ROADMAP 2025

[yingfeng]

v0.23.0

• Replace Flask with Quart to increase service concurrency. #11277
• Orchestrating data transform nodes for data pipeline. [Feature Request]: Enhance orchestrating data transform nodes for data pipeline. #10427
• Support memory for agent. [Feature Request]: Support memory for Agent #4213

v0.22.0

• [Feature Request] Support multiple data sources synchronizations, like S3, Confluence, Discord, etc; #10953
• Integrate MinerU as one of the document parser. Feat: add MinerU parser #10621
• Admin UI

v0.21.0

• Support pipeline orchestration for data ingestion. [Feature Request]: Use data pipeline to visualize the parsing configuration of the knowledge base #9869
• Support long-context RAG for long documents with much better retrieval qualities. #10436
• Manually trigger knowledge graph and Raptor generation for the entire dataset. [Feature Request]: Manually trigger knowledge graph and Raptor generation #10443
• Add admin CLI and admin service. [Feature Request]: Add admin CLI and admin service #10241

v0.20.0

• A brand new agent and workflow builder: support agent and workflow orchestration in a unified canvas. [Feature Request]: Redesign and refactor our agent module #9082 [Feature Request]: Is it possible to edit the model's system prompt words when creating an agent? #6365
• Enable agents to call MCP tools [Feature Request]: When can the agent support the invocation of MCP services? #7936 [Feature Request]: Expanded the capabilities of the MCP Server #8644 [Feature Request]: To make the workflow more powerful, apply for an MCP node #7742
• Knowledge Retrieval enhancement, including using as a tool to Agent, cross-lang search, etc [Feature Request]: Redesign and refactor our agent module #9082 [Feature Request]: Retrieval component support cross-lang search #8107
• Support sharing agents with initial variables pre-configured [Feature Request]: When the Agent uses variables at the beginning, that is, when the Begin component adopts variables, this Agent cannot embed into other websites #7291 [Question]: Upload file function after embedding webpage #7486 [Feature Request]: Begin component takes a variable so I cannot embed it into a webpage, but translation agent needs to get user's input #7322
• Agent now can see the world with vision feature [Feature Request]: Redesign and refactor our agent module #9082
• Display runtime logs when interacting with agents [Feature Request]: Redesign and refactor our agent module #9082
• Developers now can see the chat logs when their users chat with Agent on the embedded page [Feature Request]: Audit logs to record my users' questions and system replies #4577
• Users can clear chat history during chating with an Agent [Question]: Why do the results obtained by accessing the agent through the api are different from those obtained by accessing the console #8371
• Redesign the UI and introduce dark mode [Feature Request]: Redesign and refactor our agent module #9082

v0.19.0

• Support cross-language search in the knowledge base and chat modules. [Feature Request]: Cross-language query #7376
• Add a new code component to the Agent module. [Feature Request]: Add the JS code (or other) executor component to Agent. #4977
• [Feature Request]: Display inline (non-quoted) images in the chat and search modules #7623
• [Feature Request]: Add image preview for images in documents(Retrieval test). #7608

v0.18.0

• [Feature Request]: Implement Model Context Protocol (MCP) server for RAGFlow #4344
• [Feature Request]: Citation in answer optimization. #5905
• Agent can be shared betweem team members. Feat: add agent share team viewer #6222
• Supports agent version history. Feat: support agent version history. #6130
• [Feature Request]: interrupt streaming #6515
• OpenAI compatible API for chat assistant and agent. Feat: add OpenAI compatible API for agent #6329
• Supports disabling user sign-up. Feat: Add user registration toggle feature #6327
• Figures from PDF/Docx can be parsed via VLMs. Feat: add VLM-boosted PDF parser #6278

v0.17.0

• Supports reasoning triggered RAG #5173
• Improve prompt reference form Agent [Feature Request]: Inner prompt parameter setting. #4764
• [Feature Request]: How to enable internet search? #5198

v0.16.0

• Refactor GraphRAG #4543
• Supports Loop within Agent #4242
• Supports tagging/labeling for document chunk and query #4367
• [Feature Request]: Metadata in documents for improve the prompt #3690

Long-term plan

• RAGFlow documents [Feature Request]: Add guides, references, best practices, and other documentation that may facilitate user journey with RAGFlow #720
• APIs [Feature Request]: RAGFlow API proposal #1102

Backlogs

• Implement EraRAG as an alternative for RAPTOR/GraphRAG. [Feature Request]: Implement EraRAG as an alternative for RAPTOR and GraphRAG #8614

Previous roadmap

ROADMAP 2024

[Jeocy]

Does RAGFlow not have an administrator entrance?

[alexxchen]

How about supporting multi-modal RAG?

[chminsc]

mobile ui please
option to disable user registration

[Zeta-ZhaoLi]

user management method, currently if I expose it to the internet, anyone can register
PostgresSQL support

[hoharald]

UX improvements for the chat

• multiple lines for chat input (HTML tag input => textarea)
• markdown rendering, since response from LLMs supports it

Mentioned here

[15053285989]

希望支持导出和导入agent工作流
Hope to support exporting and importing Agent workflows.

[yingfeng]

How about supporting multi-modal RAG?

Under planning, it requires Infinity to be production ready such that Tensor based reranker as well as multi vector retrieval could be available.

[Echoole]

希望知识库可以分享一个连接，这样别人点看，可以询问问题，不用登录平台，类似与分享了一个聊天界面，上面可以定义机器人可以提供什么服务

[cike8899]

希望支持导出和导入agent工作流 Hope to support exporting and importing Agent workflows.

Already supported

[Andy-Zyu]

How do I need to be able to call the visual understanding model in the agent? Can the user call the corresponding visual understanding model after uploading the picture?

[zmxccxy]

1、新建助理，可以不选择关联知识库，就只与大模型对话。
2、助理的对话页面美化下，比如对话框颜色、小组件按钮、进行中/加载中的提示。助理的对话框显示效果支持更多格式，比如md、code等。对话输入框支持语音输入。设置了助理头像后，新建对话，还是默认的头像，需要过一段时间才行。
3、agent，可以自定义头像。
4、模型供应商，已加入的，显示面积可以调整下，太宽了。
5、完善第三方嵌入，要显示助理标题和头像。
6、知识库，加入启用、解析等筛选项，知识库中文件多了就不好找了。

感谢🙏

我也想加入开发中，编码实现某些功能，但是我不会 😭。

[rplescia]

user management method, currently if I expose it to the internet, anyone can register PostgresSQL support

I agree with user management, this feature is needed very much. Even if it is basic user management to start with, the ability to disable user registration and only invite users by email.

Ideally, it would be great if we could get the following user management features

• User management page with the ability to invite/disable/remove users and add users to groups
  • Admin group (full system access with user management)
  • Standard group (can access everything except 'user management' and 'system' features[similar to user-settings page on the demo site])
• Multiple teams
  • The ability to create multiple teams and add different people to each team. I don't always want to share a knowledge base with everyone in my team, I need a way to manage my teams and the people in each team so that I can maintain some knowledge base security

[n40k264]

SSO/IAM support pls

[yumu7777777]

请支持 SSO/IAM

您好SSO/IAM 有好用的开源项目吗