Core Stability 🔥🔥🔥🔥🔥

Objective: Ensure users can use the product stably and smoothly, and continuously fix issues submitted by users

• Regular release plan
• Package extensions
• Sophisticated unit tests / integration tests / e2e tests
• Benchmark

Performance and Experience Optimization 🔥🔥🔥🔥

Goal: Ensure users experience no waiting or lag

• Support concurrent user requests. Continue to improve the relevant interfaces of /api/langgraph-compat and support capabilities such as stream join.
  • feat(persistence): add unified persistence layer with event store, token tracking, and feedback #1851
• FrontEnd performance improvement

Channel and Integration Expansion 🔥🔥🔥

Goal: Make Agent ubiquitous

• Implement the built-in Cron scheduler
  • feat(cron): add cron scheduling system for reminders and recurring tasks #1092
• LLM Provider extensions
  • fix(kimi): preserve reasoning content across follow-up turns #1821
  • feat(models): add vLLM provider support #1860
• new channels (Matrix, DingTalk, JIRA, WeCom)
  • feat: add WeChat channel integration #1869
  • feat(channels): add Discord channel integration #1806
  • fix(channels): normalize slack allowed user ids #1802
  • fix: Improve thinking placeholder implementation for IM channels #1514
  • feat: support wecom channel #1390

Security and Permission Strengthening 🔥🔥🔥🔥

Objective: Meet enterprise-level security compliance requirements

• Implement Role-Based Access Control (RBAC)
  • RFC: DeerFlow user Authentication Module DeerFlow 用户认证模块设计 #1721
• Improve the sandbox security mechanism
  • feat(sandbox): add read-only support for local sandbox path mappings #1808
  • feat(sandbox): strengthen bash command auditing with compound splitting and expanded patterns #1881
• MCP and Tools calling auditing
  • feat(guardrails): add pre-tool-call authorization middleware with pluggable providers #1240

Hierarchical Memory Architecture 🔥🔥🔥🔥🔥

Objective: Enable the Agent to remember long-term conversations and become smarter as the conversation progresses

RFCs:

• Layered Memory Architecture for DeerFlow | 为 DeerFlow 引入分层记忆架构 #1590
• RFC: Self-Improving Agent — MVP (structured reflection + correction detection in MemoryMiddleware) #1620

Skills and Extensions 🔥🔥🔥🔥🔥

Objective: Provide a way for the user to extend the functions of DeerFlow to solve their daily problem

• auditing customer skills
• Skills Hub?
• RFCs
  • RFC: Skill Self-Evolution — Agent-Managed Skill Creation & Improvement #1865
• Skills implementations:
  • feat(skills): add academic-paper-review, code-documentation, and newsletter-generation skills #1861

Documentation and Developer Experience 🔥🔥🔥🔥

Objective: Enable new users to get started within 30 minutes and reduce the number of issues by 40%

• Document website
• User guide
• Video Tutorials