Highlights of main changes since 0.2.4:
- AI Providers & Ollama Integration:
- Added Ollama AI provider support with configuration options for
model context size, thinking mode, and JSON response formatting.
- Added comprehensive error handling mapping Ollama API error codes
and transient errors.
- Fixed reviewer concurrency handling when concurrency is set to 1.
- Webhook Security & Endpoint Hardening:
- Implemented HMAC signature verification (SHA256/SHA1) for GitHub
and GitLab webhook endpoints.
- Added input validation and payload body size limits for webhook
endpoints.
- Added security guide and design document for webhook input
hardening.
- Patch, Email, & NNTP Processing:
- Preserved and injected patch changelogs below the separator
into review patch context.
- Handled quoted and unquoted author names with special characters to
ensure proper email delivery.
- Fixed NNTP article ingestion to prevent premature skipping of tip
articles due to server replication lag.
- Updated mailing list tracking and email policies for linux-mtd,
linux-clk, linux-media, media-ci, and linux-s390.
- Git Operations & Baseline Resolution:
- Improved baseline commit resolution to use mainline
remote-tracking refs instead of HEAD.
- Added automatic fetching of missing base commits and tags from
mainline remote.
- Made git status checks independent of user git configuration.
- Web UI & Diagnostics:
- Preserved LLM interaction logs and metrics in review worker
outputs.
- Fixed copy to clipboard for inline review comments in Web UI.
- Restored default about dialog content when project description is
empty and safely updated dynamic layout titles.
- Fixed UI progress bar capacity overflow.
- Subsystem Prompts & Documentation:
- Added subsystem guides for fuse, hwmon (target directory
guidance), and i2c (struct i2c_device_id initialization).
- Updated prompt guidelines for io_uring and media subsystems.
- Added Docker, Cloud Build, and K8s configuration fixes, updated
press coverage, and updated dependencies.
