Update app.py

app.py

@@ -58,6 +58,7 @@ class CustomLLM(LLM):
 llm = CustomLLM(model=model)
 
 import gradio as gr
+from langchain.prompts import PromptTemplate
 from langchain.docstore.document import Document
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain.chains.question_answering import load_qa_chain
@@ -66,21 +67,48 @@ from langchain.chains.question_answering import load_qa_chain
 
 # embeddings = HuggingFaceEmbeddings()
 query = "總結並以點列形式舉出重點"
-chain = load_qa_chain(llm, chain_type="map_reduce")
-# chain = load_qa_chain(llm, chain_type="refine")
+prompt_template = """總結下文並列舉出重點：
+
+
+{text}
+
+
+摘要及各項重點："""
+PROMPT = PromptTemplate(template=prompt_template, input_variables=["text"])
+# chain = load_summarize_chain(llm, chain_type="stuff", prompt=PROMPT)
+chain = load_summarize_chain(llm, chain_type="map_reduce", map_prompt=PROMPT, combine_prompt=PROMPT)
+# refine_template = (
+#     "你的任務是整理出一段摘要以及例舉所有重點\n"
+#     "我們之前已經整理出這些內容: {existing_answer}\n"
+#     "請再整合這些摘要並將重點整理到一個列表"
+#     "(如果需要) 下文這裡有更多的參考資料：\n"
+#     "------------\n"
+#     "{text}\n"
+#     "------------\n"
+#     "根據新的資料，完善原有的摘要和重點列表"
+#     "如果新資料對已經整理出的文字沒有補充，請重複原來的重點文字。"
+# )
+# refine_prompt = PromptTemplate(
+#     input_variables=["existing_answer", "text"],
+#     template=refine_template,
+# )
+# chain = load_summarize_chain(llm, chain_type="refine", question_prompt=PROMPT, refine_prompt=refine_prompt)
+# chain = load_qa_chain(llm, chain_type="map_reduce", map_prompt=PROMPT, combine_prompt=PROMPT)
+# chain = load_qa_chain(llm, chain_type="refine", question_prompt=PROMPT, refine_prompt=refine_prompt)
 
 def greet(text):
     docs = [Document(page_content=text)]
 
     text_splitter = RecursiveCharacterTextSplitter(
-        chunk_size=1024, # 分割最大尺寸
+        chunk_size=512, # 分割最大尺寸
         chunk_overlap=64, # 重复字数
         length_function=len
     )
     texts = text_splitter.split_documents(docs)
     # docsearch = Chroma.from_texts(texts, embeddings).as_retriever()
     # docs = docsearch.get_relevant_documents(query)
-    return chain.run(input_documents=texts, question=query)
+    return chain.run(texts)
+    # return chain.run(input_documents=texts, question=query)
 
 iface = gr.Interface(fn=greet,
                      inputs=gr.Textbox(lines=20,