.env.example

# Realtime voice provider: openai (default), grok, or gemini. Blank means openai.
TALK_PROVIDER=

# xAI key for the grok provider. The Talk-scoped TALK_XAI_API_KEY wins when set.
# Leave both unset to ride the host's SuperGrok / X Premium+ login
# (`hermes auth add xai-oauth`). Set TALK_PREFER_XAI_OAUTH=true to require
# that login and refuse metered key fallback.
XAI_API_KEY=
TALK_PREFER_XAI_OAUTH=

# Gemini key for the gemini provider (free-tier keys work). The Talk-scoped
# TALK_GEMINI_API_KEY wins when set. The key rides the socket URL query on this
# lane — keep it out of logs.
GEMINI_API_KEY=
TALK_GEMINI_API_KEY=

# Gemini Live model/voice overrides. Voices are case-sensitive:
# Puck, Charon, Kore, Fenrir, Aoede.
TALK_GEMINI_MODEL=
TALK_GEMINI_VOICE=

# Custom-voice cascade: native (default; provider voices) or cascade (the
# provider thinks in text and ElevenLabs speaks — your own or a cloned voice).
# Cascade requires TALK_PROVIDER=openai and a TALK_ELEVENLABS_VOICE_ID.
TALK_VOICE_MODE=
TALK_CASCADE_TTS=

# ElevenLabs key for the cascade lane. The Talk-scoped TALK_ELEVENLABS_API_KEY
# wins when set. The key rides the xi-api-key header — never the URL, never a
# log line.
ELEVENLABS_API_KEY=
TALK_ELEVENLABS_API_KEY=

# Voice the cascade speaks with (VoiceLab -> your voice -> ID) and the TTS model.
TALK_ELEVENLABS_VOICE_ID=
TALK_ELEVENLABS_MODEL=