Release Notes - July 2025
Smarter reviews, new AI commands, and stronger security for gitStream.
More accurate AI reviews with better context
We've enhanced Retrieval-Augmented-Generation (RAG) on-the-fly to detect issues when code calls functions defined outside the current PR context. This helps identify:
- Incorrect argument types in function calls
- Mismatched number of arguments
- Missing
await
for async functions - Issues in JavaScript, Python, and other dynamic languages
PR reviews now include project ticket details to instantly identify mismatches between requirements and implementation. Here's what's new:
- Automatically detect ticket keys in PRs
- Pull relevant details from your project management tool
- Flag scope gaps early in the process review
Trigger AI actions with commands - no config file needed
You can now run AI actions on demand by adding straightforward commands to your PR comments without needing a config file in the repo. Use commands such as:
/gs review
to trigger an AI code review/gs desc
to generate a PR description/gs help
to see supported actions
This is currently available for GitHub, with GitLab and Bitbucket support coming soon. Read the docs.
More secure with built-in AI security guardrails
AI features now validate every prompt before processing. Guardrails, powered by Prompt-Guard-66M, block jailbreak attempts and protect against malicious prompt injections, keeping your systems more secure without slowing you down.
Improved AI for generating PR descriptions
AI-generated PR descriptions are now powered by Sonnet 4 to improve context and clarity. We've also migrated the PR description functionality to the pr_agent
service to improve performance and maintainability.
Support for PRs with large files
The maximum file size for PRs has increased from 1MB to 5MB, and compression functionality is now supported to process PRs with larger files. When a PR exceeds the processing limit, you’ll also see clearer, actionable guidance on best practices to help keep reviews moving smoothly.
Metrics for AI code reviews
A new AI Code Review Metrics Dashboard gives you complete visibility into how AI performs across your teams, pulling in data from PRs monitored by gitStream. Track key metrics such as:
- Number of reviewed PRs and total reviews to measure adoption
- AI-generated suggestions and lines accepted to assess quality and trust
- Issues detected across security, performance, maintainability, and readability
Filter by repos, contributor, or time range and share insights that matter most. Read the docs on metrics and findings.

Real-time system status
You can now monitor gitStream system statuses across GitHub, GitLab, and Bitbucket in real time. Check out linearbstatus.com.

Sunset: Resource Allocation emails
To reduce noise in your inbox, we've stopped sending email summaries on Resource Allocation. You'll still find everything you need in your LinearB dashboard, which has real-time data and filtering capabilities.

Archived release notes
View our previous release notes.
How did we do?
Archived Release Notes