Revora v1.2.0 — Expanded AI Providers and Performance Improvements

Release Notes (v1.2.0)

This release focuses on expanding Revora's AI provider ecosystem, improving model discovery and integration reliability, fixing bugs across the application, and optimizing the review execution pipeline.

New Features

Expanded AI Provider Integrations

Revora continues to expand its Bring Your Own Key (BYOK) architecture, giving users greater flexibility in choosing AI providers and models for repository-aware code reviews.

Groq Integration: Added support for Groq, expanding the available inference providers and enabling users to configure compatible Groq models using their own API keys.

Expanded Model Availability: Continued improvements to AI model discovery and provider configuration, making it easier to identify and configure supported models.

Ongoing Provider Expansion: Additional AI provider integrations and model support are under development, with further improvements planned for future releases.

Bug Fixes and Improvements

• Fixed bugs affecting application functionality and user experience.
• Improved reliability across provider configuration and AI review workflows.
• Refined error handling for provider and model-related failures.
• Improved consistency across existing AI provider integrations.

Performance Optimizations

• Optimized selected backend workflows to improve execution efficiency.
• Refined review processing and resource utilization.
• Improved application stability and reliability during AI-powered review execution.

Security and Stability

• Continued improvements to secure provider configuration and API key handling.
• Strengthened reliability across AI request processing and error handling.
• Preserved compatibility with existing providers and review workflows.

What's Next

Revora is continuing to build a broader AI provider ecosystem, with additional providers, expanded model availability, and further improvements to the repository intelligence and code review pipeline.

The Model Lab is also under development as part of Revora's longer-term roadmap, with the goal of bringing more AI capabilities and project-specific models into the platform.

Thank you to everyone contributing to Revora and helping improve open-source AI-powered code review.

What's Changed

• Feature/ollama cloud integration by @d-kavinraja in #53
• feat(seo): improve Revora search discoverability by @d-kavinraja in #54
• fix: reuse single Review card on PR reopen with new ReviewExecution by @d-kavinraja in #55
• Fix-Queue and Sync by @d-kavinraja in #56
• UI Update and Bug Fix by @sureshkumar2704 in #57
• Feature/revora v1 enhancements by @d-kavinraja in #58
• Feature/platform UI enhancements by @d-kavinraja in #59
• Feature/groq provider by @d-kavinraja in #60

New Contributors

• @sureshkumar2704 made their first contribution in #57

Full Changelog: v1.1.0...v1.2.0

What's Changed

• Feature/ollama cloud integration by @d-kavinraja in #53
• feat(seo): improve Revora search discoverability by @d-kavinraja in #54
• fix: reuse single Review card on PR reopen with new ReviewExecution by @d-kavinraja in #55
• Fix-Queue and Sync by @d-kavinraja in #56
• UI Update and Bug Fix by @sureshkumar2704 in #57
• Feature/revora v1 enhancements by @d-kavinraja in #58
• Feature/platform UI enhancements by @d-kavinraja in #59
• Feature/groq provider by @d-kavinraja in #60

New Contributors

• @sureshkumar2704 made their first contribution in #57

Full Changelog: v1.1.0...v1.2.0

Release Notes (v1.1.0)

This release focuses heavily on expanding AI provider availability, overhauling the review execution lifecycle, enhancing the frontend UI/UX, and significantly optimizing backend memory usage for single-service deployments.

🚀 New Features

Expanded AI Provider Integrations

We've significantly expanded the Bring Your Own Key (BYOK) architecture to natively support more providers:

• Cohere Integration: Full support for Cohere's latest models.
• Ollama Cloud Integration: Support for routing requests to Ollama Cloud.
• OpenRouter Integration: Access to dozens of community models via OpenRouter.
• Model Discovery Engine: Dynamically fetches and discovers available models.

Review Execution & Lifecycle

• Repository Sync Engine: Hardened the repository syncing and review queue lifecycle.
• Precise Execution Tracking: Granular tracking of review states, giving real-time feedback on queue position, execution time, and potential delays.
• Persistent Review State: The frontend now properly maintains persistent state across review details UX.

Frontend & UI Enhancements

• Profile Page & Dashboard Refinements: New Profile page and improved dashboard layouts and color schemes.
• Landing Page Polish: Substantial UI enhancements to the public-facing landing page.
• Accessibility Fixes: Improved text visibility specifically in light mode, and removed several UI anti-patterns.
• Updated Documentation: Completely overhauled the README.md to reflect the actual pipeline architecture, including the latest provider availability statuses.

🛡️ Security & Stability

• Security Mitigations: Applied security mitigations and refinements across the application.
• Queue Concurrency: Resolved review generation failures, potential JSON leaks, and race conditions within the background queue concurrency.
• 410 EOL Error Handling: Added clean and explicit handling for models that have reached End Of Life.

⚡ Performance Optimizations

• Memory Usage Reduction: Optimized memory consumption specifically for single-service deployments.
• Embedded Queue Worker: The background queue worker is now cleanly embedded into the FastAPI lifespan, drastically reducing baseline memory overhead.
• Unblocked Uvicorn Startup: Moved startup database syncing to a background task, ensuring the API is instantly available.
• Thread Leak Fixes: Resolved an async thread timeout leak affecting dashboard statistics.

🧹 Maintenance & Chore

• Extensive backend formatting using black and ruff.
• Fixed multiple broken tests (test_api_keys.py, prompt builder mock case sensitivity).
• Added missing Alembic migrations for database timestamps.

Revora - Initial Release

We are excited to announce the first official release of Revora!

Revora is an advanced, open-source AI Pull Request Reviewer designed to act as a deeply integrated team member. Built from the ground up for performance and accuracy, Revora automatically analyzes pull requests, understands repository architecture, and provides actionable code feedback directly within the GitHub UI.

Key Features

Repository Intelligence Engine

Revora goes beyond surface-level code scanning. Our custom Intelligence Engine performs deterministic analysis of your repository before involving the LLM, automatically detecting:

• Programming languages and frameworks.
• System architecture patterns.
• Database and ORM technologies.
• CI/CD pipelines, package managers, and testing frameworks.
• Security layers and cloud providers.

Advanced Context Retrieval

Instead of just sending the raw diff to an LLM, Revora intelligently maps the changes against your entire repository context:

• AST (Abstract Syntax Tree) parsing using Tree-sitter.
• Dependency graph resolution to understand how changes impact other files.
• Intelligent code chunking and vector retrieval to feed the AI highly relevant context.

Bring Your Own Key (BYOK)

You are in complete control of your data and AI providers. Revora supports multiple LLM orchestrators out of the box:

• OpenAI Models (GPT-4o)
• Anthropic Models (Claude 3.5 Sonnet)
• NVIDIA NIM Integration

PostgreSQL-Native Job Queue

Designed for lightweight deployments, Revora operates without the need for external brokers like Redis or RabbitMQ. Our highly concurrent, asynchronous queue runs natively on PostgreSQL using SKIP LOCKED mechanics, allowing for isolated API and Worker processing on minimal infrastructure.

GitHub App Native

• Zero Config Setup: Install the GitHub App on your organization or repository, and Revora automatically listens for PR webhooks.
• Inline Comments: Reviews are posted directly to the lines of code in the PR diff.
• Check Runs: Seamlessly integrates with GitHub Actions / Check Runs to block or pass PRs based on the review outcome.

Architecture Highlights

• Backend: Built with FastAPI, SQLAlchemy, and asyncpg.
• Worker: decoupled Python-based queue consumer designed to run efficiently on low-memory environments (like Render Free tier).
• Database: PostgreSQL.
• Frontend: Next.js deployment.

Getting Started

To self-host Revora or contribute to the project, please refer to our Installation Guide in the README. The backend can be easily deployed via Render or any Docker-compatible environment, while the frontend is optimized for Vercel.

What's Next

This release establishes the core AI orchestrator and review pipeline. Future updates will focus on:

• Multi-file context improvements.
• Custom user prompts and review rubrics.
• Extended language support for AST parsing.

Thank you to all contributors who helped make this initial release possible!