v0.9.0

What's Changed

• fix: render transcript text safely in browser extensions by @nightcityblade in #435
• CI: expand test suite coverage by @boxerab in #436
• audio: add support for raw pcm input via server flag by @boxerab in #437
• Add thread safety to client manager with threading lock by @boxerab in #438
• Bound transcript memory and translation queue size by @boxerab in #439
• Validate server parameters on startup by @boxerab in #440
• Avoid clear screen shell injection by using ANSI escape codes by @boxerab in #441
• Add custom vocabulary / hotwords support by @boxerab in #489
• feat: add Prometheus metrics instrumentation by @boxerab in #492
• Add real-time speaker diarization support by @boxerab in #490
• Document all new features in README by @boxerab in #491
• Add word-level timestamps and confidence scores by @boxerab in #488
• feat: add segment_post_processor hook for external plugins by @boxerab in #487
• feat: add SSE streaming for REST transcription endpoint by @boxerab in #452
• docs: fix setup instructions reported in #405 by @nightcityblade in #434
• Clarify --translate vs --enable_translation CLI flags by @boxerab in #442
• Extract hardcoded buffer constants into class attributes by @boxerab in #443
• Add reconnect logic to WebSocket client by @boxerab in #444
• Improve REST API unsupported param warnings by @boxerab in #445
• Add optional API key auth and rate limiting for REST API by @boxerab in #446
• Add WebSocket authentication via api_key by @boxerab in #449
• fix: add av to package requirements by @nightcityblade in #497
• fix: make caption overlay line count configurable by @nightcityblade in #494
• fix: support Python 3.13 server installs by @nightcityblade in #484
• test: add clean virtualenv install smoke test by @nightcityblade in #482

Full Changelog: v0.8.0...v0.9.0