Clients/testrr/bot

bot

testrr1-bot · ● live · v

ModelLLM
VoiceAudio
TranscriberAudio
PromptBrain
Turn-takingTiming
Call limitsTiming
TelephonyRouting
ToolsActions
AnalysisData

Model

The language model that powers the conversation.

Which LLM runs the conversation. Requesty and OpenRouter are pools (no per-account rate cap); Azure can be EU-pinned.
LLM_PROVIDER
The model id on Requesty. Must support tools if this bot uses them.
Lower is more consistent and on-script; higher is more varied. Phone support usually wants low (0.3–0.5).
LLM_TEMPERATURE
0.40
Cap on how long a single reply can be (sent as max_completion_tokens). Lower keeps the bot crisp on a phone call.
LLM_MAX_TOKENS
tokens
Conversation memory

How the bot remembers what it said — which matters most when the caller interrupts.

When the bot is interrupted mid-sentence, it generated a full reply but the caller only heard the first part. OFF (recommended) means the bot's memory holds only what was actually SPOKEN — so it never thinks it told the caller something they never heard. ON means it remembers the full text it generated, even the unspoken part.
modelOutputInMessagesEnabled
On
Pass detected caller emotion to the model so it can adapt its tone.
emotionRecognitionEnabled
Off
All changes saved