ha-conversation-jev GitHub avatar by luxus

luxus /

Voice & Conversation

ha-conversation-jev

A Home Assistant conversation integration routing simple lighting commands to services and other requests to Grok.

License: MITLanguage: PythonLast push: 9/18/2026#domain-workflows#voice-conversation#llm-routing-cost

What Jev Does in this Architecture

Jev classifies the utterance and exposed devices; code chooses a lighting service or conversation path.

Key Architecture Benefit

Provides a bounded lighting fast path rather than support for every smart-home command.

Review & Benchmark Note: Reviewed the fixed-version README and call path.
Clone Repository
git clone https://github.com/luxus/ha-conversation-jev.git

Inspect on GitHub

Check out the upstream repository README, issues, and commit log.

luxus/ha-conversation-jev

Architecture Category

Voice & Conversation

Voice turn-taking, interruption detection, and real-time dialogue routing.

View all in this category →

Related Projects in Voice & Conversation

jev-system-one repository icon by haseeb-heaven

haseeb-heaven

jev-system-one

2
Voice & ConversationMITTypeScript

A terminal Q&A app where OpenAI writes answers and Jev sets response policy and reviews drafts.

What Jev does here

Judges response mode, depth, uncertainty and draft quality inside a LangGraph workflow.

Displays the answer alongside a structured decision report.

#voice-conversation#evaluation-benchmarks#typed-decisions
2 evidence links
Inspect details
Voice & ConversationApache-2.0TypeScript

An optional AIAvatarKit component uses Jev to judge turn endings from speech transcripts.

What Jev does here

Assesses utterance completeness and whether the user is likely to continue speaking.

Adds a semantic signal to silence-based turn detection.

#voice-conversation#typed-decisions
663 evidence links
Inspect details
OpenWhisper repository icon by Knuckles92

Knuckles92

OpenWhisper

185
Voice & ConversationMITPython

A dictation and meeting-notes app with optional Jev checks for topic changes, note-taker instructions and sensitive text.

What Jev does here

Makes bounded transcript judgments for checkpoints, note-taking actions or screening before remote text cleanup.

Keeps model probabilities separate from local thresholds, switches and failure fallbacks.

#voice-conversation#security-guardrails#classification-ranking
254 evidence links
Inspect details