v 0.1

- добавлен llama-index

app.py

@@ -30,9 +30,9 @@ set_global_tokenizer(
     AutoTokenizer.from_pretrained("Qwen/Qwen2.5-7B-Instruct").encode
 )
 
-embed_model = HuggingFaceEmbedding(model_name="sentence-transformers/all-mpnet-base-v2")
-
-documents = SimpleDirectoryReader("./data/").load_data()
+# embed_model = HuggingFaceEmbedding(model_name="sentence-transformers/all-mpnet-base-v2")
+#
+# documents = SimpleDirectoryReader("./data/").load_data()
 
 def messages_to_prompt(messages):
     messages = [{"role": m.role.value, "content": m.content} for m in messages]
@@ -74,9 +74,9 @@ llm = LlamaCPP(
 )
 
 memory = ChatMemoryBuffer.from_defaults(token_limit=3900)
-index = VectorStoreIndex.from_documents(documents, embed_model=embed_model)
+index = VectorStoreIndex
 chat_engine = index.as_chat_engine(
-    chat_mode="condense_plus_context",
+    chat_mode="beast",
     memory=memory,
     llm=llm,
     context_propt=(