add topic, qa answer

app.py

@@ -14,9 +14,10 @@ from langchain.prompts.chat import (
 )
 
 # Streaming endpoint
-API_URL = "https://api.openai.com/v1/chat/completions"  # os.getenv("API_URL") + "/generate_stream"
+API_URL = "https://api.openai.com/v1/chat/completions"
 cohere_key = '5IRbILAbjTI0VcqTsktBfKsr13Lych9iBAFbLpkj'
-faiss_store = './indexer'
+faiss_store = './indexer/{}'
+
 
 def gen_conversation(conversations):
     messages = []
@@ -32,22 +33,28 @@ def gen_conversation(conversations):
     return messages
 
 
-def predict(inputs, top_p, temperature, openai_api_key, enable_index, max_tokens, model,
+def predict(inputs, top_p, temperature, openai_api_key, enable_index, max_tokens, model, topic,
             chat_counter, chatbot=[], history=[]):
     model = model[0]
+    topic = topic[0]
     headers = {
         "Content-Type": "application/json",
         "Authorization": f"Bearer {openai_api_key}"
     }
 
     print(f"chat_counter - {chat_counter}")
-    #Debugging
+    print(f'Histroy - {history}')  # History: Original Input and Output in flatten list
+    print(f'chatbot - {chatbot}')  # Chat Bot： 上一轮回复的[[user, AI]]
+
+    history.append(inputs)
+    # Debugging
     if enable_index:
         # Faiss 检索最近的embedding
-        if model =='openai':
-            docsearch = FAISS.load_local(faiss_store, OpenAIEmbeddings(openai_api_key=openai_api_key))
+        store = faiss_store.format(topic)
+        if model == 'openai':
+            docsearch = FAISS.load_local(store, OpenAIEmbeddings(openai_api_key=openai_api_key))
         else:
-            docsearch = FAISS.load_local(faiss_store, CohereEmbeddings(cohere_api_key=cohere_key ))
+            docsearch = FAISS.load_local(store, CohereEmbeddings(cohere_api_key=cohere_key))
         # 构建模板
         llm = ChatOpenAI(openai_api_key=openai_api_key, max_tokens=max_tokens)
         messages_combine = [
@@ -67,66 +74,62 @@ def predict(inputs, top_p, temperature, openai_api_key, enable_index, max_tokens
                                            )
         result = chain({"query": inputs})
         print(result)
-
-    if chat_counter == 0:
-        messages = [{"role": "user", "content": f"{inputs}"}]
-
+        result = result['result']
+        # 生成返回值
+        history.append(result)
+        chat = [(history[i], history[i + 1]) for i in range(0, len(history) - 1, 2)]
+        chat_counter += 1
+        yield chat, history, chat_counter
     else:
-        # 如果有历史对话，把对话拼接进入上下文
-        messages = gen_conversation(chatbot)
-
-        temp3 = {}
-        temp3["role"] = "user"
-        temp3["content"] = inputs
-        messages.append(temp3)
-
-    # messages
-    payload = {
-        "model": "gpt-3.5-turbo",
-        "messages": messages,  # [{"role": "user", "content": f"{inputs}"}],
-        "temperature": temperature,  # 1.0,
-        "top_p": top_p,  # 1.0,
-        "n": 1,
-        "stream": True,
-        "presence_penalty": 0,
-        "frequency_penalty": 0,
-    }
-
-    chat_counter += 1
-
-    # History: Original Input and Output
-    history.append(inputs)
-    print(f"payload is - {payload}")
-    #上一轮回复的[[user, AI]]
-    print(f'chatbot - {chatbot}')
-    print(f'Histroy - {history}')
-    # 请求OpenAI
-    response = requests.post(API_URL, headers=headers, json=payload, stream=True)
-    token_counter = 0
-    partial_words = ""
-
-    # 逐字返回
-    counter = 0
-    for chunk in response.iter_lines():
-        if counter == 0:
+        if chat_counter == 0:
+            messages = [{"role": "user", "content": f"{inputs}"}]
+        else:
+            # 如果有历史对话，把对话拼接进入上下文
+            messages = gen_conversation(chatbot)
+            messages.append({'role': 'user', 'content': inputs})
+        # messages
+        payload = {
+            "model": "gpt-3.5-turbo",
+            "messages": messages,  # [{"role": "user", "content": f"{inputs}"}],
+            "temperature": temperature,  # 1.0,
+            "top_p": top_p,  # 1.0,
+            "n": 1,
+            "stream": True,
+            "presence_penalty": 0,
+            "frequency_penalty": 0,
+        }
+        print(f"payload is - {payload}")
+
+        chat_counter += 1
+
+        # 请求OpenAI
+        response = requests.post(API_URL, headers=headers, json=payload, stream=True)
+        token_counter = 0
+        partial_words = ""
+
+        # 逐字返回
+        counter = 0
+        for chunk in response.iter_lines():
+            if counter == 0:
+                counter += 1
+                continue
             counter += 1
-            continue
-        counter += 1
-        # check whether each line is non-empty
-        if chunk:
-            # decode each line as response data is in bytes
-            delta = json.loads(chunk.decode()[6:])['choices'][0]["delta"]
-            if len(delta) == 0:
-                break
-            partial_words += delta["content"]
-            # Keep Updating history
-            if token_counter == 0:
-                history.append(" " + partial_words)
-            else:
-                history[-1] = partial_words
-            chat = [(history[i], history[i + 1]) for i in range(0, len(history) - 1, 2)]  # convert to tuples of list
-            token_counter += 1
-            yield chat, history, chat_counter
+            # check whether each line is non-empty
+            if chunk:
+                # decode each line as response data is in bytes
+                delta = json.loads(chunk.decode()[6:])['choices'][0]["delta"]
+                if len(delta) == 0:
+                    break
+                partial_words += delta["content"]
+                # Keep Updating history
+                if token_counter == 0:
+                    history.append(" " + partial_words)
+                else:
+                    history[-1] = partial_words
+                chat = [(history[i], history[i + 1]) for i in
+                        range(0, len(history) - 1, 2)]  # convert to tuples of list
+                token_counter += 1
+                yield chat, history, chat_counter
 
 
 def reset_textbox():
@@ -138,30 +141,40 @@ with gr.Blocks(css="""#col_container {width: 1000px; margin-left: auto; margin-r
     gr.HTML("""<h1 align="center">🚀Finance ChatBot🚀</h1>""")
     with gr.Column(elem_id="col_container"):
         openai_api_key = gr.Textbox(type='password', label="输入OPEN API Key")
+
+        # inputs, top_p, temperature, top_k, repetition_penalty
+        with gr.Accordion("Parameters", open=True):
+            with gr.Row():
+                top_p = gr.Slider(minimum=-0, maximum=1.0, value=0.9, step=0.05, interactive=True,
+                                  label="Top-p (nucleus sampling)", )
+
+                temperature = gr.Slider(minimum=-0, maximum=5.0, value=0.8, step=0.1, interactive=True,
+                                        label="Temperature", )
+            with gr.Row():
+                model = gr.CheckboxGroup(["cohere", "openai", "mpnet"])
+                max_tokens = gr.Slider(minimum=100, maximum=2000, value=200, step=100, interactive=True,
+                                       label="Max Tokens", )
+                chat_counter = gr.Number(value=0, precision=0, label='对话轮数')
+            with gr.Row():
+                enable_index = gr.Checkbox(label='是', info='开启基于文档问答模式/关闭为聊天模式')
+                enable_search = gr.Checkbox(label='是', info='是否使用搜索结果')
+                topic = gr.CheckboxGroup(["两会", "数字经济", "硅谷银行"])
+
         chatbot = gr.Chatbot(elem_id='chatbot')
         inputs = gr.Textbox(placeholder="您有什么问题可以问我", label="输入数字经济，两会，硅谷银行相关的提问")
         state = gr.State([])
 
-        clear = gr.Button("Clear Conversation")
-        run = gr.Button("Run")
+        with gr.Row():
+            clear = gr.Button("Clear Conversation")
+            run = gr.Button("Run")
 
-        # inputs, top_p, temperature, top_k, repetition_penalty
-        with gr.Accordion("Parameters", open=True):
-            top_p = gr.Slider(minimum=-0, maximum=1.0, value=1.0, step=0.05, interactive=True,
-                              label="Top-p (nucleus sampling)", )
-            max_tokens = gr.Slider(minimum=512, maximum=3000, value=3000, step=100, interactive=True,
-                              label="Max Tokens", )
-            temperature = gr.Slider(minimum=-0, maximum=5.0, value=1.0, step=0.1, interactive=True,
-                                    label="Temperature", )
-            model = gr.CheckboxGroup(["cohere", "openai", "mpnet"])
-            chat_counter = gr.Number(value=0, precision=0)
-            enable_index = gr.Checkbox(label='是', info='是否使用研报等金融数据')
-            # 后续考虑加入搜索结果
-            enable_search = gr.Checkbox(label='是', info='是否使用搜索结果')
-
-    inputs.submit(predict, [inputs, top_p, temperature, openai_api_key, enable_index, max_tokens, model, chat_counter, chatbot, state],
+    inputs.submit(predict,
+                  [inputs, top_p, temperature, openai_api_key, enable_index, max_tokens, model, topic, chat_counter, chatbot,
+                   state],
                   [chatbot, state, chat_counter], )
-    run.click(predict, [inputs, top_p, temperature, openai_api_key, enable_index, max_tokens, model, chat_counter, chatbot, state],
+    run.click(predict,
+              [inputs, top_p, temperature, openai_api_key, enable_index, max_tokens, model, topic, chat_counter, chatbot,
+               state],
               [chatbot, state, chat_counter], )
 
     # 每次对话结束都重置对话

prompts/chat_combine_prompt.txt

@@ -1,4 +1,4 @@
-You are a DocsGPT, friendly and helpful AI assistant by Arc53 that provides help with documents. You give thorough answers with code examples if possible.
-Use the following pieces of context to help answer the users question.
+You are a DocsGPT, friendly and helpful AI assistant by TianHong Asset Managementthat provides help with documents and financial news. You give thorough answers with detail number and illustrated examples if possible.
+Use the following pieces of context to help answer the users question, always answer in chinese.
 ----------------
 {summaries}