Release v0.1.6

Summary of Changes

Release v0.1.6 introduces full support for OpenAI-compatible Image Generations, major dashboard UI redesigns, modular API key and quota management, agent memory CLI integrations, streaming responsiveness improvements, and a completely modularized fallback policy cascade engine.

🚀 New Features & Enhancements

• Image Generation Support (/v1/images/generations):
  • Implemented OpenAI-compatible image generation endpoint (/v1/images/generations) supporting prompt, n, size, and response format (#108).
  • Enforced modality verification via models.jsonc output types (requiring explicit "image" capability).
  • Added Neosantara image generation models (sdxl-lightning, agnes-image-2.*-flash).
• Dashboard Operations Redesign:
  • Sleek, modern card-based devtools redesign for gateway operations dashboard (#113).
  • Re-architected API key modals with responsive mobile layout and specialized model selector dialog (#113).
  • Modularized API key dialogs, quota table layout, and quota query caching (#107).
• CLI Adapters:
  • Added Hindsight CLI adapter for agent memory integration (#114).
  • Fixed Claude Code configuration resolution across multiple candidate directory paths (#115).
• Catalog & Maintenance:
  • Pruned dead and unsupported OpenCode Zen free models (#116).
  • Streamlined model catalog metadata in @srouter/constants.

🛠️ Refactoring & Architectural Hardening

• Modular Fallback Policy & Cascade Engine:
  • Extracted status code parsing and error matching to apps/api/src/logic/fallback.policy.ts (#118).
  • Unified candidate tracking, follow-up search message construction, and error logging across streaming and non-streaming cascade loops in ChatLogic (#118).
  • Shared fallback logic with image generation (ImagesLogic).
• Database & Isolation:
  • Fully isolated test databases from the production database during test runner execution (#110).

🐛 Bug Fixes

• CodeBuddy Upstream Compatibility:
  • Enforced response_format via user-message schema injection (#117) by @pandamoon21.
  • Preserved system prompts and provided fallback to reasoning_content on empty deltas (#117).
• Streaming & Interception:
  • Restored tool call interception logic and missing imports in ChatLogic (#106).
  • Immediately streamed SSE chunks without proxy buffering by setting appropriate downstream headers (#104).

---

🔄 Pull Requests Included

PR	Title	Author
#118	refactor(api): modularize fallback policy and streamline cascade execution	@seaavey
#117	fix(codebuddy): enforce response_format via user-message schema injection	@pandamoon21
#116	chore(constants): prune dead and unsupported opencode zen free models	@seaavey
#115	fix(cli): resolve claude configuration across multiple candidate paths	@seaavey
#114	feat(cli): add hindsight adapter for agent memory integration	@seaavey
#113	feat(web): redesign dashboard with modern sleek devtools style	@seaavey
#111	chore: remove image generation spec and plan docs	@seaavey
#110	fix(db): isolate test databases from production database	@seaavey
#108	feat(api): add image generation endpoint with models.jsonc modality validation	@seaavey
#107	refactor(web): modularize API key dialogs, quota page, and add quota caching	@seaavey
#106	fix(api): restore tool call interception logic and missing imports in ChatLogic	@seaavey
#104	fix(api): stream SSE chunks immediately without buffering	@seaavey

---

👥 Contributors

• @seaavey (Muhammad Adriansyah)
• @pandamoon21 (PR #117)

Full Changelog: v0.1.5...v0.1.6