import gradio as gr
import os
import openai
from auto_backgrounds import generate_backgrounds, generate_draft
from utils.file_operations import hash_name, list_folders
from references_generator import generate_top_k_references
# todo:
# 6. get logs when the procedure is not completed. *
# 7. 自己的文件库； 更多的prompts
# 2. 实现别的功能
# 3. Check API Key GPT-4 Support.
# future:
# generation.log sometimes disappears (ignore this)
# 1. Check if there are any duplicated citations
# 2. Remove potential thebibliography and bibitem in .tex file
#######################################################################################################################
# Check if openai and cloud storage available
#######################################################################################################################
openai_key = os.getenv("OPENAI_API_KEY")
access_key_id = os.getenv('AWS_ACCESS_KEY_ID')
secret_access_key = os.getenv('AWS_SECRET_ACCESS_KEY')
GPT4_ENABLE = os.getenv("GPT4_ENABLE") # by default None.
if access_key_id is None or secret_access_key is None:
print("Access keys are not provided. Outputs cannot be saved to AWS Cloud Storage.\n")
IS_CACHE_AVAILABLE = False
else:
IS_CACHE_AVAILABLE = True
if openai_key is None:
print("OPENAI_API_KEY is not found in environment variables. The output may not be generated.\n")
IS_OPENAI_API_KEY_AVAILABLE = False
else:
openai.api_key = openai_key
try:
openai.Model.list()
IS_OPENAI_API_KEY_AVAILABLE = True
# except Exception as e:
except openai.error.AuthenticationError:
IS_OPENAI_API_KEY_AVAILABLE = False
DEFAULT_MODEL = "gpt-4" if GPT4_ENABLE else 'gpt-3.5-turbo-16k'
GPT4_INTERACTIVE = True if GPT4_ENABLE else False
DEFAULT_SECTIONS = ["introduction", "related works", "backgrounds", "methodology", "experiments",
"conclusion", "abstract"] if GPT4_ENABLE \
else ["introduction", "related works"]
MODEL_LIST = ['gpt-4', 'gpt-3.5-turbo', 'gpt-3.5-turbo-16k']
#######################################################################################################################
# Load the list of templates & knowledge databases
#######################################################################################################################
ALL_TEMPLATES = list_folders("latex_templates")
ALL_DATABASES = ["(None)"] + list_folders("knowledge_databases")
#######################################################################################################################
# Gradio UI
#######################################################################################################################
theme = gr.themes.Default(font=gr.themes.GoogleFont("Questrial"))
# .set(
# background_fill_primary='#E5E4E2',
# background_fill_secondary = '#F6F6F6',
# button_primary_background_fill="#281A39"
# )
ANNOUNCEMENT = """
# Auto-Draft: 学术写作辅助工具
本Demo提供对[Auto-Draft](https://github.com/CCCBora/auto-draft)的学术论文模板生成功能的测试. 学术综述和Github文档功能正在开发中.
## 主要功能
通过输入想要生成的论文名称（比如Playing atari with deep reinforcement learning)，即可由AI辅助生成论文模板.
***2023-06-13 Update***:
- 增加了最新的gpt-3.5-turbo-16k模型的支持.
***2023-06-13 Update***:
1. 新增‘高级选项-Prompts模式’. 这个模式仅会输出用于生成论文的Prompts而不会生成论文本身. 可以根据自己的需求修改Prompts, 也可以把Prompts复制给其他语言模型.
2. 把默认的ICLR 2022模板改成了Default模板. 不再显示ICLR的页眉页尾.
3. 中文支持: 暂不支持. 建议使用英文生成论文, 然后把输出结果送入[GPT 学术优化](https://github.com/binary-husky/gpt_academic)中的Latex全文翻译、润色功能即可.
4. 使用GPT-4模型：
- 点击Duplicate this Space, 进入Settings-> Repository secrets, 点击New Secret添加OPENAI_API_KEY为自己的OpenAI API Key. 添加GPT4_ENBALE为1.
- 或者可以访问[Auto-Draft-Private](https://huggingface.co/spaces/auto-academic/auto-draft-private).
如果有更多想法和建议欢迎加入QQ群里交流, 如果我在Space里更新了Key我会第一时间通知大家. 群号: ***249738228***."""
ACADEMIC_PAPER = """## 一键生成论文初稿
1. 在Title文本框中输入想要生成的论文名称（比如Playing Atari with Deep Reinforcement Learning).
2. 点击Submit. 等待大概十五分钟(全文).
3. 在右侧下载.zip格式的输出，在Overleaf上编译浏览.
"""
REFERENCES = """## 一键搜索相关论文
(此功能已经被整合进一键生成论文初稿)
1. 在Title文本框中输入想要搜索文献的论文（比如Playing Atari with Deep Reinforcement Learning).
2. 点击Submit. 等待大概十分钟.
3. 在右侧JSON处会显示相关文献.
"""
REFERENCES_INSTRUCTION = """### References
这一栏用于定义AI如何选取参考文献. 目前是两种方式混合:
1. GPT自动根据标题生成关键字，使用Semantic Scholar搜索引擎搜索文献，利用Specter获取Paper Embedding来自动选取最相关的文献作为GPT的参考资料.
2. 用户上传bibtex文件，使用Google Scholar搜索摘要作为GPT的参考资料.
关于有希望利用本地文件来供GPT参考的功能将在未来实装.
"""
DOMAIN_KNOWLEDGE_INSTRUCTION = """### Domain Knowledge
这一栏用于定义AI的知识库. 将提供两种选择:
1. 各个领域内由专家预先收集资料并构建的的FAISS向量数据库. 目前实装的数据库
* (None): 不使用任何知识库
* ml_textbook_test: 包含两本机器学习教材The Elements of Statistical Learning和Reinforcement Learning Theory and Algorithms. 仅用于测试知识库Pipeline.
2. 自行构建的使用OpenAI text-embedding-ada-002模型创建的FAISS向量数据库. (暂未实装)
"""
OUTPUTS_INSTRUCTION = """### Outputs
这一栏用于定义输出的内容：
* Template: 用于填装内容的LaTeX模板.
* Models: 使用GPT-4或者GPT-3.5-Turbo生成内容.
* Prompts模式: 不生成内容, 而是生成用于生成内容的Prompts. 可以手动复制到网页版或者其他语言模型中进行使用.
"""
OTHERS_INSTRUCTION = """### Others
"""
style_mapping = {True: "color:white;background-color:green",
False: "color:white;background-color:red"} # todo: to match website's style
availability_mapping = {True: "AVAILABLE", False: "NOT AVAILABLE"}
STATUS = f'''## Huggingface Space Status
当`OpenAI API`显示AVAILABLE的时候这个Space可以直接使用.
当`OpenAI API`显示NOT AVAILABLE的时候这个Space可以通过在左侧输入OPENAI KEY来使用. 需要有GPT-4的API权限.
当`Cache`显示AVAILABLE的时候, 所有的输入和输出会被备份到我的云储存中. 显示NOT AVAILABLE的时候不影响实际使用.
`OpenAI API`: {availability_mapping[IS_OPENAI_API_KEY_AVAILABLE]}. `Cache`: {availability_mapping[IS_CACHE_AVAILABLE]}.'''
def clear_inputs(*args):
return "", ""
def clear_inputs_refs(*args):
return "", 5
def wrapped_generator(
paper_title, paper_description, # main input
openai_api_key=None, openai_url=None, # key
tldr=True, max_kw_refs=10, bib_refs=None, max_tokens_ref=2048, # references
knowledge_database=None, max_tokens_kd=2048, query_counts=10, # domain knowledge
paper_template="ICLR2022", selected_sections=None, model="gpt-4", prompts_mode=False, # outputs parameters
cache_mode=IS_CACHE_AVAILABLE # handle cache mode
):
# if `cache_mode` is True, then follow the following steps:
# check if "title"+"description" have been generated before
# if so, download from the cloud storage, return it
# if not, generate the result.
if bib_refs is not None:
bib_refs = bib_refs.name
if openai_api_key is not None:
openai.api_key = openai_api_key
try:
openai.Model.list()
except Exception as e:
raise gr.Error(f"Key错误. Error: {e}")
if cache_mode:
from utils.storage import list_all_files, download_file
# check if "title"+"description" have been generated before
input_dict = {"title": paper_title, "description": paper_description,
"generator": "generate_draft"}
file_name = hash_name(input_dict) + ".zip"
file_list = list_all_files()
# print(f"{file_name} will be generated. Check the file list {file_list}")
if file_name in file_list:
# download from the cloud storage, return it
download_file(file_name)
return file_name
try:
output = generate_draft(
paper_title, description=paper_description, # main input
tldr=tldr, max_kw_refs=max_kw_refs, bib_refs=bib_refs, max_tokens_ref=max_tokens_ref, # references
knowledge_database=knowledge_database, max_tokens_kd=max_tokens_kd, query_counts=query_counts, # domain knowledge
sections=selected_sections, model=model, template=paper_template, prompts_mode=prompts_mode, # outputs parameters
)
if cache_mode:
from utils.storage import upload_file
upload_file(output)
except Exception as e:
raise gr.Error(f"生成失败. Error: {e}")
return output
def wrapped_references_generator(paper_title, num_refs, openai_api_key=None):
if openai_api_key is not None:
openai.api_key = openai_api_key
openai.Model.list()
return generate_top_k_references(paper_title, top_k=num_refs)
with gr.Blocks(theme=theme) as demo:
gr.Markdown(ANNOUNCEMENT)
with gr.Row():
with gr.Column(scale=2):
key = gr.Textbox(value=openai_key, lines=1, max_lines=1, label="OpenAI Key",
visible=not IS_OPENAI_API_KEY_AVAILABLE)
url = gr.Textbox(value=None, lines=1, max_lines=1, label="URL",
visible=False)
# 每个功能做一个tab
with gr.Tab("学术论文"):
gr.Markdown(ACADEMIC_PAPER)
title = gr.Textbox(value="Playing Atari with Deep Reinforcement Learning", lines=1, max_lines=1,
label="Title", info="论文标题")
description_pp = gr.Textbox(lines=5, label="Description (Optional)", visible=True,
info="这篇论文的主要贡献和创新点. (生成所有章节时共享这个信息, 保持生成的一致性.)")
with gr.Accordion("高级设置", open=False):
with gr.Row():
with gr.Column(scale=1):
gr.Markdown(OUTPUTS_INSTRUCTION)
with gr.Column(scale=2):
with gr.Row():
template = gr.Dropdown(label="Template", choices=ALL_TEMPLATES, value="Default",
interactive=True,
info="生成论文的模板.")
model_selection = gr.Dropdown(label="Model", choices=MODEL_LIST,
value=DEFAULT_MODEL,
interactive=GPT4_INTERACTIVE,
info="生成论文用到的语言模型.")
prompts_mode = gr.Checkbox(value=False, visible=True, interactive=True,
label="Prompts模式",
info="只输出用于生成论文的Prompts, 可以复制到别的地方生成论文.")
sections = gr.CheckboxGroup(
choices=["introduction", "related works", "backgrounds", "methodology", "experiments",
"conclusion", "abstract"],
type="value", label="生成章节", interactive=True, info="选择生成论文的哪些章节.",
value=DEFAULT_SECTIONS)
with gr.Row():
with gr.Column(scale=1):
gr.Markdown(REFERENCES_INSTRUCTION)
with gr.Column(scale=2):
max_kw_ref_slider = gr.Slider(minimum=1, maximum=20, value=10, step=1,
interactive=True, label="MAX_KW_REFS",
info="每个Keyword搜索几篇参考文献", visible=False)
max_tokens_ref_slider = gr.Slider(minimum=256, maximum=8192, value=2048, step=2,
interactive=True, label="MAX_TOKENS",
info="参考文献内容占用Prompts中的Token数")
tldr_checkbox = gr.Checkbox(value=True, label="TLDR;",
info="选择此筐表示将使用Semantic Scholar的TLDR作为文献的总结.",
interactive=True)
gr.Markdown('''
上传.bib文件提供AI需要参考的文献.
''')
bibtex_file = gr.File(label="Upload .bib file", file_types=["text"],
interactive=True)
with gr.Row():
with gr.Column(scale=1):
gr.Markdown(DOMAIN_KNOWLEDGE_INSTRUCTION)
with gr.Column(scale=2):
query_counts_slider = gr.Slider(minimum=1, maximum=20, value=10, step=1,
interactive=True, label="QUERY_COUNTS",
info="从知识库内检索多少条内容", visible=False)
max_tokens_kd_slider = gr.Slider(minimum=256, maximum=8192, value=2048, step=2,
interactive=True, label="MAX_TOKENS",
info="知识库内容占用Prompts中的Token数")
# template = gr.Dropdown(label="Template", choices=ALL_TEMPLATES, value="Default",
# interactive=True,
# info="生成论文的参考模板.")
domain_knowledge = gr.Dropdown(label="预载知识库",
choices=ALL_DATABASES,
value="(None)",
interactive=True,
info="使用预先构建的知识库.")
local_domain_knowledge = gr.File(label="本地知识库 (暂未实装)", interactive=False)
with gr.Row():
clear_button_pp = gr.Button("Clear")
submit_button_pp = gr.Button("Submit", variant="primary")
# with gr.Tab("文献搜索"):
# gr.Markdown(REFERENCES)
#
# title_refs = gr.Textbox(value="Playing Atari with Deep Reinforcement Learning", lines=1, max_lines=1,
# label="Title", info="论文标题")
# slider_refs = gr.Slider(minimum=1, maximum=100, value=5, step=1,
# interactive=True, label="最相关的参考文献数目")
# with gr.Row():
# clear_button_refs = gr.Button("Clear")
# submit_button_refs = gr.Button("Submit", variant="primary")
with gr.Tab("文献综述 (Coming soon!)"):
gr.Markdown('''
Coming soon!
''')
with gr.Tab("Github文档 (Coming soon!)"):
gr.Markdown('''
Coming soon!
''')
with gr.Column(scale=1):
gr.Markdown(STATUS)
file_output = gr.File(label="Output")
json_output = gr.JSON(label="References")
# def wrapped_generator(
# paper_title, paper_description, # main input
# openai_api_key=None, openai_url=None, # key
# tldr=True, max_kw_refs=10, bib_refs=None, max_tokens_ref=2048, # references
# knowledge_database=None, max_tokens_kd=2048, query_counts=10, # domain knowledge
# paper_template="ICLR2022", selected_sections=None, model="gpt-4", prompts_mode=False, # outputs parameters
# cache_mode=IS_CACHE_AVAILABLE # handle cache mode
# ):
clear_button_pp.click(fn=clear_inputs, inputs=[title, description_pp], outputs=[title, description_pp])
submit_button_pp.click(fn=wrapped_generator,
inputs=[title, description_pp, key, url,
tldr_checkbox, max_kw_ref_slider, bibtex_file, max_tokens_ref_slider,
domain_knowledge, max_tokens_kd_slider, query_counts_slider,
template, sections, model_selection, prompts_mode], outputs=file_output)
# clear_button_refs.click(fn=clear_inputs_refs, inputs=[title_refs, slider_refs], outputs=[title_refs, slider_refs])
# submit_button_refs.click(fn=wrapped_references_generator,
# inputs=[title_refs, slider_refs, key], outputs=json_output)
demo.queue(concurrency_count=1, max_size=5, api_open=False)
demo.launch(show_error=True)