PulseChat – OmniChannel AI Customer Service module for Perfex CRM
PulseChat OmniChannel AI Customer Service Module Overview & Functional Engine
PulseChat is a CodeIgniter-based module for Perfex CRM, engineered for agencies and in-house operations teams to consolidate WhatsApp, Telegram, Email, SMS, Slack and Viber conversations into one inbox with AI-assisted replies. The core bottleneck it resolves is context fragmentation: agents traditionally jump between browser tabs, shared mailboxes and phone apps while the CRM holds the client history in a separate silo, so identities and threads stay disconnected. In a production workflow the module installs as a first-class Perfex extension, inheriting the host CRM’s permission model, cron scheduler and client/lead records, so every external conversation is stored, routed and answered without leaving the existing infrastructure.
Technical Specifications & Deployment Matrix
| Engineering Metric | Deployment Specification |
|---|---|
| Base Technology / Framework | CodeIgniter (native Perfex CRM module architecture) |
| Runtime / Language Requirements | PHP 8.x (including PHP 8.5 support) and PHP 7.x; MySQL 8.x or MySQL 5.x |
| Communication Protocols | Webhook listeners for WhatsApp/Meta, Facebook Pages, Instagram, Telegram, Slack and Viber; IMAP/SMTP polling for email channels |
| AI Provider Integration | OpenAI (ChatGPT, gpt-4o, gpt-4o-mini) and Anthropic (Claude 3.5) with configurable temperature, max tokens and system prompts |
| Files Included | JavaScript JS, JavaScript JSON, HTML, CSS, PHP, SQL |
| Database / Persistence Footprint | Module-scoped tables applied through guarded migrations that run automatically on upgrade without touching existing rows |
| Admin Configuration Surface | Dedicated settings area for channel enablement, AI provider selection, automation rules and capability assignments |
Omnichannel Messaging Engine & Execution Pipeline
Inbound messages arrive through channel webhooks and IMAP polling, get normalized into unified conversations, then matched to Perfex clients, contacts or leads.
Architectural & Functional Highlights
- Multi-Channel Adapter Layer: Each external network (WhatsApp Business number, Telegram bot, IMAP mailbox, Facebook Page, Instagram account, SMS, Slack, Viber) is configured as an adapter in the Channels tab. Since version 1.3.6, several WhatsApp numbers, Instagram accounts and Facebook Pages can operate under a single Meta App, with inbound routing matched to the exact receiving number or Page so replies always originate from the correct identity.
- Dual-Provider AI Assistant: The engine calls either OpenAI or Anthropic endpoints per configured model. Agents trigger a one-click Draft Reply from recent thread context, receive three ready-to-send suggestion chips, or run rewrite operations (professional, friendly, concise), grammar cleanup, thread summarization, sentiment analysis and on-the-fly translation. Every call is logged with model, token count, estimated cost and duration at the conversation level.
- Automation Rules Engine: Conditional rules evaluate inbound traffic against channel, subject content, customer attributes and detected language, then execute actions such as assigning a staff member or team, setting status and priority, applying tags for filtering and reporting, triggering AI auto-replies for low-risk queries, or creating a Perfex ticket directly from a conversation.
- Identity Unification & CRM Context: External contacts are automatically mapped to existing client, contact or lead records where possible. The side panel surfaces name, company, role, open tickets, invoices and prior omnichannel history in a single timeline, eliminating manual lookups during live conversations.
- Internal Collaboration Layer: Beyond external channels, the module provides staff-to-staff direct messages, staff-to-client portal messaging and group conversations, with filters for status, priority, channel type and tags, pinned threads, unread counters and canned responses used alongside AI drafts.
For operations running WordPress as a separate site and evaluating an independent helpdesk path rather than extending Perfex, Fluent Support Pro functions as a self-contained ticketing alternative, whereas PulseChat keeps ticket creation, invoicing and messaging coupled to the CRM record itself.
Production Deployment & Implementation Considerations
Deployment requires a functioning Perfex CRM installation on PHP 7.x or 8.x with MySQL 5.x/8.x, plus valid API credentials for each channel you intend to activate: a verified WhatsApp Business number or Meta App configuration, a Telegram bot token, IMAP/SMTP credentials for mailboxes, and OpenAI or Anthropic API keys for AI features. Email polling runs through the standard Perfex cron and background job system, so a correctly configured cron entry is a prerequisite for timely inbox delivery rather than an optional optimization. Channel credentials and tokens are stored encrypted, and granular capabilities decide which staff members may invoke AI actions, manage adapters or view analytics.
Implementation boundaries worth planning around: webhook-based channels need a publicly reachable endpoint for Meta, Telegram and similar providers; inbound traffic that cannot be matched to a configured channel falls back to the first verified channel rather than being dropped; and each AI operation consumes provider tokens, which the built-in cost logging helps monitor. Customization scope covers the AI system prompt for brand tone, model selection, temperature and max token limits, automation rule definitions, and per-channel enablement. Database schema changes ship through guarded migrations, so upgrading from earlier releases applies automatically without data loss. For product teams that also need public feature voting and roadmap publishing as a decoupled service outside the CRM stack, Votlie runs as an independent Laravel application and does not share PulseChat’s CodeIgniter runtime.
Frequently Asked Questions
Can PulseChat use both OpenAI and Anthropic models at the same time?
Yes. The module ships with a multi-provider assistant supporting OpenAI (including gpt-4o and gpt-4o-mini) and Anthropic (Claude 3.5), with configurable model selection, temperature, max tokens and a customizable system prompt per deployment.
Does email channel polling depend on the Perfex cron system?
Yes. Incoming mail is fetched through background jobs that respect Perfex cron scheduling. Since version 1.3.8, each message is saved before the poller advances, mail is not marked read until safely stored, and per-message results are written to the email-channel log.
Can multiple WhatsApp numbers and Facebook Pages run under one Meta App?
Yes. Version 1.3.6 added multi-Page, multi-Instagram and multi-WhatsApp support under a single Meta App, with inbound routing matched to the exact receiving number or Page so conversations stay separated and replies use the correct identity.
Is AI usage cost tracked per conversation?
Yes. The module logs model, token consumption, estimated cost and call duration for each AI operation, giving administrators visibility into spending alongside granular capabilities that control which staff members may use AI features.
-
v1.3.5 Latest