feat(release): v2.4.2 — task-aware routing, HuggingFace/Vertex providers, streaming fixes, token tracking, playground uploads

- feat: Task-Aware Smart Routing (T05) — auto-select model by task type
- feat: HuggingFace and Vertex AI provider support
- feat: Playground audio/image file uploads for transcription and vision
- feat: ModelSelectModal shows ✓ for already-added models (#180)
- fix: Claude Haiku routed to OpenAI without provider prefix (#73)
- fix: Token counts always 0 for Antigravity/Claude streaming (#74)
- fix: OpenAI SDK stream=False drops tool_calls (#302)
- fix: Media page generation errors — inline rendering for images/transcription
- fix: Round-robin state management for excluded accounts (#349)
- fix: Qwen user agent and CLI fingerprint compatibility (#352)
- deps: undici→7.24.2, dompurify→3.3.3, docker actions v4
- docs: CHANGELOG 2.4.2 with full feature/fix list
- docs: README with Task-Aware Routing table entry

docs/openapi.yaml

@@ -1,7 +1,7 @@
 openapi: 3.1.0
 info:
   title: OmniRoute API
-  version: 2.4.1
+  version: 2.4.2
   description: |
     OmniRoute is a local-first AI API proxy router. It provides an OpenAI-compatible
     endpoint that routes requests to multiple AI providers with load balancing,