WorldwideChat v1.50: Redesigned for 26.1, Add Guardrails, New AI Options

Hello!
This is a major release focused on backend stability, AI reliability, and user-facing translation experience improvements.

Changes:

1. Added guardrail support for all AI providers. Essentially a 'guardrail' prompt will run before each request with your arbitrary rules. This model can be set to be separate from the main one.
2. Added OpenAI-compatible provider support alongside existing ChatGPT and Ollama paths. This is a third AI option meant for services such as vLLM/OpenRouter.
3. Enabled OpenAI Responses API defaults:
   • Endpoint: https://api.openai.com/v1/responses
   • Default ChatGPT model: gpt-5.3-chat-latest
4. Added optional Guidelines AI checks with configurable prompts and model selection.
5. Added the new loading/translation indicator experience (action bar based progress instead of chat, smoother spinner flow).
6. Added Paper Dialog API input support with configurable General.inputMethod:
   • auto
   • paper-dialog
   • conversation
   • none
7. Added translation controls for capacity and object throughput for better performance:
   • General.translationCapacityLimit
   • General.objectTranslationConcurrencyLimit
8. Added clearer feedback from translation failures and make them less vague.
9. Much better concurrency support for entity/world translation (meaning faster responses with sign/book translation.)
10. Expanded migration + status checks for SQL and MongoDB cleanup/deduplication and lookup performance.
11. Bumped runtime baseline to Java 21 and plugin compatibility docs to Spigot/Paper/Folia 1.20+. 1.19 and below are no longer supported.
12. Fixed bugs across /wwct stop, /wwcg, /wwcs, rate limiting, folia scheduling issues, and more
13. New testing server :)

As always take care!

---

Technical Notes (for those inclined, see per-commit for more details.)

• Added migration engine wiring:

  • StorageMigrationUtils and migration status model
  • Flyway-backed SQL/PostgreSQL migration path
  • Mongock-backed MongoDB migration path
  • New normalization migration for cache dedupe/index/key consistency
  • Migration failure handling disables plugin on unsafe state rather than continuing partial startup

• AI/prompt pipeline changes:

  • OpenAI-compatible API support (URL/API key/model/prompt integration)
  • Shared guardrail/prompt templates across providers
  • Better OpenAI response parsing for both strict JSON and failure paths
  • Optional Guidelines AI is now explicit and configurable instead of implicit behavior

• Input/interaction refactor:

  • Input abstraction for conversations/paper dialogs/null-input modes
  • General.inputMethod selection exposed in config + GUI paths
  • Paper dialog availability checks by version/classpath plus fallback strategy

• Loading/progress revamp details:

  • Action bar indicator now behaves as a proper lifecycle:
    • delayed start display
    • pulsing frame updates while active
    • safe clear path on completion
    • hard timeout guard and explicit error state behavior
  • Separate handling for object translation UX so long-running object workflows don’t stall or spam chat output

• Throughput and concurrency:

  • Added bounded translation capacity + object concurrency controls
  • Broader race-condition and queue fixes in translation/test flows
  • Threading and scheduler behavior improved for async/folia-safe execution paths

• Storage and startup:

  • Cache loading and sync logic tightened across YAML/SQL/Postgres/Mongo
  • Better startup task orchestration during reload to reduce partial-init races
  • More deterministic plugin bootstrap and background task cancellation/restart behavior

• Maintenance + hygiene:

  • Large dependency refresh across modules
  • GitHub workflow security/CI adjustments
  • Expanded test coverage for migration paths and platform-specific scheduler/performance scenarios