0.3.0 更新规划 #2660
Replies: 176 comments 162 replies
-
|
实现对RAG结果的评估 |
Beta Was this translation helpful? Give feedback.
-
|
这个目前都是各大大模型厂商在做的内容，估计不好去竞争。。。 |
Beta Was this translation helpful? Give feedback.
-
咱们现在的应用程序，用开源的哪个评估框架比较合适方便一点呀 |
Beta Was this translation helpful? Give feedback.
-
|
实现知识库文档内容追加 |
Beta Was this translation helpful? Give feedback.
-
|
这个现在也可以做呀 |
Beta Was this translation helpful? Give feedback.
-
|
@zRzRzRzRzRzRzR 请问现在怎么实现？好像只能重新上传新的文档？ |
Beta Was this translation helpful? Give feedback.
-
|
如果是对URL地址的所在内容进行知识库加载，这个没有 |
Beta Was this translation helpful? Give feedback.
-
|
增加多模态Agent功能 |
Beta Was this translation helpful? Give feedback.
-
|
这个有的 |
Beta Was this translation helpful? Give feedback.
-
|
考虑api的稳定性和可扩展性，二开的话有这方面需求 |
Beta Was this translation helpful? Give feedback.
-
|
可以，之后API估计会精简很多 |
Beta Was this translation helpful? Give feedback.
-
|
es混合检索 |
Beta Was this translation helpful? Give feedback.
-
|
也不一定是ES，只要双路召回即可提升。 |
Beta Was this translation helpful? Give feedback.
-
|
多路召回挺好做的，改下代码就行，就是没做分布并发式多路召回 |
Beta Was this translation helpful? Give feedback.
-
|
怎么个改下代码就行？ |
Beta Was this translation helpful? Give feedback.
-
|
用chatglm3-6b实现数据分析的功能。。。。感觉很难，但是值得尝试吧 |
Beta Was this translation helpful? Give feedback.
-
|
chatglm3-6b本身没有这个能力，4具备但是暂时没有开源消息，会进行跟进 |
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
|
现在就有... |
Beta Was this translation helpful? Give feedback.
-
|
有这个逻辑的现在，只是不知道有啥用，因为这个项目跟微调没有联起来 |
Beta Was this translation helpful? Give feedback.
-
|
建议增加文本切片后的人工干预功能，类似FastGPT的可以人工上调整分片后的内容。 |
Beta Was this translation helpful? Give feedback.
-
|
没有太理解，可视化切片吗 |
Beta Was this translation helpful? Give feedback.
-
|
类似这样可以调吧
|
Beta Was this translation helpful? Give feedback.
-
|
@zRzRzRzRzRzRzR 对就是您说的可视化切片，就像楼上这位朋友贴的图中的功能。 |
Beta Was this translation helpful? Give feedback.
-
|
对对对，这个需要的 |
Beta Was this translation helpful? Give feedback.
-
|
增加对话管理功能，支持本地多个对话 |
Beta Was this translation helpful? Give feedback.
-
|
mark一下 |
Beta Was this translation helpful? Give feedback.
-
|
未来的大语言模型，可以预见上下文的会提高的很快，是否可以针对上下文支持比较好的大模型，比如达到32k以上，将知识库匹配的全文给到上下文，而不是给切片 |
Beta Was this translation helpful? Give feedback.
-
|
那成本会很高，这个还是不太现实目前 |
Beta Was this translation helpful? Give feedback.
-
|
Qwen1.5-72B-Chat-AWQ支持32K上下文，Yi-34B-200K支持200K，我是售前，写的方案经常要上万字（厂商白皮书也经常是万字起），现有切片对知识库上万字的文档解析结果差强人意，基本查不到里面的特定内容，如果能 优化知识库全文匹配真的能更实用的，求优化升级！ |
Beta Was this translation helpful? Give feedback.
-
|
我也有这个诉求，KB的召回准确率很大程度取决于你录入的文档质量，知识库也需要合理划分，防止相互干扰。另外还需要多语言的支持。如果能有更好的RAG机制，比如Self RAG，可以让结果更好些 |
Beta Was this translation helpful? Give feedback.
-
|
chatglm3-6b 目前也有32K的 |
Beta Was this translation helpful? Give feedback.
-
|
增加对mysql等数据库的问答 |
Beta Was this translation helpful? Give feedback.
-
|
本地模型扛不住，这个内容可能作为一个tools开放给开发者自己适配，并加入到Agent系统内 |
Beta Was this translation helpful? Give feedback.
-
|
没毛病 |
Beta Was this translation helpful? Give feedback.
-
|
新版本是否可以支持GLM-4的在线api调用？ |
Beta Was this translation helpful? Give feedback.
-
|
这个肯定可以 |
Beta Was this translation helpful? Give feedback.
-
|
这个现在就支持了吧 |
Beta Was this translation helpful? Give feedback.
-
|
支持多agent 增加agent的使用场景和实用性 |
Beta Was this translation helpful? Give feedback.
-
|
本地模型扛不住，我们大概率不会再0.3.0推出，会在0.3.x后期版本做 |
Beta Was this translation helpful? Give feedback.
-
|
多agent 现在可以实现吗？
***@***.***
…------------------ 原始邮件 ------------------
发件人: ***@***.***>;
发送时间: 2024年1月18日(星期四) 下午5:29
收件人: ***@***.***>;
抄送: ***@***.***>; ***@***.***>;
主题: Re: [chatchat-space/Langchain-Chatchat] 0.3.0 更新规划 (Discussion #2660)
现在就有...
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you commented.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
本地模型扛不住，我们大概率不会再0.3.0推出，会在0.3.x后期版本做 |
Beta Was this translation helpful? Give feedback.
-
|
希望有Tools的开关功能，可以放在API的请求参数里面，用户在使用Agent问答时可以自由控制需要用到哪些Tools，降低每次请求时使用错误工具的风险以及无用工具的数量。 |
Beta Was this translation helpful? Give feedback.
-
|
简单梳理一下需求：
|
Beta Was this translation helpful? Give feedback.
-
|
Beta Was this translation helpful? Give feedback.
-
|
个人觉得需要修改的： |
Beta Was this translation helpful? Give feedback.
-
|
其实大家都忽略了一个重要的问题： 匹配相似度阈值的问题。 在项目里这个匹配相似度很混乱，有的模型是0-1，有的模型是0-2；而且并不是值越大相似度就越高，这样用起来就不好统一了。 建议全部统一成0-1，值越高表示相似度越高； |
Beta Was this translation helpful? Give feedback.
-
|
我也发现这个问题，但看了代码，里面有作正则化。按理，如果是cosine距离，取值应该是-1~1的区间。实际中距离可以超过1，不知道怎么得到的。 |
Beta Was this translation helpful? Give feedback.
-
|
感觉可以像text-generation-webui那样, 把配置文件挪到网页里去设置, 如多显存分层比例之类的. 要是能修改AI回答, 以此来引导AI就更好了. 国内的RWKV-Runne的小说续写模式也蛮不错, 但就是模型不通用. |
Beta Was this translation helpful? Give feedback.
-
|
qwlong 有计划支持吗 |
Beta Was this translation helpful? Give feedback.
-
|
都可，改model config文件路径就行
…---原始邮件---
发件人: ***@***.***>
发送时间: 2024年6月4日(周二) 上午10:29
收件人: ***@***.***>;
抄送: ***@***.******@***.***>;
主题: Re: [chatchat-space/Langchain-Chatchat] 0.3.0 更新规划 (Discussion #2660)
image.png (view on web)
现chatglm3-6b模型，是否只能放在Langchain-Chatchat目录下，还是能在哪里可配置到其他路径下
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you commented.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
知识库问答，可以根据问题去检索回答图片或图片地址吗？ |
Beta Was this translation helpful? Give feedback.
-
可否支持上传的知识库有图片，然后回答输出库里面的图片 |
Beta Was this translation helpful? Give feedback.
-
|
上传文件提问对文件进行摘要总结 |
Beta Was this translation helpful? Give feedback.
-
|
希望新版本考虑对于多模态RAG的支持，也就是对于文档中的图片，比如流程图，脑图，表格等图片能按照逻辑顺序解析出其中的文字。 |
Beta Was this translation helpful? Give feedback.
-
|
希望知识库问答同时调用多个知识库 |
Beta Was this translation helpful? Give feedback.
-
|
希望新版本中添加text2sql功能 |
Beta Was this translation helpful? Give feedback.
-
|
希望新版本中添加text-to-sql功能 |
Beta Was this translation helpful? Give feedback.
-
|
已经有啦 |
Beta Was this translation helpful? Give feedback.
-
|
自定义Agent的详细教程。 |
Beta Was this translation helpful? Give feedback.
-
|
你好，请问新版本能够使LLM与进行知识图谱连接吗 |
Beta Was this translation helpful? Give feedback.
and others
-
随着 LangChain 0.1.x 的更新和 LLM 技术的迭代，
Langchain-Chatchat 0.2.x已经无法满足日益增长的入门实战需求。随着
Langchain-Chatchat 0.2.10的发布，0.2.x内容在不在进行维护和支持。我们预计会在 2024 Q1 推出 Langchain-Chatchat 0.3.0.
Langchain-Chatchat 0.3.x系列将进行大规模的重写，包括但不限于请注意，Langchain-Chatchat 作为 开源，非商业性 的 RAG 框架，其主要目的是让更多人能体验到最新的大模型RAG相关技术和应用，项目更多在内容上进行更新，因此不会优先考虑以下需求
如果你对我们有什么期待，可以在这个公告下面提出。
Beta Was this translation helpful? Give feedback.
All reactions