diff --git a/.gitignore b/.gitignore
index be959f73..be33f58c 100644
--- a/.gitignore
+++ b/.gitignore
@@ -160,3 +160,4 @@ test.*
 temp.*
 objdump*
 *.min.*.js
+TODO
\ No newline at end of file
diff --git a/TODO b/TODO
deleted file mode 100644
index 4ab3721b..00000000
--- a/TODO
+++ /dev/null
@@ -1 +0,0 @@
-RAG忘了触发保存了！
\ No newline at end of file
diff --git a/crazy_functions/Rag_Interface.py b/crazy_functions/Rag_Interface.py
index 9e1d9075..0de42b0c 100644
--- a/crazy_functions/Rag_Interface.py
+++ b/crazy_functions/Rag_Interface.py
@@ -1,7 +1,14 @@
 from toolbox import CatchException, update_ui, get_conf, get_log_folder, update_ui_lastest_msg
 from crazy_functions.crazy_utils import input_clipping
 from crazy_functions.crazy_utils import request_gpt_model_in_new_thread_with_ui_alive
-from crazy_functions.rag_fns.llama_index_worker import LlamaIndexRagWorker
+
+VECTOR_STORE_TYPE = "Milvus"
+
+if VECTOR_STORE_TYPE == "Simple":
+    from crazy_functions.rag_fns.llama_index_worker import LlamaIndexRagWorker
+if VECTOR_STORE_TYPE == "Milvus":
+    from crazy_functions.rag_fns.milvus_worker import MilvusRagWorker as LlamaIndexRagWorker
+
 
 RAG_WORKER_REGISTER = {}
 
@@ -14,16 +21,25 @@ def Rag问答(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, u
 
     # 1. we retrieve rag worker from global context
     user_name = chatbot.get_user()
+    checkpoint_dir = get_log_folder(user_name, plugin_name='experimental_rag')
     if user_name in RAG_WORKER_REGISTER:
         rag_worker = RAG_WORKER_REGISTER[user_name]
     else:
         rag_worker = RAG_WORKER_REGISTER[user_name] = LlamaIndexRagWorker(
             user_name, 
             llm_kwargs, 
-            checkpoint_dir=get_log_folder(user_name, plugin_name='experimental_rag'), 
+            checkpoint_dir=checkpoint_dir, 
             auto_load_checkpoint=True)
+    current_context = f"{VECTOR_STORE_TYPE} @ {checkpoint_dir}"
+    tip = "提示：输入“清空向量数据库”可以清空RAG向量数据库"
+    if txt == "清空向量数据库":
+        chatbot.append([txt, f'正在清空 ({current_context}) ...'])
+        yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
+        rag_worker.purge()
+        yield from update_ui_lastest_msg('已清空', chatbot, history, delay=0) # 刷新界面
+        return
 
-    chatbot.append([txt, '正在召回知识 ...'])
+    chatbot.append([txt, f'正在召回知识 ({current_context}) ...'])
     yield from update_ui(chatbot=chatbot, history=history) # 刷新界面
 
     # 2. clip history to reduce token consumption
@@ -68,8 +84,8 @@ def Rag问答(txt, llm_kwargs, plugin_kwargs, chatbot, history, system_prompt, u
     )
 
     # 5. remember what has been asked / answered
-    yield from update_ui_lastest_msg(model_say + '</br></br>' + '对话记忆中, 请稍等 ...', chatbot, history, delay=0.5) # 刷新界面
+    yield from update_ui_lastest_msg(model_say + '</br></br>' + f'对话记忆中, 请稍等 ({current_context}) ...', chatbot, history, delay=0.5) # 刷新界面
     rag_worker.remember_qa(i_say_to_remember, model_say)
     history.extend([i_say, model_say])
 
-    yield from update_ui_lastest_msg(model_say, chatbot, history, delay=0) # 刷新界面
+    yield from update_ui_lastest_msg(model_say, chatbot, history, delay=0, msg=tip) # 刷新界面
diff --git a/crazy_functions/rag_fns/llama_index_worker.py b/crazy_functions/rag_fns/llama_index_worker.py
index de1ef38d..761d6943 100644
--- a/crazy_functions/rag_fns/llama_index_worker.py
+++ b/crazy_functions/rag_fns/llama_index_worker.py
@@ -1,4 +1,7 @@
 import llama_index
+import os
+import atexit
+from typing import List
 from llama_index.core import Document
 from llama_index.core.schema import TextNode
 from request_llms.embed_models.openai_embed import OpenAiEmbeddingModel
@@ -38,6 +41,7 @@ class SaveLoad():
         return True
 
     def save_to_checkpoint(self, checkpoint_dir=None):
+        print(f'saving vector store to: {checkpoint_dir}')
         if checkpoint_dir is None: checkpoint_dir = self.checkpoint_dir
         self.vs_index.storage_context.persist(persist_dir=checkpoint_dir)
 
@@ -65,7 +69,8 @@ class LlamaIndexRagWorker(SaveLoad):
         if auto_load_checkpoint:
             self.vs_index = self.load_from_checkpoint(checkpoint_dir)
         else:
-            self.vs_index = self.create_new_vs()
+            self.vs_index = self.create_new_vs(checkpoint_dir)
+        atexit.register(lambda: self.save_to_checkpoint(checkpoint_dir))
 
     def assign_embedding_model(self):
         pass
@@ -117,6 +122,3 @@ class LlamaIndexRagWorker(SaveLoad):
         buf = "\n".join(([f"(No.{i+1} | score {n.score:.3f}): {n.text}" for i, n in enumerate(nodes)]))
         if self.debug_mode: print(buf)
         return buf
-
-
-
diff --git a/crazy_functions/rag_fns/milvus_worker.py b/crazy_functions/rag_fns/milvus_worker.py
new file mode 100644
index 00000000..4b5b0ad9
--- /dev/null
+++ b/crazy_functions/rag_fns/milvus_worker.py
@@ -0,0 +1,107 @@
+import llama_index
+import os
+import atexit
+from typing import List
+from llama_index.core import Document
+from llama_index.core.schema import TextNode
+from request_llms.embed_models.openai_embed import OpenAiEmbeddingModel
+from shared_utils.connect_void_terminal import get_chat_default_kwargs
+from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
+from crazy_functions.rag_fns.vector_store_index import GptacVectorStoreIndex
+from llama_index.core.ingestion import run_transformations
+from llama_index.core import PromptTemplate
+from llama_index.core.response_synthesizers import TreeSummarize
+from llama_index.core import StorageContext
+from llama_index.vector_stores.milvus import MilvusVectorStore
+from crazy_functions.rag_fns.llama_index_worker import LlamaIndexRagWorker
+
+DEFAULT_QUERY_GENERATION_PROMPT = """\
+Now, you have context information as below:
+---------------------
+{context_str}
+---------------------
+Answer the user request below (use the context information if necessary, otherwise you can ignore them):
+---------------------
+{query_str}
+"""
+
+QUESTION_ANSWER_RECORD = """\
+{{
+    "type": "This is a previous conversation with the user",
+    "question": "{question}",
+    "answer": "{answer}",
+}}
+"""
+
+
+class MilvusSaveLoad():
+
+    def does_checkpoint_exist(self, checkpoint_dir=None):
+        import os, glob
+        if checkpoint_dir is None: checkpoint_dir = self.checkpoint_dir
+        if not os.path.exists(checkpoint_dir): return False
+        if len(glob.glob(os.path.join(checkpoint_dir, "*.json"))) == 0: return False
+        return True
+
+    def save_to_checkpoint(self, checkpoint_dir=None):
+        print(f'saving vector store to: {checkpoint_dir}')
+        # if checkpoint_dir is None: checkpoint_dir = self.checkpoint_dir
+        # self.vs_index.storage_context.persist(persist_dir=checkpoint_dir)
+
+    def load_from_checkpoint(self, checkpoint_dir=None):
+        if checkpoint_dir is None: checkpoint_dir = self.checkpoint_dir
+        if self.does_checkpoint_exist(checkpoint_dir=checkpoint_dir):
+            print('loading checkpoint from disk')
+            from llama_index.core import StorageContext, load_index_from_storage
+            storage_context = StorageContext.from_defaults(persist_dir=checkpoint_dir)
+            try:
+                self.vs_index = load_index_from_storage(storage_context, embed_model=self.embed_model)
+                return self.vs_index
+            except:
+                return self.create_new_vs(checkpoint_dir)
+        else:
+            return self.create_new_vs(checkpoint_dir)
+
+    def create_new_vs(self, checkpoint_dir, overwrite=False):
+        vector_store = MilvusVectorStore(
+            uri=os.path.join(checkpoint_dir, "milvus_demo.db"), 
+            dim=self.embed_model.embedding_dimension(),
+            overwrite=overwrite
+        )
+        storage_context = StorageContext.from_defaults(vector_store=vector_store)
+        index = GptacVectorStoreIndex.default_vector_store(storage_context=storage_context, embed_model=self.embed_model)
+        return index
+
+    def purge(self):
+        self.vs_index = self.create_new_vs(self.checkpoint_dir, overwrite=True)
+
+class MilvusRagWorker(MilvusSaveLoad, LlamaIndexRagWorker):
+
+    def __init__(self, user_name, llm_kwargs, auto_load_checkpoint=True, checkpoint_dir=None) -> None:
+        self.debug_mode = True
+        self.embed_model = OpenAiEmbeddingModel(llm_kwargs)
+        self.user_name = user_name
+        self.checkpoint_dir = checkpoint_dir
+        if auto_load_checkpoint:
+            self.vs_index = self.load_from_checkpoint(checkpoint_dir)
+        else:
+            self.vs_index = self.create_new_vs(checkpoint_dir)
+        atexit.register(lambda: self.save_to_checkpoint(checkpoint_dir))
+
+    def inspect_vector_store(self):
+        # This function is for debugging
+        try:
+            self.vs_index.storage_context.index_store.to_dict()
+            docstore = self.vs_index.storage_context.docstore.docs
+            if not docstore.items():
+                raise ValueError("cannot inspect")
+            vector_store_preview = "\n".join([ f"{_id} | {tn.text}" for _id, tn in docstore.items() ])
+        except:
+            dummy_retrieve_res: List["NodeWithScore"] = self.vs_index.as_retriever().retrieve(' ')
+            vector_store_preview = "\n".join(
+                [f"{node.id_} | {node.text}" for node in dummy_retrieve_res]
+            )
+        print('\n++ --------inspect_vector_store begin--------')
+        print(vector_store_preview)
+        print('oo --------inspect_vector_store end--------')
+        return vector_store_preview
diff --git a/request_llms/embed_models/openai_embed.py b/request_llms/embed_models/openai_embed.py
index c559e1c3..9d565173 100644
--- a/request_llms/embed_models/openai_embed.py
+++ b/request_llms/embed_models/openai_embed.py
@@ -71,7 +71,13 @@ class OpenAiEmbeddingModel(EmbeddingModel):
             embedding = res.data[0].embedding
         return embedding
 
-    def embedding_dimension(self, llm_kwargs):
+    def embedding_dimension(self, llm_kwargs=None):
+        # load kwargs
+        if llm_kwargs is None:
+            llm_kwargs = self.llm_kwargs
+        if llm_kwargs is None:
+            raise RuntimeError("llm_kwargs is not provided!")
+
         from .bridge_all_embed import embed_model_info
         return embed_model_info[llm_kwargs['embed_model']]['embed_dimension']
 
diff --git a/requirements.txt b/requirements.txt
index 757df774..99c841e2 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -7,7 +7,9 @@ tiktoken>=0.3.3
 requests[socks]
 pydantic==2.5.2
 llama-index==0.10.47
-protobuf==3.18
+llama-index-vector-stores-milvus==0.1.16
+pymilvus==2.4.2
+protobuf==3.20
 transformers>=4.27.1,<4.42
 scipdf_parser>=0.52
 anthropic>=0.18.1
diff --git a/toolbox.py b/toolbox.py
index 6b2f4c10..900cf234 100644
--- a/toolbox.py
+++ b/toolbox.py
@@ -178,7 +178,7 @@ def update_ui(chatbot:ChatBotWithCookies, history, msg="正常", **kwargs):  # 
     yield cookies, chatbot_gr, history, msg
 
 
-def update_ui_lastest_msg(lastmsg:str, chatbot:ChatBotWithCookies, history:list, delay=1):  # 刷新界面
+def update_ui_lastest_msg(lastmsg:str, chatbot:ChatBotWithCookies, history:list, delay=1, msg="正常"):  # 刷新界面
     """
     刷新用户界面
     """
@@ -186,7 +186,7 @@ def update_ui_lastest_msg(lastmsg:str, chatbot:ChatBotWithCookies, history:list,
         chatbot.append(["update_ui_last_msg", lastmsg])
     chatbot[-1] = list(chatbot[-1])
     chatbot[-1][-1] = lastmsg
-    yield from update_ui(chatbot=chatbot, history=history)
+    yield from update_ui(chatbot=chatbot, history=history, msg=msg)
     time.sleep(delay)