changed default split values

app.py

@@ -9,6 +9,7 @@ from langchain_openai import ChatOpenAI, OpenAIEmbeddings
 from langchain_text_splitters import RecursiveCharacterTextSplitter
 from langchain.callbacks import get_openai_callback
 from langchain_core.prompts import ChatPromptTemplate
+from langchain_core.runnables import RunnableParallel
 
 from langchain_community.document_loaders import UnstructuredFileLoader
 
@@ -26,11 +27,11 @@ for file in files:
 # Config
 with st.sidebar:
     st.write(f"Injected documents: \n\n {'\n'.join('\n'+file for file in files)}")
-    model = st.selectbox("Model name", ["gpt-3.5-turbo"])
+    model = st.selectbox("Model name", ["gpt-3.5-turbo"], disabled=True)
     temperature = st.number_input("Temperature", value=0.0, min_value=0.0, step=0.2, max_value=1.0, placeholder=0.0)
-    if st.toggle("Splitting"):
-        chunk_size = st.number_input("Chunk size", value=500, step=250, placeholder=500) # Defines the chunks in amount of tokens in which the files are split. Also defines the amount of tokens that are feeded into the context. 
-        chunk_overlap = st.number_input("Chunk overlap", value=100, step=10, placeholder=100)
+    if st.toggle("Splitting", value=True):
+        chunk_size = st.number_input("Chunk size", value=750, step=250, placeholder=750) # Defines the chunks in amount of tokens in which the files are split. Also defines the amount of tokens that are feeded into the context. 
+        chunk_overlap = st.number_input("Chunk overlap", value=0, step=10, placeholder=0)
         text_splitter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
         splits = text_splitter.split_documents(docs)
         vectorstore = Chroma.from_documents(documents=splits, embedding=OpenAIEmbeddings())
@@ -67,13 +68,26 @@ def format_docs(docs):
     return "\n\n".join(doc.page_content for doc in docs)
 
 
-rag_chain = (
-    {"context": retriever | format_docs, "question": RunnablePassthrough()}
+# rag_chain = (
+#     {"context": retriever | format_docs, "question": RunnablePassthrough()}
+#     | prompt
+#     | llm
+#     | StrOutputParser()
+# )
+
+
+
+rag_chain_from_docs = (
+    RunnablePassthrough.assign(context=(lambda x: format_docs(x["context"])))
     | prompt
     | llm
     | StrOutputParser()
 )
 
+rag_chain = RunnableParallel(
+    {"context": retriever, "question": RunnablePassthrough()}
+).assign(answer=rag_chain_from_docs)
+
 
 st.title("🐔 Volker-Chat")
 
@@ -81,6 +95,7 @@ def click_button(prompt):
     st.session_state.clicked = True
     st.session_state['prompt'] = prompt
 
+
 c = st.container()
 c.write("Beispielfragen")
 col1, col2, col3 = c.columns(3)
@@ -104,19 +119,27 @@ if st.session_state.clicked:
     st.chat_message("user").write(prompt)
     with get_openai_callback() as cb:
         response = rag_chain.invoke(prompt)
-        st.chat_message("assistant").write(response)
+        st.chat_message("assistant").write(response['answer'])
+        with st.expander("Kontext ansehen"):
+            st.write(response["context"])
         with st.sidebar:
             sidebar_c = st.container()
             sidebar_c.success(cb)
-    st.session_state.clicked = False
+
 
 if prompt := st.chat_input():
     st.chat_message("user").write(prompt)
     with get_openai_callback() as cb:
         response = rag_chain.invoke(prompt)
-        st.chat_message("assistant").write(response)
+        st.chat_message("assistant").write(response['answer'])
+        with st.expander("Kontext ansehen"):
+            st.write(response["context"])
         with st.sidebar:
             sidebar_c = st.container()
             sidebar_c.success(cb)
-    st.session_state.clicked = False
+
+
+# cleanup
+st.session_state.clicked = False
+vectorstore.delete_collection()