去除UI微信化

app.py

@@ -1,14 +1,17 @@
 # -*- coding:utf-8 -*-
-import gradio as gr
 import os
 import logging
 import sys
-import argparse
+
+import gradio as gr
+
 from utils import *
 from presets import *
+from overwrites import *
+from chat_func import *
 
 logging.basicConfig(
-    level=logging.INFO,
+    level=logging.DEBUG,
     format="%(asctime)s [%(levelname)s] [%(filename)s:%(lineno)d] %(message)s",
 )
 
@@ -49,72 +52,13 @@ else:
                 authflag = True
 
 gr.Chatbot.postprocess = postprocess
+PromptHelper.compact_text_chunks = compact_text_chunks
 
 with open("custom.css", "r", encoding="utf-8") as f:
     customCSS = f.read()
 
 with gr.Blocks(
     css=customCSS,
-    theme=gr.themes.Soft(
-        primary_hue=gr.themes.Color(
-            c50="#02C160",
-            c100="rgba(2, 193, 96, 0.2)",
-            c200="#02C160",
-            c300="rgba(2, 193, 96, 0.32)",
-            c400="rgba(2, 193, 96, 0.32)",
-            c500="rgba(2, 193, 96, 1.0)",
-            c600="rgba(2, 193, 96, 1.0)",
-            c700="rgba(2, 193, 96, 0.32)",
-            c800="rgba(2, 193, 96, 0.32)",
-            c900="#02C160",
-            c950="#02C160",
-        ),
-        secondary_hue=gr.themes.Color(
-            c50="#576b95",
-            c100="#576b95",
-            c200="#576b95",
-            c300="#576b95",
-            c400="#576b95",
-            c500="#576b95",
-            c600="#576b95",
-            c700="#576b95",
-            c800="#576b95",
-            c900="#576b95",
-            c950="#576b95",
-        ),
-        neutral_hue=gr.themes.Color(
-            name="gray",
-            c50="#f9fafb",
-            c100="#f3f4f6",
-            c200="#e5e7eb",
-            c300="#d1d5db",
-            c400="#B2B2B2",
-            c500="#808080",
-            c600="#636363",
-            c700="#515151",
-            c800="#393939",
-            c900="#272727",
-            c950="#171717",
-        ),
-        radius_size=gr.themes.sizes.radius_sm,
-    ).set(
-        button_primary_background_fill="#06AE56",
-        button_primary_background_fill_dark="#06AE56",
-        button_primary_background_fill_hover="#07C863",
-        button_primary_border_color="#06AE56",
-        button_primary_border_color_dark="#06AE56",
-        button_primary_text_color="#FFFFFF",
-        button_primary_text_color_dark="#FFFFFF",
-        button_secondary_background_fill="#F2F2F2",
-        button_secondary_background_fill_dark="#2B2B2B",
-        button_secondary_text_color="#393939",
-        button_secondary_text_color_dark="#FFFFFF",
-        # background_fill_primary="#F7F7F7",
-        # background_fill_primary_dark="#1F1F1F",
-        block_title_text_color="*primary_500",
-        block_title_background_fill = "*primary_100",
-        input_background_fill="#F6F6F6",
-    ),
 ) as demo:
     history = gr.State([])
     token_count = gr.State([])
@@ -156,7 +100,7 @@ with gr.Blocks(
                         value=hide_middle_chars(my_api_key),
                         type="password",
                         visible=not HIDE_MY_KEY,
-                        label="API-Key",
+                        label="API-Key(按Enter提交)",
                     )
                     model_select_dropdown = gr.Dropdown(
                         label="选择模型", choices=MODELS, multiselect=False, value=MODELS[0]
@@ -165,7 +109,7 @@ with gr.Blocks(
                         label="实时传输回答", value=True, visible=enable_streaming_option
                     )
                     use_websearch_checkbox = gr.Checkbox(label="使用在线搜索", value=False)
-                    index_files = gr.File(label="上传索引文件", type="file", multiple=True)
+                    index_files = gr.Files(label="上传索引文件", type="file", multiple=True)
 
                 with gr.Tab(label="Prompt"):
                     systemPromptTxt = gr.Textbox(
@@ -286,6 +230,7 @@ with gr.Blocks(
             use_streaming_checkbox,
             model_select_dropdown,
             use_websearch_checkbox,
+            index_files
         ],
         [chatbot, history, status_display, token_count],
         show_progress=True,
@@ -306,6 +251,7 @@ with gr.Blocks(
             use_streaming_checkbox,
             model_select_dropdown,
             use_websearch_checkbox,
+            index_files
         ],
         [chatbot, history, status_display, token_count],
         show_progress=True,

presets.py

@@ -1,4 +1,23 @@
 # -*- coding:utf-8 -*-
+# 错误信息
+standard_error_msg = "☹️发生了错误："  # 错误信息的标准前缀
+error_retrieve_prompt = "请检查网络连接，或者API-Key是否有效。"  # 获取对话时发生错误
+connection_timeout_prompt = "连接超时，无法获取对话。"  # 连接超时
+read_timeout_prompt = "读取超时，无法获取对话。"  # 读取超时
+proxy_error_prompt = "代理错误，无法获取对话。"  # 代理错误
+ssl_error_prompt = "SSL错误，无法获取对话。"  # SSL 错误
+no_apikey_msg = "API key长度不是51位，请检查是否输入正确。"  # API key 长度不足 51 位
+
+max_token_streaming = 3500  # 流式对话时的最大 token 数
+timeout_streaming = 30  # 流式对话时的超时时间
+max_token_all = 3500  # 非流式对话时的最大 token 数
+timeout_all = 200  # 非流式对话时的超时时间
+enable_streaming_option = True  # 是否启用选择选择是否实时显示回答的勾选框
+HIDE_MY_KEY = False  # 如果你想在UI中隐藏你的 API 密钥，将此值设置为 True
+
+SIM_K = 5
+INDEX_QUERY_TEMPRATURE = 1.0
+
 title = """<h1 align="left" style="min-width:200px; margin-top:0;">川虎ChatGPT 🚀</h1>"""
 description = """\
 <div align="center" style="margin:16px 0">
@@ -12,6 +31,7 @@ description = """\
 """
 
 summarize_prompt = "你是谁？我们刚才聊了什么？"  # 总结对话时的 prompt
+
 MODELS = [
     "gpt-3.5-turbo",
     "gpt-3.5-turbo-0301",
@@ -21,7 +41,8 @@ MODELS = [
     "gpt-4-32k-0314",
 ]  # 可选的模型
 
-websearch_prompt = """\
+
+WEBSEARCH_PTOMPT_TEMPLATE = """\
 Web search results:
 
 {web_results}
@@ -31,18 +52,29 @@ Instructions: Using the provided web search results, write a comprehensive reply
 Query: {query}
 Reply in 中文"""
 
-# 错误信息
-standard_error_msg = "☹️发生了错误："  # 错误信息的标准前缀
-error_retrieve_prompt = "请检查网络连接，或者API-Key是否有效。"  # 获取对话时发生错误
-connection_timeout_prompt = "连接超时，无法获取对话。"  # 连接超时
-read_timeout_prompt = "读取超时，无法获取对话。"  # 读取超时
-proxy_error_prompt = "代理错误，无法获取对话。"  # 代理错误
-ssl_error_prompt = "SSL错误，无法获取对话。"  # SSL 错误
-no_apikey_msg = "API key长度不是51位，请检查是否输入正确。"  # API key 长度不足 51 位
+PROMPT_TEMPLATE = """\
+Context information is below.
+---------------------
+{context_str}
+---------------------
+Current date: {current_date}.
+Using the provided context information, write a comprehensive reply to the given query.
+Make sure to cite results using [number] notation after the reference.
+If the provided context information refer to multiple subjects with the same name, write separate answers for each subject.
+Use prior knowledge only if the given context didn't provide enough information.
+Answer the question: {query_str}
+Reply in 中文
+"""
 
-max_token_streaming = 3500  # 流式对话时的最大 token 数
-timeout_streaming = 30  # 流式对话时的超时时间
-max_token_all = 3500  # 非流式对话时的最大 token 数
-timeout_all = 200  # 非流式对话时的超时时间
-enable_streaming_option = True  # 是否启用选择选择是否实时显示回答的勾选框
-HIDE_MY_KEY = False  # 如果你想在UI中隐藏你的 API 密钥，将此值设置为 True
+REFINE_TEMPLATE = """\
+The original question is as follows: {query_str}
+We have provided an existing answer: {existing_answer}
+We have the opportunity to refine the existing answer
+(only if needed) with some more context below.
+------------
+{context_msg}
+------------
+Given the new context, refine the original answer to better
+Answer in the same language as the question, such as English, 中文, 日本語, Español, Français, or Deutsch.
+If the context isn't useful, return the original answer.
+"""

requirements.txt

@@ -6,4 +6,6 @@ socksio
 tqdm
 colorama
 duckduckgo_search
-Pygments
+Pygments
+llama_index
+langchain

utils.py

@@ -3,23 +3,16 @@ from __future__ import annotations
 from typing import TYPE_CHECKING, Any, Callable, Dict, List, Tuple, Type
 import logging
 import json
-import gradio as gr
-
-# import openai
 import os
-import traceback
-import requests
-
-# import markdown
+import datetime
+import hashlib
 import csv
-import mdtex2html
+
+import gradio as gr
 from pypinyin import lazy_pinyin
-from presets import *
 import tiktoken
-from tqdm import tqdm
-import colorama
-from duckduckgo_search import ddg
-import datetime
+
+from presets import *
 
 # logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(levelname)s] [%(filename)s:%(lineno)d] %(message)s")
 
@@ -37,27 +30,6 @@ HISTORY_DIR = "history"
 TEMPLATES_DIR = "templates"
 
 
-def postprocess(
-    self, y: List[Tuple[str | None, str | None]]
-) -> List[Tuple[str | None, str | None]]:
-    """
-    Parameters:
-        y: List of tuples representing the message and response pairs. Each message and response should be a string, which may be in Markdown format.
-    Returns:
-        List of tuples representing the message and response. Each message and response will be a string of HTML.
-    """
-    if y is None:
-        return []
-    for i, (message, response) in enumerate(y):
-        y[i] = (
-            # None if message is None else markdown.markdown(message),
-            # None if response is None else markdown.markdown(response),
-            None if message is None else message,
-            None if response is None else mdtex2html.convert(response, extensions=['fenced_code','codehilite','tables']),
-        )
-    return y
-
-
 def count_token(message):
     encoding = tiktoken.get_encoding("cl100k_base")
     input_str = f"role: {message['role']}, content: {message['content']}"
@@ -102,389 +74,6 @@ def construct_token_message(token, stream=False):
     return f"Token 计数: {token}"
 
 
-def get_response(
-    openai_api_key, system_prompt, history, temperature, top_p, stream, selected_model
-):
-    headers = {
-        "Content-Type": "application/json",
-        "Authorization": f"Bearer {openai_api_key}",
-    }
-
-    history = [construct_system(system_prompt), *history]
-
-    payload = {
-        "model": selected_model,
-        "messages": history,  # [{"role": "user", "content": f"{inputs}"}],
-        "temperature": temperature,  # 1.0,
-        "top_p": top_p,  # 1.0,
-        "n": 1,
-        "stream": stream,
-        "presence_penalty": 0,
-        "frequency_penalty": 0,
-    }
-    if stream:
-        timeout = timeout_streaming
-    else:
-        timeout = timeout_all
-
-    # 获取环境变量中的代理设置
-    http_proxy = os.environ.get("HTTP_PROXY") or os.environ.get("http_proxy")
-    https_proxy = os.environ.get("HTTPS_PROXY") or os.environ.get("https_proxy")
-
-    # 如果存在代理设置，使用它们
-    proxies = {}
-    if http_proxy:
-        logging.info(f"Using HTTP proxy: {http_proxy}")
-        proxies["http"] = http_proxy
-    if https_proxy:
-        logging.info(f"Using HTTPS proxy: {https_proxy}")
-        proxies["https"] = https_proxy
-
-    # 如果有代理，使用代理发送请求，否则使用默认设置发送请求
-    if proxies:
-        response = requests.post(
-            API_URL,
-            headers=headers,
-            json=payload,
-            stream=True,
-            timeout=timeout,
-            proxies=proxies,
-        )
-    else:
-        response = requests.post(
-            API_URL,
-            headers=headers,
-            json=payload,
-            stream=True,
-            timeout=timeout,
-        )
-    return response
-
-
-def stream_predict(
-    openai_api_key,
-    system_prompt,
-    history,
-    inputs,
-    chatbot,
-    all_token_counts,
-    top_p,
-    temperature,
-    selected_model,
-):
-    def get_return_value():
-        return chatbot, history, status_text, all_token_counts
-
-    logging.info("实时回答模式")
-    partial_words = ""
-    counter = 0
-    status_text = "开始实时传输回答……"
-    history.append(construct_user(inputs))
-    history.append(construct_assistant(""))
-    chatbot.append((parse_text(inputs), ""))
-    user_token_count = 0
-    if len(all_token_counts) == 0:
-        system_prompt_token_count = count_token(construct_system(system_prompt))
-        user_token_count = (
-            count_token(construct_user(inputs)) + system_prompt_token_count
-        )
-    else:
-        user_token_count = count_token(construct_user(inputs))
-    all_token_counts.append(user_token_count)
-    logging.info(f"输入token计数: {user_token_count}")
-    yield get_return_value()
-    try:
-        response = get_response(
-            openai_api_key,
-            system_prompt,
-            history,
-            temperature,
-            top_p,
-            True,
-            selected_model,
-        )
-    except requests.exceptions.ConnectTimeout:
-        status_text = (
-            standard_error_msg + connection_timeout_prompt + error_retrieve_prompt
-        )
-        yield get_return_value()
-        return
-    except requests.exceptions.ReadTimeout:
-        status_text = standard_error_msg + read_timeout_prompt + error_retrieve_prompt
-        yield get_return_value()
-        return
-
-    yield get_return_value()
-    error_json_str = ""
-
-    for chunk in tqdm(response.iter_lines()):
-        if counter == 0:
-            counter += 1
-            continue
-        counter += 1
-        # check whether each line is non-empty
-        if chunk:
-            chunk = chunk.decode()
-            chunklength = len(chunk)
-            try:
-                chunk = json.loads(chunk[6:])
-            except json.JSONDecodeError:
-                logging.info(chunk)
-                error_json_str += chunk
-                status_text = f"JSON解析错误。请重置对话。收到的内容: {error_json_str}"
-                yield get_return_value()
-                continue
-            # decode each line as response data is in bytes
-            if chunklength > 6 and "delta" in chunk["choices"][0]:
-                finish_reason = chunk["choices"][0]["finish_reason"]
-                status_text = construct_token_message(
-                    sum(all_token_counts), stream=True
-                )
-                if finish_reason == "stop":
-                    yield get_return_value()
-                    break
-                try:
-                    partial_words = (
-                        partial_words + chunk["choices"][0]["delta"]["content"]
-                    )
-                except KeyError:
-                    status_text = (
-                        standard_error_msg
-                        + "API回复中找不到内容。很可能是Token计数达到上限了。请重置对话。当前Token计数: "
-                        + str(sum(all_token_counts))
-                    )
-                    yield get_return_value()
-                    break
-                history[-1] = construct_assistant(partial_words)
-                chatbot[-1] = (parse_text(inputs), parse_text(partial_words))
-                all_token_counts[-1] += 1
-                yield get_return_value()
-
-
-def predict_all(
-    openai_api_key,
-    system_prompt,
-    history,
-    inputs,
-    chatbot,
-    all_token_counts,
-    top_p,
-    temperature,
-    selected_model,
-):
-    logging.info("一次性回答模式")
-    history.append(construct_user(inputs))
-    history.append(construct_assistant(""))
-    chatbot.append((parse_text(inputs), ""))
-    all_token_counts.append(count_token(construct_user(inputs)))
-    try:
-        response = get_response(
-            openai_api_key,
-            system_prompt,
-            history,
-            temperature,
-            top_p,
-            False,
-            selected_model,
-        )
-    except requests.exceptions.ConnectTimeout:
-        status_text = (
-            standard_error_msg + connection_timeout_prompt + error_retrieve_prompt
-        )
-        return chatbot, history, status_text, all_token_counts
-    except requests.exceptions.ProxyError:
-        status_text = standard_error_msg + proxy_error_prompt + error_retrieve_prompt
-        return chatbot, history, status_text, all_token_counts
-    except requests.exceptions.SSLError:
-        status_text = standard_error_msg + ssl_error_prompt + error_retrieve_prompt
-        return chatbot, history, status_text, all_token_counts
-    response = json.loads(response.text)
-    content = response["choices"][0]["message"]["content"]
-    history[-1] = construct_assistant(content)
-    chatbot[-1] = (parse_text(inputs), parse_text(content))
-    total_token_count = response["usage"]["total_tokens"]
-    all_token_counts[-1] = total_token_count - sum(all_token_counts)
-    status_text = construct_token_message(total_token_count)
-    return chatbot, history, status_text, all_token_counts
-
-
-def predict(
-    openai_api_key,
-    system_prompt,
-    history,
-    inputs,
-    chatbot,
-    all_token_counts,
-    top_p,
-    temperature,
-    stream=False,
-    selected_model=MODELS[0],
-    use_websearch_checkbox=False,
-    should_check_token_count=True,
-):  # repetition_penalty, top_k
-    logging.info("输入为：" + colorama.Fore.BLUE + f"{inputs}" + colorama.Style.RESET_ALL)
-    if use_websearch_checkbox:
-        results = ddg(inputs, max_results=3)
-        web_results = []
-        for idx, result in enumerate(results):
-            logging.info(f"搜索结果{idx + 1}：{result}")
-            web_results.append(f'[{idx+1}]"{result["body"]}"\nURL: {result["href"]}')
-        web_results = "\n\n".join(web_results)
-        today = datetime.datetime.today().strftime("%Y-%m-%d")
-        inputs = (
-            websearch_prompt.replace("{current_date}", today)
-            .replace("{query}", inputs)
-            .replace("{web_results}", web_results)
-        )
-    if len(openai_api_key) != 51:
-        status_text = standard_error_msg + no_apikey_msg
-        logging.info(status_text)
-        chatbot.append((parse_text(inputs), ""))
-        if len(history) == 0:
-            history.append(construct_user(inputs))
-            history.append("")
-            all_token_counts.append(0)
-        else:
-            history[-2] = construct_user(inputs)
-        yield chatbot, history, status_text, all_token_counts
-        return
-    if stream:
-        yield chatbot, history, "开始生成回答……", all_token_counts
-    if stream:
-        logging.info("使用流式传输")
-        iter = stream_predict(
-            openai_api_key,
-            system_prompt,
-            history,
-            inputs,
-            chatbot,
-            all_token_counts,
-            top_p,
-            temperature,
-            selected_model,
-        )
-        for chatbot, history, status_text, all_token_counts in iter:
-            yield chatbot, history, status_text, all_token_counts
-    else:
-        logging.info("不使用流式传输")
-        chatbot, history, status_text, all_token_counts = predict_all(
-            openai_api_key,
-            system_prompt,
-            history,
-            inputs,
-            chatbot,
-            all_token_counts,
-            top_p,
-            temperature,
-            selected_model,
-        )
-        yield chatbot, history, status_text, all_token_counts
-    logging.info(f"传输完毕。当前token计数为{all_token_counts}")
-    if len(history) > 1 and history[-1]["content"] != inputs:
-        logging.info(
-            "回答为："
-            + colorama.Fore.BLUE
-            + f"{history[-1]['content']}"
-            + colorama.Style.RESET_ALL
-        )
-    if stream:
-        max_token = max_token_streaming
-    else:
-        max_token = max_token_all
-    if sum(all_token_counts) > max_token and should_check_token_count:
-        status_text = f"精简token中{all_token_counts}/{max_token}"
-        logging.info(status_text)
-        yield chatbot, history, status_text, all_token_counts
-        iter = reduce_token_size(
-            openai_api_key,
-            system_prompt,
-            history,
-            chatbot,
-            all_token_counts,
-            top_p,
-            temperature,
-            stream=False,
-            selected_model=selected_model,
-            hidden=True,
-        )
-        for chatbot, history, status_text, all_token_counts in iter:
-            status_text = f"Token 达到上限，已自动降低Token计数至 {status_text}"
-            yield chatbot, history, status_text, all_token_counts
-
-
-def retry(
-    openai_api_key,
-    system_prompt,
-    history,
-    chatbot,
-    token_count,
-    top_p,
-    temperature,
-    stream=False,
-    selected_model=MODELS[0],
-):
-    logging.info("重试中……")
-    if len(history) == 0:
-        yield chatbot, history, f"{standard_error_msg}上下文是空的", token_count
-        return
-    history.pop()
-    inputs = history.pop()["content"]
-    token_count.pop()
-    iter = predict(
-        openai_api_key,
-        system_prompt,
-        history,
-        inputs,
-        chatbot,
-        token_count,
-        top_p,
-        temperature,
-        stream=stream,
-        selected_model=selected_model,
-    )
-    logging.info("重试完毕")
-    for x in iter:
-        yield x
-
-
-def reduce_token_size(
-    openai_api_key,
-    system_prompt,
-    history,
-    chatbot,
-    token_count,
-    top_p,
-    temperature,
-    stream=False,
-    selected_model=MODELS[0],
-    hidden=False,
-):
-    logging.info("开始减少token数量……")
-    iter = predict(
-        openai_api_key,
-        system_prompt,
-        history,
-        summarize_prompt,
-        chatbot,
-        token_count,
-        top_p,
-        temperature,
-        stream=stream,
-        selected_model=selected_model,
-        should_check_token_count=False,
-    )
-    logging.info(f"chatbot: {chatbot}")
-    for chatbot, history, status_text, previous_token_count in iter:
-        history = history[-2:]
-        token_count = previous_token_count[-1:]
-        if hidden:
-            chatbot.pop()
-        yield chatbot, history, construct_token_message(
-            sum(token_count), stream=stream
-        ), token_count
-    logging.info("减少token数量完毕")
-
-
 def delete_last_conversation(chatbot, history, previous_token_count):
     if len(chatbot) > 0 and standard_error_msg in chatbot[-1][1]:
         logging.info("由于包含报错信息，只删除chatbot记录")
@@ -643,6 +232,7 @@ def reset_state():
 def reset_textbox():
     return gr.update(value="")
 
+
 def reset_default():
     global API_URL
     API_URL = "https://api.openai.com/v1/chat/completions"
@@ -650,6 +240,7 @@ def reset_default():
     os.environ.pop("https_proxy", None)
     return gr.update(value=API_URL), gr.update(value=""), "API URL 和代理已重置"
 
+
 def change_api_url(url):
     global API_URL
     API_URL = url
@@ -657,22 +248,37 @@ def change_api_url(url):
     logging.info(msg)
     return msg
 
+
 def change_proxy(proxy):
     os.environ["HTTPS_PROXY"] = proxy
     msg = f"代理更改为了{proxy}"
     logging.info(msg)
     return msg
 
+
 def hide_middle_chars(s):
     if len(s) <= 8:
         return s
     else:
         head = s[:4]
         tail = s[-4:]
-        hidden = '*' * (len(s) - 8)
+        hidden = "*" * (len(s) - 8)
         return head + hidden + tail
 
+
 def submit_key(key):
+    key = key.strip()
     msg = f"API密钥更改为了{hide_middle_chars(key)}"
     logging.info(msg)
     return key, msg
+
+
+def sha1sum(filename):
+    sha1 = hashlib.sha1()
+    sha1.update(filename.encode("utf-8"))
+    return sha1.hexdigest()
+
+
+def replace_today(prompt):
+    today = datetime.datetime.today().strftime("%Y-%m-%d")
+    return prompt.replace("{current_date}", today)