chatbot

app.py

@@ -48,11 +48,7 @@ def process_llm_response(llm_response):
     # return json.dumps(response_data)
     return response_data
     
-def get_answer(question):
-    llm_response = qa_chain(question)
-    response = process_llm_response(llm_response)
 
-    return response["result"], response["sources"]
 
 # @app.route('/question', methods=['POST'])
 # def answer():
@@ -64,51 +60,7 @@ def get_answer(question):
 #         return response
 #     else:
 #         return 'Content-Type not supported!'
-    
-# @app.route('/', methods=['GET'])
-# def default():
-#     return "Hello World!"
-
-
-# if __name__ == '__main__':
-#     ip=get_local_ip()
-#     os.environ["OPENAI_API_KEY"] = "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"
-#     # Embed and store the texts
-#     # if(torch.cuda.is_available() == False):
-#     #     print("No GPU available")
-#     #     exit(1)
-    
-#     torch.cuda.empty_cache()
-#     torch.max_split_size_mb = 100
-#     instructor_embeddings = HuggingFaceInstructEmbeddings(model_name="hkunlp/instructor-xl", 
-#                                                       model_kwargs={"device": "cpu"})
-#     # Supplying a persist_directory will store the embeddings on disk
-#     persist_directory = 'db'
-#     vectordb2 = Chroma(persist_directory=persist_directory, 
-#                   embedding_function=instructor_embeddings,
-#                    )
-#     retriever = vectordb2.as_retriever(search_kwargs={"k": 3})
-#     vectordb2.persist()
-
-#     # Set up the turbo LLM
-#     turbo_llm = ChatOpenAI(
-#         temperature=0,
-#         model_name='gpt-3.5-turbo'
-#     )
-#     qa_chain = RetrievalQA.from_chain_type(llm=turbo_llm, 
-#                                   chain_type="stuff", 
-#                                   retriever=retriever, 
-#                                   return_source_documents=True)
-#     qa_chain.combine_documents_chain.llm_chain.prompt.messages[0].prompt.template= """
-#     Use only the following pieces of context and think step by step to answer. Answer the users question only if they are related to the context given.
-#     If you don't know the answer, just say that you don't know, don't try to make up an answer. Make your answer very detailed and long. 
-#     Use bullet points to explain when required. 
-#     Use only text found in the context as your knowledge source for the answer. 
-#     ----------------
-#     {context}"""
-#     app.run(host=ip, port=5000)
-# 
-# 
+
 
 ip=get_local_ip()
 os.environ["OPENAI_API_KEY"] = "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"
@@ -151,18 +103,37 @@ def print_array(arr):
     arr_str = str(arr)
     return arr_str
 
-def getanswer(question):
-    llm_response = qa_chain(question)
-    response = process_llm_response(llm_response)
-    sources= print_array(response["sources"])
-    return response["result"], sources
+# def getanswer(question):
+#     llm_response = qa_chain(question)
+#     response = process_llm_response(llm_response)
+#     sources= print_array(response["sources"])
+#     return response["result"], sources
 
 # iface = gr.Interface(fn=getanswer, inputs="text", outputs="text")
 # iface.launch()
 
-demo = gr.Interface(
-    fn=getanswer,
-    inputs=["text"],
-    outputs=["text", "text"],
-)
+# demo = gr.Interface(
+#     fn=getanswer,
+#     inputs=["text"],
+#     outputs=["text", "text"],
+#     api="question"
+# )
+# demo.launch()
+
+with gr.Blocks() as demo:
+    chatbot = gr.Chatbot()
+    msg = gr.Textbox()
+    clear = gr.Button("Clear")
+
+    def getanswer(question, chat_history):
+        llm_response = qa_chain(question)
+        response = process_llm_response(llm_response)
+        sources= print_array(response["sources"])
+        processed_response = response["result"] +" \n\nSources: " + sources
+        chat_history.append((question, processed_response))
+        return "", chat_history
+
+    msg.submit(getanswer, [msg, chatbot], [msg, chatbot])
+    clear.click(lambda: None, None, chatbot, queue=False)
+
 demo.launch()