search

lobehub / lobe-chat

🤯 Lobe Chat - an open-source, modern-design AI chat framework. Supports Multi AI Providers( OpenAI / Claude 3 / Gemini / Ollama / Azure / DeepSeek), Knowledge Base (file upload / knowledge management / RAG ), Multi-Modals (Vision/TTS) and plugin system. One-click FREE deployment of your private ChatGPT/ Claude application.
https://chat-preview.lobehub.com
Other
42.93k stars 9.65k forks source link

[Request] Lobe Chat 的知识库功能超赞，但是可不可以支持完整文件内容推理？ #4005

Open ShinChven opened 4 weeks ago

ShinChven commented 4 weeks ago

🥰 Feature Description

我用了Lobe Chat 的知识库功能，对我查找信息很有帮助。但是每次只能从5个片段中总结信息，当我想从一整个文件里面总结信息的时候，发现会缺失信息。有没有可能像 Google AI Studio 那样利用超大上下文窗口直接挂在一整个或多个文件来生成？

🧐 Proposed Solution

在支持库推理时支持将1个或多个文件的完整内容作业附加知识库，而不是仅使用搜索到的相关片段。

📝 Additional Information

No response

lobehubbot commented 4 weeks ago

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

🥰 Feature Description

I used the knowledge base function of Lobe Chat, which was very helpful for me to find information. But I can only summarize information from 5 fragments at a time. When I want to summarize information from an entire file, I find that information is missing. Is it possible to directly hang one or more files to generate using a large context window like Google AI Studio does?

🧐 Proposed Solution

Supports attaching the complete content job of 1 or more files to the knowledge base when supporting library inference, instead of just using the relevant fragments that were searched.

📝 Additional Information

No response

lobehubbot commented 4 weeks ago

👀 @ShinChven

Thank you for raising an issue. We will investigate into the matter and get back to you as soon as possible. Please make sure you have given us as much context as possible.\ 非常感谢您提交 issue。我们会尽快调查此事，并尽快回复您。 请确保您已经提供了尽可能多的背景信息。

arvinxx commented 4 weeks ago

可以的，这个的确有计划。会准备做成可选的，比如是所有文件内容都塞进去，还是只找部分片段

lobehubbot commented 4 weeks ago

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

Yes, there is indeed a plan for this. Will prepare to make it optional, such as whether to stuff all the file contents or only find some fragments

arvinxx commented 4 weeks ago

现在模型的 context 这么大，其实不差钱的情况下，都塞进去大概率效果更好

lobehubbot commented 4 weeks ago

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

The context of the model is so large now. In fact, if you don’t have a lot of money, you will probably get better results if you just put it all in.

ShinChven commented 4 weeks ago

现在模型的 context 这么大，其实不差钱的情况下，都塞进去大概率效果更好

最近在用 LLM 读论文，确实有这个需求。

lobehubbot commented 4 weeks ago

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

Now that the context of the model is so large, in fact, if you don’t have a lot of money, you will probably get better results if you put it all in.

I am currently reading a thesis using LLM, and I really have this need.

Zoumachuan commented 3 weeks ago

+1，希望能推出这个功能

lobehubbot commented 3 weeks ago

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

+1, I hope this feature can be launched

muhanstudio commented 2 weeks ago

另一个 issue 最后我引用了一个可以做到文件级别的嵌入的项目，希望也可以看看，大概是一个类似的解决方案

https://github.com/lobehub/lobe-chat/issues/4102

lobehubbot commented 2 weeks ago

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

In another issue, I finally quoted a project that can achieve file-level embedding. I hope you can also take a look. It is probably a similar solution.

https://github.com/lobehub/lobe-chat/issues/4102