暂无图片
暂无图片
暂无图片
暂无图片
暂无图片

RAG as a Service?

484

目前我们知道有好多中间层的产品,例如:Dify[1]、扣子[2]、vectorshift[3] 等等。

最近还看到了有一些公司在做 RAG as a Service,还不知道这个新概念的产品是做什么,本篇文章就是来了解一下目前在做 RAG as a Service 的产品都有哪些。

RAG as a Service (RAGaaS) is a service model that provides the benefits of RAG technology via cloud-based platforms. --- MAIA[4]

MAIA

With MAIA[5], you can easily complete your daily tasks with access to the entire company knowledge at the touch of a button for better results in less time - your AI platform for company knowledge.
MAIA 从介绍上来看主打的是安全性与隐私保护,从介绍的功能上来看有:文档上传、不同部门之间的数据权限与共享、对话。

Epsilla

Epsilla[6] is a one-stop RAG-as-a-Service platform for building production ready LLM applications connected with your proprietary data.
从演示视频里可以看到其能够选择 LLM,上传数据,选择场景:Chat bot、Agent、search、recommendation。

Nuclia

Nuclia[7] automatically indexes files and documents, from internal and external sources, to fuel diverse company use cases with LLMs. Get modular RAG in different Nuclia deployment options.
  • 对任何类型的数据进行索引,无论这些数据使用何种语言

  • 自定义数据检索的策略

  • 自定义 Chunking 策略

  • 不同 embedding 模型选择

Ragie

RAG. Right out of the box. Meet Ragie[8], fully managed RAG-as-a-Service for developers.

  • 可以连接 Google Drive, Notion, Confluence 上的数据,并保持持续更新
  • 内置re-ranking, summary index, entity extraction, flexible filtering, and hybrid semantic and keyword search 策略

Vectara[9]

Mitigating Hallucinations and Copyright Concerns, Minimizing Bias, Enhancing Explainability, and Broadening Cross-Lingual Reach. Your TRUSTED entry point for GenAI.
能够处理 emails, social media comments, web pages, customer support tickets, messages, survey responses, tweets 和 documents,

总结

RAG as a Service 的相关产品,看下来更像是 Dify 类产品的抽象,Langchain 的可视化,更像一个知识库的解决方案,但是功能上又比 Glean[10] 弱些。
现在的一些向量数据库产品正在朝这个方向走,比如 RAGFlow[11] 其实也可以叫做 RAG as a Service 的产品。
但是用户到底需要个什么样的知识库产品?光知识问答足够么?其他的业务场景如何满足?到底有多少的开放度更合适?Workflow 适合什么样的用户使用?这些问题目前看下来还是在探索的过程中。

相关连接

[1] https://dify.ai/
[2] https://www.coze.cn/
[3] https://vectorshift.ai/
[4] https://en.getmaia.ai/glossary/rag-as-a-service#:~:text=RAGaaS%20is%20based%20on%20the,the%20needs%20of%20different%20applications
[5] https://en.getmaia.ai/
[6] https://epsilla.com/
[7] https://nuclia.com/
[8] https://www.ragie.ai/
[9] https://vectara.com/
[10] https://www.glean.com/
[11] https://github.com/infiniflow/ragflow

👆 关注 AI 搜索引擎,获取更多专业技术分享 ~

文章转载自向量检索实验室,如果涉嫌侵权,请发送邮件至:contact@modb.pro进行举报,并提供相关证据,一经查实,墨天轮将立刻删除相关内容。

评论