Latency issue in response

[jagiboy]

We are trying out Nem guardrails in one of our bots. We are seeing that there seems to be a large delay in our response. This is an issue for us because we are using it in a voice bot. Can anyone tell us why there seems to be a consistent 4 secs delay between these two actions (UtteranceBotActionFinished and StartUtteranceBotAction)?

22:45:43.808 | SetLanguageActionFinished | {"action_uid": "3767...", "action_name": "SetLanguageAction", "status": "success", "is_success": true, "return_value": "EN", "events": []}
22:45:43.830 | UtteranceBotActionFinished | {"action_uid": "7e2d...", "action_name": "UtteranceBotAction", "status": "success", "is_success": true, "return_value": null, "events": [], "final_script": "I
can help you with Application Status, Benefit Amount, Appointment Information, and other questions. How may I help you today?"}
22:45:47.493 | StartUtteranceBotAction | {"action_uid": "7e2d...", "script": "I can help you with Application Status, Benefit Amount, Appointment Information, and other questions. How may I help you
today?"}
22:45:47.957 | Total processing took 8.56 seconds. LLM Stats: 1 total calls, 0.41 total time, 416 total tokens, 412 total prompt tokens, 4 total completion tokens, [0.41] as latencies
INFO Raw rails response: [{"role": "assistant", "content": "I can help you with Application Status, Benefit Amount, Appointment Information, and other questions. How may I help you today?",
"tool_calls": [{"id": "33a0d52a-0a0f-4c55-95f0-2c3c6c46c196", "type": "function", "function": {"name": "TimerBotAction", "arguments": {"timer_name": "llm_response_polling", "duration": 1.0,
"action_info_modality": "time", "action_info_modality_policy": "parallel"}}}], "events": [{"type": "StopTimerBotAction", "uid": "a6c5c7fb-fccd-479a-90f9-df8c0677a9cc", "event_created_at":
"2025-03-20T02:45:43.788437+00:00", "source_uid": "NeMoGuardrails-Colang-2.x", "action_uid": "33a0d52a-0a0f-4c55-95f0-2c3c6c46c196", "action_info_modality": "time",
"action_info_modality_policy": "parallel"}]}]
Summary: 1 LLM call(s) took 0.41 seconds and used 416 tokens.

[Pouyanpi]

Hi @jagiboy thanks for sharing this. Would you please open a bug and complete the required steps so we can reproduce this issue? Thanks!