Skip to content

Overview

This guide explains how to set up and operate MiaRec Conversation Analytics in a multi-tenant deployment.

It is written for platform operators (service providers / partners) who manage:

  • tenant lifecycle (create/configure/deprovision tenants)
  • system-wide ingestion and transcription configuration
  • AI engines (LLM providers/models)
  • global Custom Fields and AI Tasks
  • the AI Assistant job (continuous processing pipeline)
  • monitoring, troubleshooting, upgrades, and governance

If you are an organization (tenant) administrator, use the Conversation Analytics – Administration Guide instead.
If you are an end user (supervisor/analyst/agent), use the Conversation Analytics – User Guide.

What Conversation Analytics does (operator view)

At a high level, the platform:

  1. Ingests conversations (voice calls today; omni-channel text sources such as chat/email/tickets as you enable them).
  2. Produces a Transcript (for voice calls) or a normalized Text Thread (for text channels).
  3. Runs tenant-enabled AI Tasks (LLM prompts + output mapping + filters) via the AI Assistant job.
  4. Stores outputs into Custom Fields, which power dashboards, search filters, and conversation details.

Key operator concept: AI Tasks are executed per-tenant based on tenant activation and filters. You manage the global defaults and the underlying processing pipeline.

How this guide is organized

  • Deployment Models and Responsibilities – SaaS vs Partner-hosted, and who owns which settings.
  • Platform Setup – prerequisites, tenant lifecycle, ingestion, transcription, AI engines, global tasks/fields, AI Assistant job.
  • Operations – monitoring, cost controls, security & governance, upgrades.
  • Troubleshooting – common problems and runbooks.
  • Reference – naming standards, prompt/schema standards, default filter patterns.

Fast-start checklist (operator)

Use this as a sanity path when bringing up a new environment:

  1. ✅ Validate infrastructure prerequisites (databases, storage, queues/workers, secrets).
  2. ✅ Configure channel ingestion (voice and/or text sources).
  3. ✅ Configure transcription for voice calls and verify transcripts are produced.
  4. ✅ Configure one or more AI Engines (LLM providers/models).
  5. ✅ Create/publish baseline global Custom Fields (CX + Sales + Ops).
  6. ✅ Create/publish baseline global AI Tasks (CSAT, summarization, sentiment, etc.), default-disabled.
  7. ✅ Configure and start the AI Assistant job.
  8. ✅ Create a test tenant → enable a couple AI Tasks → run a smoke test.
  9. ✅ Set up monitoring, alerting, and a change-management process.

Definitions used in this guide

  • Platform operator / system admin: manages global settings and multiple tenants.
  • Tenant admin: administers one tenant (enables tasks, overrides prompt/filters, dashboards).
  • Custom Field: typed storage for an insight (number/text/date/dropdown).
  • AI Task: an analysis definition (prompt + mapping + optional filters + AI engine selection).
  • AI Assistant job: continuous worker/job that runs enabled tasks and writes results.