Improving Code Reviewer with Atlassian PR Context

This post details the integration of similar past PR context (both diff and comments) into the Rovo Dev Code Reviewer. Experiments showed that combining diff and comment context significantly improved precision and recall, with precisionPerfect reaching statistical significance. A deliberate design decision was made to restrict context retrieval to the same repository due to quality trade-offs with cross-repo context. Online experiments demonstrated a lift in resolution rate (from 31.92% to 41.90% externally) and increased engineer feedback. Case studies illustrate how this context helps with type signature correctness and understanding team framework patterns and culture. The post also highlights efficiency gains, with fewer LLM requests and function calls despite increased input tokens.
More from the AI-Native Software Development Capabilities in Jira trail
- A simpler way to run AI agents in Bitbucket Pipelines
- We’re bringing governed agent loops to the AI-Native SDLC
- How we made vulnerability fixes review-ready with Agentic Pipelines
- Agentic automation in practice: putting standard engineering work on autopilot
- Opening the door to agent autonomy: The architecture behind Rovo’s agent harness