0.1.37

[0.1.37] - 2026-02-20

Fixed

• Improved conversation continuity by sending session_id in chat completion JSON payloads (in addition to X-Session-Id header), for both regular and streaming requests.
• Reduces cases where the gateway treats each message as a new conversation when custom headers are ignored upstream.

Full Changelog: 0.1.36...0.1.37

0.1.36

[0.1.36] - 2026-02-20

Fixed

• Voice mode now auto-falls back to browser speech when voice_provider is assist_stt, instead of blocking continuous mode with an error message.
• Reduced duplicated assistant replies in the chat card by deduplicating repeated openclaw_message_received payloads.

Full Changelog: 0.1.35...0.1.36

0.1.35

[0.1.35] - 2026-02-20

Fixed

• Improved chat-card reliability when using voice send flow by re-subscribing to openclaw_message_received after card reconnects.
• Added backend history-sync fallback after message send so user/assistant messages still appear when an event is missed.

Full Changelog: 0.1.34...0.1.35

0.1.34

[0.1.34] - 2026-02-20

Changed

• Assist STT microphone capture now uses AudioWorkletNode when available, with automatic fallback to ScriptProcessorNode for older browsers.
• Reduced browser deprecation noise by avoiding ScriptProcessorNode on modern browser engines.

Documentation

• Expanded README voice documentation with practical guidance for voice_provider (browser vs assist_stt) and provider-specific troubleshooting.

[0.1.33] - 2026-02-20

Fixed

• Reduced 415 Unsupported Media Type failures for assist_stt by fetching STT provider capabilities and negotiating metadata before upload.
• Assist STT now auto-matches provider-supported language values (for example bg vs bg-BG) when submitting transcription audio.
• Assist STT now aligns upload metadata sample rate/channels with provider-supported values when available.

Full Changelog: 0.1.32...0.1.34

0.1.32

[0.1.32] - 2026-02-20

Added

• Added configurable voice input provider option: browser or assist_stt.
• Chat card now supports Home Assistant STT transcription mode (assist_stt) for manual mic input.

Changed

• Voice provider is now exposed through integration settings websocket payload and card configuration handling.
• Continuous voice mode remains available only for browser voice provider.

Full Changelog: 0.1.31...0.1.32

0.1.31

[0.1.31] - 2026-02-20

Fixed

• Fixed Assist pipeline crash during intent recognition on some Home Assistant versions:
  • Replaced conversation.IntentResponse / conversation.IntentResponseErrorCode usage with homeassistant.helpers.intent equivalents in the OpenClaw conversation agent.
  • Resolves AttributeError: module 'homeassistant.components.conversation' has no attribute 'IntentResponse'.

Full Changelog: 0.1.30...0.1.31

0.1.30

[0.1.30] - 2026-02-20

Fixed

• Treated SpeechRecognition no-speech as a normal listening condition instead of an error.
• Reduced voice error noise by avoiding retry scheduling for no-speech events.
• Added clearer in-card status text for silence/no-speech scenarios.

[0.1.29] - 2026-02-20

Fixed

• Voice language selection now prioritizes the preferred Assist pipeline language (assist_pipeline/pipeline/list) instead of only using Home Assistant UI language.
• Added separate TTS language resolution so spoken replies follow Assist pipeline TTS language when available.
• Retained safe fallbacks to integration/UI/browser language when Assist pipeline data is unavailable.

Full Changelog: 0.1.28...0.1.30

0.1.28

[0.1.28] - 2026-02-20

Fixed

• Treat SpeechRecognition aborted events as expected stop behavior (no error status/no noisy console error) when voice is intentionally stopped.
• Added a stop-request guard to avoid restart/error churn during recognition shutdown.
• Synchronized release versioning so manifest, frontend loader URL, and backend card resource URL all use the same cache-busting version.

[0.1.27] - 2026-02-20

Changed

• Improved backward compatibility for older Home Assistant Core builds by removing Python 3.12-only type alias syntax in integration runtime code.
• Added fallback import handling for ConfigFlowResult in config flow type hints.

Full Changelog: 0.1.26...0.1.28

0.1.26

[0.1.26] - 2026-02-20

Added

• Added integration option allow_brave_webspeech to Settings → Devices & Services → OpenClaw → Configure.
• Frontend card now reads this option via openclaw/get_settings and applies it automatically.

Changed

• Card version bumped to 0.2.6 and cache-busting URL updated to v=0.1.26.

[0.1.25] - 2026-02-20

Fixed

• Voice input language now prioritizes integration/HA locale settings more reliably (including frontend locale fallback), reducing unwanted fallback to English.
• Voice-mode assistant replies now use improved speech synthesis voice selection for the active language and better voice-loading handling.
• Reworked chat pending-response tracking to support multiple in-flight messages without leaving stuck typing indicators.

Full Changelog: 0.1.24...0.1.26

0.1.24

Full Changelog: 0.1.23...0.1.24