AI coding is fast, but breaks down when teams need review, context, and control - especially on large, long-lived codebases.
With Kavia v1.0.9, we focused on tightening the core workflow so teams can move faster without losing context or governance.
What’s new
Interactive Agent Mode
Code generation is now fully conversational, with a proper review loop. Teams can generate changes, inspect diffs before anything is applied, refine iteratively, and approve with confidence—no blind edits.
Simpler SCM onboarding
Projects can now be onboarded via Bitbucket OAuth, with foundational support in place for self-hosted GitLab environments. The goal is to reduce setup friction without compromising enterprise constraints.
Persistent maintenance context (Kdiff with tags)
Change context across branches and releases is now queryable using tags. This makes it easier to understand why something changed, not just what changed—especially during long-term maintenance.
Workflow and UX improvements
Manifest generation, chat responsiveness, editor loading states, and file ingestion visibility have all been refined to provide clearer feedback throughout the setup and development process.
Why this matters
As codebases grow, speed alone isn’t enough. Teams need workflows that preserve context, enable review, and scale responsibly across repos and releases.
This release is a step toward making AI-assisted development usable on real, production codebases, not just greenfield demos.
If you’re interested in exploring the workflow, it’s live at chat.kavia.ai.
Happy to answer questions or dive deeper into any of the features.
How are you handling diff review + context when using AI on production repos?
Top comments (0)