Artificial intelligence news.
Straight from the original announcements.

Alibaba Cloud DevOps Adds Review Memory for Team-Specific Code Rules

November 9, 2026Source: Alibaba Cloud
Alibaba Cloud DevOps Adds Review Memory for Team-Specific Code Rules
Image: Alibaba Cloud

Alibaba Cloud DevOps has introduced Review Memory, a feature that stores team-specific engineering rules for reuse in future AI-powered code reviews. The company says the capability is designed to reduce false positives and repeated explanations caused by missing repository context.

Key points
  • Review Memory stores repository-specific rules for future AI code reviews.
  • It covers repository rules, business constraints, team preferences and risk experience.
  • Teams add memories through comments in merge requests.
  • Alibaba Cloud reports a 96.9% reduction in incorrect comments on an internal test set.
  • Administrators can control Review Memory and require approval for candidate memories.

Turning review explanations into reusable rules

Intelligent Code Review can identify common code issues, but Alibaba Cloud says that reliable reviews also require an understanding of a repository's technical constraints, business rules and team preferences. Review Memory captures context that has already been clarified during reviews and applies it to relevant future changes.

The feature is intended to preserve rules that might otherwise depend on individual team members' knowledge. Examples include interfaces used by legacy systems, established exception-handling practices and related logic that must be updated when a particular field changes.

Teams can add a memory by telling Intelligent Code Review in a merge request comment what they want it to remember. The system then organizes the information into a candidate memory for the current repository. Once the memory takes effect, future reviews automatically use it when evaluating relevant changes.

Four types of repository context

Alibaba Cloud identifies four categories that are particularly suitable for Review Memory: repository rules, business constraints, team preferences and risk-related experience. Repository rules can describe conventions that apply to particular packages, modules, classes or interfaces. Business constraints can cover legacy compatibility paths, fields in external systems and staged rollout requirements.

Team preferences include long-standing implementation approaches, naming styles and exception-handling strategies. Risk-related experience covers types of changes that have historically caused problems and require additional checks.

The company gives several examples. A team can record that a database automatically updates modification timestamps, preventing future reviews from treating the lack of manual assignment as an omission. In an e-commerce application, it can also specify that courier deliveries require a shipping address, while in-store pickup requires a selected pickup store instead.

Another example concerns DTOs in a compatibility package. A rule can require classes whose names end in DTO to retain a public no-argument constructor because a legacy compatibility path creates objects through reflection.

Internal test reported fewer incorrect comments

Alibaba Cloud says its DevOps team built a test set and conducted comparative tests to assess whether Review Memory could reduce incorrect comments caused by missing information, limited repository context or unfamiliarity with team conventions. On the samples in that internal test set, using Review Memory reduced incorrect review comments by 96.9%.

The company explicitly says that 96.9% is a reduction in incorrect comments on the internal test set, not an overall accuracy rate for all repositories and review scenarios. Alibaba Cloud says the intended effects include fewer low-value distractions, less repeated explanation of team conventions and more consistent reviews based on the same repository rules.

The source also stresses that long-term memories need governance. Repository administrators can enable or disable Review Memory through Repository Settings, AI Assistant Settings, Intelligent Code Review, Review Memory. Administrators can also enable memory approval, after which newly generated candidate memories must be confirmed by an administrator before they take effect. This gives teams a way to review, revise and disable repository context rather than allowing every one-off explanation to remain active indefinitely.

Alibaba CloudAlibaba Cloud DevOpsReview MemoryIntelligent Code Reviewcodingenterprise AI

Related news

Products and tools
Products and toolsNovember 9, 2026, 13:07

Samepage.ai launches Bets to coordinate AI-era product teams

Samepage.ai has launched Samepage Bets, a platform that brings product research, software development and go-to-market work onto one board. The company says human team…

Neo4j launches GraphRAG kit and explorer for custom knowledge graphs
For developersNovember 9, 2026, 09:39

Neo4j launches GraphRAG kit and explorer for custom knowledge graphs

Neo4j has introduced graph-explorer-kit, a Python package and API that let developers query, extend and explore knowledge graphs built in Neo4j Aura. The accompanying…

Alibaba Cloud introduces Kitta, a domain-focused code review agent
For developersNovember 6, 2026

Alibaba Cloud introduces Kitta, a domain-focused code review agent

Alibaba Cloud has introduced Kitta, a code review agent designed to use project-specific knowledge and tailored workflows instead of applying the same review approach to…