36 docs tagged with "voice"

Provision a new church on the multi-tenant voice agent — insert the church_voice_agents DB row with config, no redeploy needed; the agent routes automatically by phone number

ChurchWiseAI brand voice, tone, visual identity, color palette, typography, messaging hierarchy, and rules for how the brand speaks across all properties

How the Coordinator agent detects pastoral/emotional signals and hands off to the Care Agent (Claude Haiku 4.5) — trigger conditions, handoff mechanics, and why two agents instead of four

TTS for voice agent calls (Cartesia Sonic) — Cartesia provides the text-to-speech voice synthesis used during live phone calls via livekit-plugins-cartesia

Diagnose voice call problems — wrong responses, routing failures, tools not executing, audio quality issues — using Railway logs, LiveKit Cloud dashboard, and voice_call_logs DB records

Full deploy cycle for voice agent code — branch, implement, test locally, push to Railway via git; LiveKit Cloud connects automatically, no manual reconnect step needed

Review calls where the voice agent couldn't complete the caller's need — query voice_callback_requests, notify church admins, and ensure follow-up actions are taken

End-to-end explanation of the ChurchWiseAI voice agent — body analogy, service map, call flow, and fallback chain

Complete lifecycle of an inbound church phone call — Twilio SIP trunk, LiveKit Cloud SIP gateway, Railway agent worker, church routing by phone number, agent selection, and post-call DB writes

SIP gateway and room management for the ChurchWiseAI voice agent — LiveKit Cloud provides the telephony bridge between Twilio SIP trunks and the Python agent worker on Railway

Respond to a LiveKit Cloud or Cartesia TTS outage that makes the voice agent unavailable — detection, distinguishing the failure point, customer notification, and recovery steps

Status: DECIDED

Review and action prayer requests from voice_prayer_requests (shared by voice + chatbot) — triage by urgency, notify church pastoral staff, and log follow-up completion

Everything that happens from "pastor hears about ChurchWiseAI" through "visitor sends first message or call is answered." Covers the church admin setup journey (Track A) and the visitor discovery journey (Track B), and traces how admin-entered config flows into every AI response.

Buy a local Twilio number for a church, configure forwarding to the Cartesia voice agent, and create the church_voice_agents row to make the church live on the agent

Supabase SQL queries for product metrics — chatbot conversation counts, voice call volume, ITW page views, premium church counts, and cross-property revenue summary

What a Starter Voice customer sees at every touchpoint — discovery through working product

Twilio phone numbers and SMS used for voice call forwarding to LiveKit Cloud, callback request notifications, ops SMS alerts, and PewSearch lead notifications

Status

Multi-tenant LiveKit Agents SDK architecture — one deployed agent serves all churches via SIP trunk phone-number routing, with per-church config loaded from Supabase at call init

Non-technical walkthrough of a complete voice call — from phone ring through greeting, RAG lookup, tool use, potential Care Agent handoff, and post-call DB writes

Curated sources for LiveKit Agents SDK and Telnyx SIP — equip subagents with these when dispatching voice work

Resilience and fallback design for the voice agent — LLM, TTS, STT provider fallbacks, routing fallbacks, cache stale recovery, and the planned Telnyx migration from Twilio

Product overview for the ChurchWiseAI Voice Agent — AI phone receptionist for churches, built on LiveKit Agents SDK, serving all plans from $39.95/mo with Gemini 2.5 Flash as primary LLM

How voice agent system prompts are built from pure functions in prompts.py — church data injection, theological lens application, HEAR protocol, and runtime context assembly

Step-by-step guide to provisioning a new church on the voice agent — church_voice_agents DB row, Twilio number setup, demo church configuration, and what is automated vs manual

How the voice agent handles provider failures (LLM, TTS, STT) and crash recovery

All voice agent tools decorated with @function_tool — prayer request capture, callback scheduling, visitor contact, event registration, and the agent handoff tool

Known voice agent issues, debugging procedures, and resolution steps — LiveKit/SIP routing errors, wrong church routing, tool failures, audio quality issues

Known issues, root causes, and fixes for LiveKit SIP + Telnyx voice infrastructure. Includes GitHub issue references, auth strategy, and operational rules.

Systematically review voice call transcripts for a church — identify agent performance gaps, off-topic responses, missed tool uses, and implement improvements via prompt tuning

Step-by-step guide to provision a new customer's voice agent phone number using Telnyx + LiveKit Cloud

Index of voice ops runbooks — add churches, debug call issues, deploy code changes, handle missed calls, review prayer requests, audit call quality, and respond to platform outages

Definitive step-by-step trace of every function, LLM call, STT/TTS step, database query, and third-party service call from SIP arrival to call termination. Required reading before touching any voice code.

What broke

Step-by-step pipeline for each voice turn — noise filtering, moderation check, RAG retrieval, LLM inference, tool execution, and TTS output — implemented via on_user_turn_completed callbacks and call_handler.py utilities