Fix Issue#2: Reduce the memory

* Add the `st.cache_resource` avoid memory abnormal increase.
* Change the HuggingFaceEmbedding model from default to `multi-qa-MiniLM-L6-cos-v1`. (Lower memory usage!)
  (see commit: 8fd68f8 )

* Using Windows task manager track the memory usages (unit: KB):

---

init the app: 15w
add pdf(8.1m): 47w
remove pdf: 47w
add same pdf(8.1m): 57w
remove pdf: 47w
add new pdf(8.8m): 58w
add original pdf(8.1m): 59w

init the app: 15w
add pdf(8.1m): 47w
remove pdf: 47w
add same pdf(8.1m): 47w
remove pdf: 47w
add new pdf(8.8m): 57w
add original pdf(8.1m): 57w

---

Reference:
[Stackoverflow](https://stackoverflow.com/questions/77013746/how-to-release-memory-correctly-in-streamlit-app/77016325?noredirect=1#comment135777962_7701632501)

Author: Lin-jun-xiang

app.py

@@ -151,8 +151,14 @@ def get_response(query: str) -> str:
 
 with doc_container:
     docs = upload_and_process_pdf()
-    model = create_doc_gpt(docs)
-    del docs
+
+    if docs:
+        model = create_doc_gpt(
+            docs,
+            {k: v for k, v in docs[0].metadata.items() if k not in ['source', 'file_path']},
+            st.session_state.g4f_provider
+        )
+        del docs
     st.write('---')
 
 if 'response' not in st.session_state:

docGPT/__init__.py

@@ -14,11 +14,13 @@
 module_logger = logger.get_logger(__name__)
 
 
-def create_doc_gpt(docs):
-    if not docs:
-        return
-
-    docGPT = DocGPT(docs=docs)
+@st.cache_resource(ttl=1800, max_entries=10)
+def create_doc_gpt(
+    _docs: list,
+    doc_metadata: str,
+    g4f_provider: str
+) -> DocGPT:
+    docGPT = DocGPT(docs=_docs)
 
     try:
         if OpenAiAPI.is_valid():
@@ -33,10 +35,8 @@ def create_doc_gpt(docs):
             )
             docGPT.llm = llm_model
             agent_.llm = llm_model
-            with st.spinner('Running...'):
-                docGPT.create_qa_chain(
-                    chain_type='refine',
-                )
+
+            docGPT.create_qa_chain(chain_type='refine')
             docGPT_tool = agent_.create_doc_chat(docGPT)
             calculate_tool = agent_.get_calculate_chain
             llm_tool = agent_.create_llm_chain()
@@ -58,14 +58,12 @@ def create_doc_gpt(docs):
             # Use gpt4free llm model without agent
             llm_model = GPT4Free(
                 provider=GPT4Free().PROVIDER_MAPPING[
-                    st.session_state.g4f_provider
+                    g4f_provider
                 ]
             )
+            print(GPT4Free().PROVIDER_MAPPING[g4f_provider])
             docGPT.llm = llm_model
-            with st.spinner('Running...(free model will take more time)'):
-                docGPT.create_qa_chain(
-                    chain_type='refine',
-                )
+            docGPT.create_qa_chain(chain_type='refine')
             return docGPT
     except Exception as e:
         module_logger.info(f'{__file__}: {e}')

docGPT/docGPT.py

@@ -159,6 +159,7 @@ def _embeddings(self):
             documents=self.docs,
             embedding=embeddings
         )
+        print('embedded...')
         return db
 
     def create_qa_chain(