Add Tencent Cloud Text Moderation System (TMS) as a fallback moderation
service when OpenAI is unavailable or returns an error. The service
now checks OpenAI first, and falls back to Tencent TMS if it fails,
providing better reliability for content moderation.
Features:
- New Tencent TMS configuration in config.yaml with environment variable overrides
- New tencent package in internal/pkg/tencent/ with TMS client implementation
- PostAIService now uses dual moderation with OpenAI primary and Tencent fallback
- Strict mode configuration passed to PostAIService for consistent behavior
BREAKING CHANGE: NewPostAIService now requires tencentClient and strictMode
parameters - update wire configuration accordingly.
- Removed Gorse-related configurations, handlers, and dependencies from the codebase.
- Introduced HotRank feature with configuration options for ranking posts based on recent activity.
- Updated application structure to support HotRank processing, including new caching mechanisms and database interactions.
- Cleaned up related DTOs and repository methods to reflect the removal of Gorse and the addition of HotRank functionality.
- Add Google Wire for compile-time dependency injection
- Replace concrete service types with interfaces across handlers
- Remove global state (DB, cache) in favor of constructor injection
- Split monolithic files into focused modules:
- config: separate files for each config domain
- dto: converters split by domain (user, post, message, group)
- cache: separate metrics.go and redis_cache.go
- Introduce unified apperrors package with string-based error codes
- Add transaction support with context-aware repository methods
- Upgrade golang.org/x/crypto from 0.17.0 to 0.26.0
