qwen_server/assistant_server.py

import json
import os
from pathlib import Path

import add_qwen_libs  # NOQA
import gradio as gr
import jsonlines
from schema import GlobalConfig

from qwen_agent.actions import RetrievalQA
from qwen_agent.llm import get_chat_model
from qwen_agent.log import logger
from qwen_agent.memory import Memory

# Read config
with open(Path(__file__).resolve().parent / 'server_config.json', 'r') as f:
    server_config = json.load(f)
    server_config = GlobalConfig(**server_config)

llm = get_chat_model(model=server_config.server.llm,
                     api_key=server_config.server.api_key,
                     model_server=server_config.server.model_server)

mem = Memory(llm=llm, stream=False)

cache_file = os.path.join(server_config.path.cache_root, 'browse.jsonl')
cache_file_popup_url = os.path.join(server_config.path.cache_root, 'popup_url.jsonl')
access_token_file = os.path.join(server_config.path.cache_root, 'access_token.jsonl')

PAGE_URL = {}

with open(Path(__file__).resolve().parent / 'css/main.css', 'r') as f:
    css = f.read()
with open(Path(__file__).resolve().parent / 'js/main.js', 'r') as f:
    js = f.read()


def add_text(history, text):
    history = history + [(text, None)]
    return history, gr.update(value='', interactive=False)


def rm_text(history):
    if not history:
        gr.Warning('No input content!')
    elif not history[-1][1]:
        return history, gr.update(value='', interactive=False)
    else:
        history = history[:-1] + [(history[-1][0], None)]
        return history, gr.update(value='', interactive=False)


def add_file(history, file):
    history = history + [((file.name,), None)]
    return history


def set_page_url(access_token):
    lines = {access_token: []}
    assert os.path.exists(cache_file_popup_url)
    for line in jsonlines.open(cache_file_popup_url):
        _access_token = line['access_token']
        if _access_token not in lines:
            lines[_access_token] = []
        lines[_access_token].append(line)
    if access_token not in PAGE_URL:
        PAGE_URL[access_token] = []
    PAGE_URL[access_token].append(lines[access_token][-1]['url'])
    logger.info('The current access page is: ' + PAGE_URL[access_token][-1])


def initialize(request: gr.Request):
    access_token = request.query_params["access_token"] or request.session["access_token"]
    is_valid = False
    if access_token:
        if os.getenv("ACCESS_TOKEN") == access_token:
            is_valid = True
        else:
            for line in jsonlines.open(access_token_file):
                if line['access_token'] == access_token:
                    is_valid = True
                    break
    if not is_valid:
        gr.Info("The token is not valid, Please reset!")
        return
    set_page_url(access_token)
    return access_token


def bot(history, access_token):
    set_page_url(access_token)
    if not history:
        yield history
    else:
        now_page = None
        _ref = ''
        if not os.path.exists(cache_file):
            gr.Info("Please add this page to LLMBB's Reading List first!")
        else:
            for line in jsonlines.open(cache_file):
                if line["access_token"] == access_token and line['url'] == PAGE_URL[access_token][-1]:
                    now_page = line

            if not now_page:
                gr.Info(
                    "This page has not yet been added to the LLMBB's reading list!"
                )
            elif not now_page['raw']:
                gr.Info('Please reopen later, LLMBB is analyzing this page...')
            else:
                _ref_list = mem.get(
                    history[-1][0], [now_page],
                    max_token=server_config.server.max_ref_token)
                if _ref_list:
                    _ref = '\n'.join(
                        json.dumps(x, ensure_ascii=False) for x in _ref_list)
                else:
                    _ref = ''

        # TODO: considering history for retrieval qa
        agent = RetrievalQA(stream=True, llm=llm)
        history[-1][1] = ''
        response = agent.run(user_request=history[-1][0], ref_doc=_ref)

        for chunk in response:
            if chunk is not None:
                history[-1][1] += chunk
            yield history

        # save history
        if now_page:
            now_page['session'] = history
            lines = []
            for line in jsonlines.open(cache_file):
                if line["access_token"] == access_token and line['url'] != PAGE_URL[access_token][-1]:
                    lines.append(line)

            lines.append(now_page)
            with jsonlines.open(cache_file, mode='w') as writer:
                for new_line in lines:
                    writer.write(new_line)


def load_history_session(history, access_token):
    now_page = None
    if not os.path.exists(cache_file):
        gr.Info("Please add this page to LLMBB's Reading List first!")
        return []
    for line in jsonlines.open(cache_file):
        if line["access_token"] == access_token and line['url'] == PAGE_URL[access_token][-1]:
            now_page = line
    if not now_page:
        gr.Info("Please add this page to LLMBB's Reading List first!")
        return []
    if not now_page['raw']:
        gr.Info('Please wait, LLMBB is analyzing this page...')
        return []
    return now_page['session']


def clear_session(access_token):
    if not os.path.exists(cache_file):
        return None
    now_page = None
    lines = []
    for line in jsonlines.open(cache_file):
        if line["access_token"] == access_token and line['url'] == PAGE_URL[access_token][-1]:
            now_page = line
        else:
            lines.append(line)
    if not now_page:
        return None
    now_page['session'] = []
    lines.append(now_page)
    with jsonlines.open(cache_file, mode='w') as writer:
        for new_line in lines:
            writer.write(new_line)

    return None


with gr.Blocks(css=css, theme='soft') as demo:
    access_token = gr.State("")
    chatbot = gr.Chatbot([],
                         elem_id='chatbot',
                         height=480,
                         avatar_images=(None, (os.path.join(
                             Path(__file__).resolve().parent,
                             'img/logo.png'))))
    with gr.Row():
        with gr.Column(scale=7):
            txt = gr.Textbox(show_label=False,
                             placeholder='Chat with LLMBB...',
                             container=False)
        # with gr.Column(scale=0.06, min_width=0):
        #     smt_bt = gr.Button('⏎')
        with gr.Column(scale=1, min_width=0):
            clr_bt = gr.Button('🧹', elem_classes='bt_small_font')
        with gr.Column(scale=1, min_width=0):
            stop_bt = gr.Button('🚫', elem_classes='bt_small_font')
        with gr.Column(scale=1, min_width=0):
            re_bt = gr.Button('🔁', elem_classes='bt_small_font')

    txt_msg = txt.submit(add_text, [chatbot, txt], [chatbot, txt],
                         queue=False).then(bot, [chatbot, access_token], chatbot)
    txt_msg.then(lambda: gr.update(interactive=True), None, [txt], queue=False)

    # txt_msg_bt = smt_bt.click(add_text, [chatbot, txt], [chatbot, txt],
    #                           queue=False).then(bot, chatbot, chatbot)
    # txt_msg_bt.then(lambda: gr.update(interactive=True),
    #                 None, [txt],
    #                 queue=False)

    clr_bt.click(clear_session, [access_token], chatbot, queue=False)
    re_txt_msg = re_bt.click(rm_text, [chatbot], [chatbot, txt],
                             queue=False).then(bot, [chatbot, access_token], chatbot)
    re_txt_msg.then(lambda: gr.update(interactive=True),
                    None, [txt],
                    queue=False)

    stop_bt.click(None, None, None, cancels=[txt_msg, re_txt_msg], queue=False)

    demo.load(initialize, [], [access_token]).then(load_history_session, [chatbot, access_token], chatbot)
    demo.queue()

# demo.queue().launch(server_name=server_config.server.server_host, server_port=server_config.server.app_in_browser_port)