Add function call feature, implement autonomous judgment on whether t…

…o search using function call, restructure code directory, add search command to enforce search.

.github/fly_deploy.yml

@@ -22,4 +22,4 @@ jobs:
         uses: actions/checkout@v2
       - name: Deploy
         run: |
-          sh ./deploy.sh
+          sh ./.github/deploy.sh

bot.py

@@ -3,17 +3,18 @@
 import config
 import logging
 import traceback
-import decorators
-from md2tgmd import escape
-from runasync import run_async
+import utils.decorators as decorators
+from utils.md2tgmd import escape
+from utils.runasync import run_async
 from chatgpt2api.chatgpt2api import Chatbot as GPT
 from telegram.constants import ChatAction
-from agent import docQA, get_doc_from_local
+from utils.agent import docQA, get_doc_from_local
 from telegram import BotCommand, InlineKeyboardButton, InlineKeyboardMarkup
 from telegram.ext import CommandHandler, MessageHandler, ApplicationBuilder, filters, CallbackQueryHandler
 
 
 logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(name)s - %(levelname)s - %(message)s")
+# logging.basicConfig(level=logging.DEBUG, format="%(asctime)s - %(name)s - %(levelname)s - %(message)s")
 logger = logging.getLogger()
 
 # 获取 httpx 的 logger
@@ -97,12 +98,65 @@ async def getChatGPT(update, context, title, robot, message, use_search=config.S
         reply_to_message_id=update.message.message_id,
     )
     messageid = message.message_id
-    if use_search and not has_command:
-        get_answer = robot.search_summary
-    else:
-        get_answer = robot.ask_stream
+    get_answer = robot.ask_stream
+    # if use_search and not has_command:
+    #     get_answer = robot.search_summary
+    # else:
+    #     get_answer = robot.ask_stream
+    if not config.API or (config.USE_G4F and not config.SEARCH_USE_GPT):
+        import utils.gpt4free as gpt4free
+        get_answer = gpt4free.get_response
+
+    try:
+        for data in get_answer(text, convo_id=str(update.message.chat_id), pass_history=config.PASS_HISTORY):
+            result = result + data
+            tmpresult = result
+            modifytime = modifytime + 1
+            if re.sub(r"```", '', result).count("`") % 2 != 0:
+                tmpresult = result + "`"
+            if result.count("```") % 2 != 0:
+                tmpresult = result + "\n```"
+            if modifytime % 20 == 0 and lastresult != tmpresult:
+                if 'claude2' in title:
+                    tmpresult = re.sub(r",", '，', tmpresult)
+                await context.bot.edit_message_text(chat_id=update.message.chat_id, message_id=messageid, text=escape(tmpresult), parse_mode='MarkdownV2', disable_web_page_preview=True)
+                lastresult = tmpresult
+    except Exception as e:
+        print('\033[31m')
+        print("response_msg", result)
+        print("error", e)
+        traceback.print_exc()
+        print('\033[0m')
+        if config.API:
+            robot.reset(convo_id=str(update.message.chat_id), system_prompt=config.systemprompt)
+        if "You exceeded your current quota, please check your plan and billing details." in str(e):
+            print("OpenAI api 已过期！")
+            await context.bot.delete_message(chat_id=update.message.chat_id, message_id=messageid)
+            messageid = ''
+            config.API = ''
+        result += f"`出错啦！{e}`"
+    print(result)
+    if lastresult != result and messageid:
+        if 'claude2' in title:
+            result = re.sub(r",", '，', result)
+        await context.bot.edit_message_text(chat_id=update.message.chat_id, message_id=messageid, text=escape(result), parse_mode='MarkdownV2', disable_web_page_preview=True)
+
+async def search(update, context, title, robot):
+    message = update.message.text if config.NICK is None else update.message.text[botNicKLength:].strip() if update.message.text[:botNicKLength].lower() == botNick else None
+    result = title
+    text = message
+    modifytime = 0
+    lastresult = ''
+    message = await context.bot.send_message(
+        chat_id=update.message.chat_id,
+        text="搜索中💭",
+        parse_mode='MarkdownV2',
+        reply_to_message_id=update.message.message_id,
+    )
+    messageid = message.message_id
+    get_answer = robot.search_summary
     if not config.API or (config.USE_G4F and not config.SEARCH_USE_GPT):
-        import gpt4free
+        import utils.gpt4free as gpt4free
         get_answer = gpt4free.get_response
 
     try:
@@ -164,16 +218,16 @@ async def delete_message(update, context, messageid, delay=10):
     # [
     #     InlineKeyboardButton("gpt-3.5-turbo-0613", callback_data="gpt-3.5-turbo-0613"),
     # ],
-    [
-        InlineKeyboardButton("gpt-4", callback_data="gpt-4"),
-        InlineKeyboardButton("gpt-4-32k", callback_data="gpt-4-32k"),
-        # InlineKeyboardButton("gpt-4-0314", callback_data="gpt-4-0314"),
-    ],
     [
         InlineKeyboardButton("gpt-4-1106-preview", callback_data="gpt-4-1106-preview"),
         # InlineKeyboardButton("gpt-4-32k", callback_data="gpt-4-32k"),
         # InlineKeyboardButton("gpt-4-32k-0314", callback_data="gpt-4-32k-0314"),
     ],
+    [
+        InlineKeyboardButton("gpt-4", callback_data="gpt-4"),
+        InlineKeyboardButton("gpt-4-32k", callback_data="gpt-4-32k"),
+        # InlineKeyboardButton("gpt-4-0314", callback_data="gpt-4-0314"),
+    ],
     # [
     #     InlineKeyboardButton("gpt-4-0613", callback_data="gpt-4-0613"),
     #     InlineKeyboardButton("gpt-4-32k-0613", callback_data="gpt-4-32k-0613"),
@@ -373,7 +427,7 @@ async def info(update, context):
     messageid = message.message_id
     await context.bot.delete_message(chat_id=update.effective_chat.id, message_id=update.message.message_id)
 
-from agent import pdfQA, getmd5, persist_emdedding_pdf
+from utils.agent import pdfQA, getmd5, persist_emdedding_pdf
 @decorators.Authorization
 async def handle_pdf(update, context):
     # 获取接收到的文件
@@ -456,14 +510,16 @@ def setup(token):
 
     run_async(application.bot.set_my_commands([
         BotCommand('info', 'basic information'),
-        BotCommand('qa', 'Document Q&A with Embedding Database Search'),
+        BotCommand('search', 'search Google or duckduckgo'),
         BotCommand('en2zh', 'translate to Chinese'),
         BotCommand('zh2en', 'translate to English'),
+        BotCommand('qa', 'Document Q&A with Embedding Database Search'),
         BotCommand('start', 'Start the bot'),
         BotCommand('reset', 'Reset the bot'),
     ]))
 
     application.add_handler(CommandHandler("start", start))
+    application.add_handler(CommandHandler("search", lambda update, context: search(update, context, title=f"`🤖️ {config.GPT_ENGINE}`\n\n", robot=config.ChatGPTbot)))
     application.add_handler(CallbackQueryHandler(button_press))
     application.add_handler(CommandHandler("reset", reset_chat))
     application.add_handler(CommandHandler("en2zh", lambda update, context: command_bot(update, context, "simplified chinese", robot=config.ChatGPTbot)))

chatgpt2api/chatgpt2api.py

@@ -13,7 +13,8 @@
 import config
 import threading
 import time as record_time
-from agent import ThreadWithReturnValue, Web_crawler, pdf_search, getddgsearchurl, getgooglesearchurl, gptsearch, ChainStreamHandler, ChatOpenAI, CallbackManager, PromptTemplate, LLMChain, EducationalLLM
+from utils.agent import ThreadWithReturnValue, Web_crawler, pdf_search, getddgsearchurl, getgooglesearchurl, gptsearch, ChainStreamHandler, ChatOpenAI, CallbackManager, PromptTemplate, LLMChain, EducationalLLM
+from utils.function_call import function_call_list
 
 def get_filtered_keys_from_object(obj: object, *keys: str) -> Set[str]:
     """
@@ -240,10 +241,8 @@ def ask_stream(
                 or "https://api.openai.com/v1/chat/completions"
             )
             headers = {"Authorization": f"Bearer {kwargs.get('api_key', self.api_key)}"}
-        response = self.session.post(
-            url,
-            headers=headers,
-            json={
+
+        json_post = {
                 "model": os.environ.get("MODEL_NAME") or model or self.engine,
                 "messages": self.conversation[convo_id] if pass_history else [{"role": "system","content": self.system_prompt},{"role": role, "content": prompt}],
                 "stream": True,
@@ -264,7 +263,13 @@ def ask_stream(
                     self.get_max_tokens(convo_id=convo_id),
                     kwargs.get("max_tokens", self.max_tokens),
                 ),
-            },
+        }
+        if config.SEARCH_USE_GPT:
+            json_post.update(function_call_list["web_search"])
+        response = self.session.post(
+            url,
+            headers=headers,
+            json=json_post,
             timeout=kwargs.get("timeout", self.timeout),
             stream=True,
         )
@@ -274,6 +279,7 @@ def ask_stream(
             )
         response_role: str or None = None
         full_response: str = ""
+        need_function_call = False
         for line in response.iter_lines():
             if not line:
                 continue
@@ -290,11 +296,20 @@ def ask_stream(
                 continue
             if "role" in delta:
                 response_role = delta["role"]
-            if "content" in delta:
+            if "content" in delta and delta["content"]:
+                need_function_call = False
                 content = delta["content"]
                 full_response += content
                 yield content
-        self.add_to_conversation(full_response, response_role, convo_id=convo_id)
+            if "function_call" in delta and config.SEARCH_USE_GPT:
+                need_function_call = True
+                function_call_content = delta["function_call"]["arguments"]
+                full_response += function_call_content
+        if need_function_call:
+            keywords = json.loads(full_response)["prompt"]
+            yield from self.search_summary(keywords, convo_id=convo_id, need_function_call=True)
+        else:
+            self.add_to_conversation(full_response, response_role, convo_id=convo_id)
 
     async def ask_stream_async(
         self,
@@ -426,13 +441,15 @@ def search_summary(
             convo_id: str = "default",
             model: str = None,
             pass_history: bool = True,
+            need_function_call: bool = False,
             **kwargs,
         ):
 
         if convo_id not in self.conversation:
             self.reset(convo_id=convo_id, system_prompt=self.system_prompt)
-        self.add_to_conversation(prompt, "user", convo_id=convo_id)
-        self.__truncate_conversation(convo_id=convo_id)
+        if need_function_call == False:
+            self.add_to_conversation(prompt, "user", convo_id=convo_id)
+            self.__truncate_conversation(convo_id=convo_id)
 
         start_time = record_time.time()
 

main.py

@@ -3,7 +3,7 @@
 from bot import setup
 from urllib import parse
 from waitress import serve
-from runasync import run_async
+from utils.runasync import run_async
 from flask import Flask, request, jsonify
 from config import BOT_TOKEN, WEB_HOOK, PORT
 

test/test.py

@@ -1,22 +1,26 @@
-my_list = [
-    {"role": "admin", "content": "This is admin content."},
-    {"role": "user", "content": "This is user content."}
-]
+# my_list = [
+#     {"role": "admin", "content": "This is admin content."},
+#     {"role": "user", "content": "This is user content."}
+# ]
+a = {"role": "admin"}
+b = {"content": "This is user content."}
+a.update(b)
+print(a)
 
-content_list = [item["content"] for item in my_list]
-print(content_list)
+# content_list = [item["content"] for item in my_list]
+# print(content_list)
 
-engine = "gpt-3.5-turbo-1106"
-truncate_limit = (
-    30500
-    if "gpt-4-32k" in engine
-    else 6500
-    if "gpt-4" in engine
-    else 14500
-    if "gpt-3.5-turbo-16k" in engine or "gpt-3.5-turbo-1106" in engine
-    else 98500
-    if ("claude-2-web" or "claude-2") in engine
-    else 3400
-)
+# engine = "gpt-3.5-turbo-1106"
+# truncate_limit = (
+#     30500
+#     if "gpt-4-32k" in engine
+#     else 6500
+#     if "gpt-4" in engine
+#     else 14500
+#     if "gpt-3.5-turbo-16k" in engine or "gpt-3.5-turbo-1106" in engine
+#     else 98500
+#     if ("claude-2-web" or "claude-2") in engine
+#     else 3400
+# )
 
-print(truncate_limit)
+# print(truncate_limit)

test/test_Faucet.py

@@ -6,6 +6,8 @@ def gptsearch(result, llm):
     response = response.content
     return response
 
-chainllm = ChatOpenAI(temperature=0.5, openai_api_base="https://openkey.cloud/v1", model_name="gpt-4", openai_api_key="sk-ucUnnmqI9DdtsAXG8OKxOFxD5dnSrU3E3ZQh4PJa1dgQ7KzE")
+chainllm = ChatOpenAI(temperature=0.5, openai_api_base="https://openkey.cloud/v1", model_name="gpt-3.5-turbo", openai_api_key="sk-ucUnnmqI9DdtsAXG8OKxOFxD5dnSrU3E3ZQh4PJa1dgQ7KzE")
+# chainllm = ChatOpenAI(temperature=0.5, openai_api_base="https://openkey.cloud/v1", model_name="gpt-4-1106-preview", openai_api_key="sk-ucUnnmqI9DdtsAXG8OKxOFxD5dnSrU3E3ZQh4PJa1dgQ7KzE")
+# chainllm = ChatOpenAI(temperature=0.5, openai_api_base="https://openkey.cloud/v1", model_name="gpt-4", openai_api_key="sk-ucUnnmqI9DdtsAXG8OKxOFxD5dnSrU3E3ZQh4PJa1dgQ7KzE")
 
 print(gptsearch("鲁迅和周树人为什么打架", chainllm))

test/test_keyword.py

@@ -4,7 +4,7 @@
 from langchain.chains import LLMChain
 from langchain.prompts import PromptTemplate
 from langchain.chat_models import ChatOpenAI
-from googlesearch import GoogleSearchAPIWrapper
+from utils.googlesearch import GoogleSearchAPIWrapper
 
 
 def getgooglesearchurl(result, numresults=3):

agent.py → utils/agent.py

@@ -34,7 +34,7 @@
 from langchain.text_splitter import CharacterTextSplitter
 from langchain.tools import DuckDuckGoSearchRun, DuckDuckGoSearchResults, Tool
 from langchain.utilities import WikipediaAPIWrapper
-from googlesearch import GoogleSearchAPIWrapper
+from utils.googlesearch import GoogleSearchAPIWrapper
 from langchain.document_loaders import UnstructuredPDFLoader
 
 def getmd5(string):
@@ -44,7 +44,7 @@ def getmd5(string):
     md5_hex = md5_hash.hexdigest()
     return md5_hex
 
-from sitemap import SitemapLoader
+from utils.sitemap import SitemapLoader
 async def get_doc_from_sitemap(url):
     # https://www.langchain.asia/modules/indexes/document_loaders/examples/sitemap#%E8%BF%87%E6%BB%A4%E7%AB%99%E7%82%B9%E5%9C%B0%E5%9B%BE-url-
     sitemap_loader = SitemapLoader(web_path=url)

utils/function_call.py

@@ -0,0 +1,64 @@
+function_call_list = {
+  "current_weather": {
+      "functions": [
+          {
+            "name": "get_current_weather",
+            "description": "Get the current weather in a given location",
+            "parameters": {
+              "type": "object",
+              "properties": {
+                "location": {
+                  "type": "string",
+                  "description": "The city and state, e.g. San Francisco, CA"
+                },
+                "unit": {
+                  "type": "string",
+                  "enum": ["celsius", "fahrenheit"]
+                }
+              },
+              "required": ["location"]
+            }
+          }
+        ],
+        "function_call": "auto"
+  },
+  "web_search": {
+      "functions": [
+          {
+            "name": "get_web_search_results",
+            "description": "Search Google to enhance knowledge.",
+            "parameters": {
+              "type": "object",
+              "properties": {
+                "prompt": {
+                  "type": "string",
+                  "description": "The prompt to search."
+                }
+              },
+              "required": ["prompt"]
+            }
+          }
+        ],
+        "function_call": "auto"
+  },
+  # "web_search": {
+  #     "functions": [
+  #         {
+  #           "name": "get_web_search_results",
+  #           "description": "Get the web page search results in a given keywords",
+  #           "parameters": {
+  #             "type": "object",
+  #             "properties": {
+  #               "keywords": {
+  #                 "type": "string",
+  #                 "description": "keywords that can yield better search results, keywords are connected with spaces, e.g. 1. The keywords of the sentence (How much does the zeabur software service cost per month?) is (zeabur price). 2. The keywords of the sentence (今天的微博热搜有哪些？) is (微博 热搜)"
+  #               }
+  #             },
+  #             "required": ["keywords"]
+  #           }
+  #         }
+  #       ],
+  #       "function_call": "auto"
+  # },
+}
+