params
Tuning options that can be passed in ai_sidecar.params of the ai_sidecar method.
All of these parameters are optional.
Properties
ai_sidecar.params
An object that accepts the following properties.
Reaction
Control how often the sidecar evaluates the conversation, how quickly it reacts to the customer, and whether tool actions take effect on the call.
params.idle_timeout_ms
How long the customer can be silent, in milliseconds, after they finish speaking before the sidecar evaluates the
conversation. Lower values make the sidecar react faster. If the agent speaks while the customer’s turn is still pending, the
sidecar evaluates immediately without waiting. Range: 50-5000.
params.min_interval_ms
The minimum time, in milliseconds, between evaluations — a throttle that limits how often the sidecar runs on a busy call.
Range: 0-60000.
params.max_iters_per_tick
The maximum number of model steps in a single evaluation, or in answering a question sent with the
calling.ai_sidecar.ask command. Each step is one model response, which can
call several tools at once. When the limit is reached the evaluation stops, whether or not the model has
produced its advice. Range: 1-20.
params.max_history_tokens
A soft token budget for the sidecar’s running conversation history. When the history grows past it, the oldest
complete exchanges are dropped, keeping the sidecar’s instructions at the start. Dropping stops once eight or fewer
messages follow the instructions, so the history can stay above the budget. An exchange is always dropped whole, so
fewer than eight messages can remain. Range: 1000-200000.
params.act_on_channel
Whether actions returned by your tools (such as transferring or hanging up the call) take effect on the call, or are only reported as callbacks.
Summaries
Generate summaries when the call ends, and set the default prompt for on-demand summaries.
params.final_summary
Whether to generate a closing summary of the sidecar’s session when the call ends. The result is included in the final callback.
params.ai_summary
Whether to generate an end-of-call summary of the conversation itself, distinct from final_summary (which summarizes the
sidecar’s session).
params.ai_summary_prompt
A custom prompt for the end-of-call conversation summary.
params.summary_model
The model used for the end-of-call conversation summary, distinct from model (the sidecar’s own model).
Suggested values: gpt-4o-mini, gpt-4.1-mini, gpt-4.1-nano.
params.transcribe_prompt
The default prompt for an on-demand summary
(action.summarize) that sets no prompt of its own. It
is not sent to the speech recognizer; use hints to help
it recognize specific terms. This is distinct from the operator prompt and from ai_summary_prompt, which is used
for the end-of-call summary.
Transcription
Configure speech recognition and per-utterance callbacks.
params.live_events
Whether to emit a callback for each utterance the speech recognizer produces.
params.verbose_utterances
Whether each utterance callback includes full speech-recognition detail, such as word timings and alternatives. This increases the callback size, so leave it off unless you need it.
params.speech_engine
The speech recognition engine to use. Possible values: deepgram, google.
params.speech_timeout
How long, in milliseconds, the recognizer waits before finalizing speech. When omitted or 0, 60000 is used.
Range: 0-600000.
params.vad_silence_ms
The amount of silence, in milliseconds, used to detect the end of speech. When omitted or 0, 700 is used with the
deepgram speech engine and 500 with google. Range: 0-60000.
params.vad_thresh
How sensitively the recognizer detects speech. When omitted or 0, 400 is used. Range: 0-10000.
Debug
Enable extra logging to help troubleshoot a sidecar.
params.debug_level
Speech-engine debug verbosity. Range: 0-100.
params.debug
Whether to enable verbose logging for the sidecar.