Update app.py

app.py

@@ -21,6 +21,7 @@ from level_classifier_tool_2 import (
 from task_temp import rag_temp, rag_cls_temp, cls_temp, gen_temp
 from all_tools import classify_and_score, QuestionRetrieverTool
 from phrases import BLOOMS_PHRASES, DOK_PHRASES
+import spaces
 # Prebuild embeddings once
 _backend = HFEmbeddingBackend(model_name="sentence-transformers/all-MiniLM-L6-v2")
 _BLOOM_INDEX = build_phrase_index(_backend, BLOOMS_PHRASES)
@@ -46,16 +47,18 @@ all_questions = (
     list(D["ScienceQA"]) +
     list(D["PubmedQA"])
 )
-texts = all_questions.
-@spaces(15)
+texts = all_questions
+@spaces.GPU(15)
 def build_indexes_on_gpu(model="google/embeddinggemma-300m"):
     device = 'cuda'
     emb = HuggingFaceEmbeddings(
         model_name="model",
         model_kwargs={"device": device},
-        encode_kwargs={"normalize_embeddings": True},
-    index = VectorStoreIndex.from_documents([Document(text=t) for t in texts], embed_model=emb)
-    return index
+        encode_kwargs={"normalize_embeddings": True})
+    idx = VectorStoreIndex.from_documents([Document(text=t) for t in texts], embed_model=emb)
+    return idx
+    device = "cuda"
+
 index = build_indexes_on_gpu(model="google/embeddinggemma-300m")
 # ------------------------ Agent setup with timeout ------------------------
 def make_agent(hf_token: str, model_id: str, provider: str, timeout: int, temperature: float, max_tokens: int):