Model parameters

Status: partial.

Use this page to find where model parameters are configured.

rtp-agent does not maintain a universal model parameter catalog. Parameters are split between app-level configuration, core option structs, and provider-specific adapter options.

App-level groups

Group	Examples
LLM	provider, model, base URL, extra headers, extra JSON body, fallback providers, response format, parallel tool calls
STT	provider, model, language, endpointing, timestamps, diarization, keywords, translation, base URL, provider-specific model options
TTS	provider, model, voice, language, speed, sample rate, text replacements, transforms, stream pacing, provider-specific JSON/model options
Realtime	realtime provider and model
Avatar	avatar provider and provider-specific IDs/options
VAD/turn detection	VAD provider/options and turn detector provider

Where to inspect exact fields

• app.AppConfig lists app-level fields.
• app.DefaultConfigFromEnv() maps environment variables to those fields.
• Adapter option types under adapter/<provider> define provider-specific constructor options.
• core/llm.ChatOptions, core/stt, and core/tts define shared core options.

Do not copy parameter names from LiveKit Python/Node docs unless the same option exists in Go source.

Evidence:

• app/app.go
• adapter option types under adapter/*
• core/llm/llm.go
• core/stt/stt.go
• core/tts/tts.go