ChatGPT席卷全球。数以百万计的人在使用它。但尽管它对通用知识很有帮助，但它只知道自己接受过哪些培训的信息，即2021年之前普遍可用的互联网数据。它不知道你的私人数据，也不知道最近的数据来源。
如果真的这样做了，那不是很有用吗？这就是LangChain的用武之地。
LangChain的目标是让每个人都更容易开发语言模型应用程序。我们最近在这里发布了一份关于如何通过您的数据创建自己的ChatGPT的指南。其中包括一个示例GitHub repo，用于启动和自定义。但即便如此，仍有一长串数据源需要集成和编写提示。我们在发出呼吁，看看什么是最有趣的集成，并得到了压倒性的回应后意识到了这一点。
在“Chat Your Data”挑战中，我们将发起为期一周的挑战，通过您的数据源创建ChatGPT。
动机
一如既往，这样做的动机是让每个人都更容易开发语言模型应用程序。特别是，我们认为示例对于帮助人们做到这一点至关重要。因此，我们希望为各种数据源提供尽可能多的示例（数据加载程序+提示）。
然后，我们将把数据加载逻辑放在LangChain中，把提示放在LangChainHub中，并把示例放在LangChain文档中，使其他人尽可能容易地开始。
如何开始
- 克隆示例GitHub 代码库
- 自定义数据源+数据提示（可以按照本教程操作）
- 额外的好处：部署一个不错的前端来配合它！在上面的教程中，我们有一个关于拥抱面部空间的示例部署。
- 使用此表格提交您的参赛作品
- 重复
示例
我们在这个示例GitHub repo的基础上创建了两个示例repo，以展示它可能是什么样子：
- Notion:：与你的理念相联系
- ReadTheDocs：连接您的ReadTheDoc网站
我们从上述推文中看到的其他消息来源的想法是：
- Obsidian
- Gong calls
- PDFs
- Audio files (can use Whisper!)
- Git repos
- Arbitrary websites
还有很多，更多！如果你正在寻找创意，只需查看这条推特的回复即可。
用您的数据聊天提交
上周，我们启动了一个为期一周的比赛，通过您自己的数据创建一个聊天机器人。目标是激发（1）大量用于加载各种类型数据的文档加载程序，（2）大量端到端示例。
我们在两方面都取得了成功！在文档加载程序方面，我们在College Confidential网站上添加了25个以上的文档加载程序，从s3到EveryNote不等。
在端到端的例子方面，我们很高兴能重点介绍下面的一些项目。作为提醒，比赛的获胜者将由本周末GitHub回购中拥有最多星星的人决定——所以如果你喜欢某个特定的一次性星星，那就去吧！
参赛者包括：
TokBot
- Author: Andrew Gleave
- Twitter: @andrewgleave
- GitHub Repo: https://github.com/andrewgleave/tokbot
- Description: A GPT-3/LangChain bot that answers questions about the TokCast podcast provides relevant video excerpts
Ask Everything About Me
- Author: Yongtae
- Twitter: @Yoooongtae
- GitHub Repo: https://github.com/Yongtae723/chat-your-data
- Description: This bot can do the following on your behalf: analyze your tweets and show you what you are interested in these days, answer questions about you from your blog or profile.
Roam QA
- Author: JimmyLv
- Twitter: @Jimmy_JingLv
- GitHub Repo: https://github.com/JimmyLv/roam-qa
- Description: Ask questions to your Roam Research graph in natural language
Chat-Your-Data Self Hosted
- Author: Misbah Syed
- Twitter: @MisbahSy
- GitHub Repo: https://github.com/misbahsy/chat-your-data-self-hosted
- Description: Create a ChatGPT like experience over your custom docs using LangChain. This repo can help you use models hosted on HuggingFace for embedding and for text generation.
Chat With Data
- Author: Echo
- Twitter: @hive_echo
- GitHub Repo: https://github.com/echohive42/chat_with_data
- Description: This app starts with the AI wikipedia article from data folder and you can chat about it.
Chat LangChain Telegram
- Author: Vidal Maxime
- Twitter: @vmaxmc2
- GitHub: https://github.com/vidalmaxime/chat-langchain-telegram
- UI: https://huggingface.co/spaces/maxime/chat-with-your-telegram-chat
- Description: Understand who you are and your relationships by creating a ChatGPT like experience over your own Telegram chat with LangChain.
Chat ISW Reports
- Author: Matthew Robinson
- Twitter: @UnstructuredIO
- GitHub: https://github.com/Unstructured-IO/chat-isw-reports
- Description: Chat over documents pulled from the Institute for the Study of War (ISW) website.
Chat Your Data Chef
- Author: Zander Chase
- Twitter: @zanderchase11
- GitHub: https://github.com/zanderchase/chat-your-data-chef
- UI: https://huggingface.co/spaces/zanderchase/chat-your-data-chef
- Description: Chat over recipes scraped from several recipes sites.
YouTuber Chatbot
- Author: James Pop
- GitHub: https://github.com/jryebread/youtuber_chatbot/tree/main
- UI: https://jryebread-youtuber-chatbot-main-v73i8j.streamlit.app/
- Description: make an ai chatbot out of any YouTube video
Investor Education ChatChain
- Author: Nicholas Muchinguri
- GitHub: https://github.com/nickmuchi87/investor-education-chatchain
- UI: https://huggingface.co/spaces/nickmuchi/Investor-Education-ChatChain
- Description: ChatBot for asking investing questions from web-scraped website of a leading Investment Manager.
Bonus (non-valid) Submissions
There were also a series of submissions that will not be counted (due to not open source or created before the submission deadline), but that we still want to highlight because they are awesome :)
ChatGPT+WolframAlpha+Whisper
- Author: James Weaver
- UI: https://huggingface.co/spaces/JavaFXpert/Chat-GPT-LangChain
- Description: A conversational agent implemented with OpenAI GPT-3.5 and LangChain
Paper QA
- Author: Andrew White
- GitHub Repo: https://github.com/whitead/paper-qa
- UI: https://huggingface.co/spaces/whitead/paper-qa
- Description: This is a simple and incomplete package for doing question and answering from PDFs or text files
Any Questions
- Author: Harry
- UI: https://www.anyquestions.ai/video
- Description: AskAI about a video
- 登录 发表评论