -131,13 +124,11 @@ def get_llm():
         huggingfacehub_api_token=HF_TOKEN
     )
 
-PROMPT = PromptTemplate.from_template("""
-You are Amina, assistant for {company}.
+PROMPT = PromptTemplate.from_template("""You are Amina, assistant for {company}.
 Use only the context below. If unsure, say: "A human agent will follow up."
 Context: {context}
 Question: {question}
-Answer:
-""")
+Answer:""")
 
 # ------------------------------------------------------------------ MAIN
 def ask_question(phone: str, question: str) -> Tuple[str, List]:
@@ -169,4 +160,4 @@ def ask_question(phone: str, question: str) -> Tuple[str, List]:
 
 def _save_chat(phone: str, q: str, a: str) -> None:
     supabase.table("chat_memory").insert({"user_phone": phone, "role": "user", "message": q}).execute()
-    supabase.table("chat_memory").insert({"user_phone": phone, "role": "assistant", "message": a}).execute()
+    supabase.table("chat_memory").insert({"user_phone": phone, "role": "assistant", "message": a}).execute()