Update app.py

app.py

@@ -19,7 +19,6 @@ from pydantic import BaseModel
 from streamlit_pills import pills
 from dense_embed import embed_text
 from ppt_chunker import ppt_chunker
-from huggingface_hub import snapshot_download
 from streamlit_navigation_bar import st_navbar
 from vllm.sampling_params import SamplingParams
 from outlines.fsm.json_schema import build_regex_from_schema
@@ -51,7 +50,6 @@ from scipy.sparse import (
 )
 from prompts import (
     outlines,
-    transform_query,
     build_prompt_conv,
     route_llm,
     open_query_prompt,
@@ -123,7 +121,7 @@ def generate_answer(query: str,
 
         answer = f"{text}\n\n\nSource :\n\n{metadata}"
     else:
-        dense_embeddings, tokens_count = asyncio.run(embed_text(dense_model[0], transform_query(query)))
+        dense_embeddings, tokens_count = asyncio.run(embed_text(dense_model[0], query))
             
         scored_points = query_hybrid_search(query, client, collection_name, dense_embeddings, sparse_embeddings).points
  
@@ -217,11 +215,11 @@ def load_models_and_documents():
         dense_model = AsyncEngineArray.from_args(
             [
                 EngineArgs(
-                    model_name_or_path='EmbeddedLLM/bge-m3-onnx-o2-cpu', 
-                    engine='optimum',
+                    model_name_or_path='BAAI/bge-m3', 
+                    engine='torch',
                     device='cpu',
                     embedding_dtype='float32',
-                    dtype='float32',
+                    dtype='float16',
                     pooling_method='cls',
                     lengths_via_tokenize=True
                 )