DoliPhone

DoliPhone for Dolibarr — WebRTC Softphone + Transcription and Artificial Intelligence

DoliPhone is the module that turns Dolibarr into a complete communications hub. It integrates a WebRTC softphone directly into the ERP interface, automatically logs every call and, optionally, transcribes and analyses the conversation content with artificial intelligence to extract summaries, classifications and follow-up actions without leaving Dolibarr.

Compatible with any Asterisk, FreeSWITCH, Kamailio, ... server with WebSocket enabled. Works from any modern browser, with no additional software to install.

---
🎯 Problem it solves

Every call with a customer generates valuable information: commitments made, issues reported, opportunities identified. Yet that information typically stays in the agent's memory or, at best, in a handwritten note after the call.

DoliPhone automates the entire cycle: the call is made from Dolibarr, logged in the CDR, recorded in stereo, and the artificial intelligence transcribes the conversation identifying each speaker, generates an executive summary, classifies the call type and proposes the next action. All of this is linked to the Dolibarr third-party or contact, visible to the team in real time.

---
📞 INTEGRATED WEBRTC SOFTPHONE

- The softphone opens in an independent window; since it is not a popup, you can open any other windows you need without interfering with the softphone
- Connection to Asterisk, FreeSWITCH, Kamailio, ... servers via secure WebSocket (WSS) or plain WebSocket (WS)
- Make and receive VoIP calls directly from the browser, no plugins required
- Automatic SIP registration on login, with manual refresh on demand
- Automatic detection and opening of the third-party/contact record on incoming calls
- Visual incoming call notification with the name and information of the identified contact
- One-click outbound call from third-party, contact and user records
- Real-time SIP registration status (Registered / Connecting / Disconnected)

---
🎛 DIALLER FEATURES

- Record all calls: automatic stereo WebM recording (left channel = agent, right channel = caller) with automatic upload when the call ends
- Hold: pauses the call with automatic microphone muting; the caller hears hold music from the PBX server
- Do Not Disturb (DND): sends DND status to the server; incoming calls are automatically rejected
- Auto-answer: answers incoming calls without manual intervention, ideal for call centres and technical support
- SIP registration refresh: reconnects to the server with one click without reloading the page
- Mute: disables the microphone during the call without hanging up

---
📋 CALL LOG (CDR)

- Complete record of all calls: date, total duration, effective talk time (billsec), origin number, destination number, direction and termination status
- Automatic linking with Dolibarr third-parties and contacts by phone number matching
- Bulk auto-association tool that processes all CDR records without an assigned third-party and links them automatically
- Built-in audio player in the call list to listen to recordings without leaving the screen
- Direct download of recordings in WebM format
- "Calls" tabs integrated in third-party, contact and internal (user) records
- Advanced filters: by date, by user, transcribed only, pending transcription, with recording
- Dashboard with KPIs: total calls, transcribed calls, pending, average duration, inbound/outbound breakdown

---
🤖 ARTIFICIAL INTELLIGENCE AND TRANSCRIPTION

DoliPhone includes a complete three-step AI pipeline: transcription → diarisation → analysis.

Automatic transcription with Whisper
- Automatic language detection from the audio (no need to configure the recording language)
- Three activation modes: manual (per-call button), automatic for all users, automatic for selected users only
- Compatible with recordings in any language

Diarisation: speaker identification
- DoliPhone identifies who is speaking at each moment and assigns a different colour to each participant
- For stereo recordings: automatically splits channels using ffmpeg (left channel = agent, right channel = customer)
- For mono recordings: uses the AI model to identify speaker turns by context and pauses
- The transcript is displayed as a visual dialogue with a colour bar per speaker and a participant legend

Analysis with customisable prompt
- The administrator configures a global prompt that defines the analysis behaviour
- The prompt controls the response language: if the prompt says "reply in English", both the summary and the transcript will appear in English
- The AI automatically generates:
  - Executive summary of the conversation (maximum 3 sentences)
  - Call type: sale, support, enquiry, complaint, follow-up, spam or other
  - Suggested next action: a concrete task for the agent
- If the transcript is in a different language from the prompt, the AI automatically translates the dialogue and keeps the original accessible with one click

---
🗂 AUTOMATIC RETENTION POLICY (GDPR)

DoliPhone includes a configurable retention system that runs via Dolibarr cron:

- Full CDR: deletes call records (including the recording) after a configurable number of days
- Recordings only: deletes the audio files from disk without touching the CDR record
- Transcriptions and AI only: sets all AI fields (transcript, summary, type, actions) to NULL, preserving the CDR and the recording

Each element has its own independent retention period. The administration panel shows in real time how many records would be affected by the current configuration before running the cleanup.

---
🧩 WHAT THE MODULE INCLUDES

- Module descriptor with granular permissions (read own / read all / create-edit)
- Full administration panel: SIP server, AI models, transcription mode, customisable prompt, advanced paths, automatic retention
- AI dashboard with KPI cards and direct access to filtered lists
- "Calls" tabs integrated in third-parties, contacts and internals
- Complete AI pipeline: Whisper + stereo/mono diarisation + analysis
- AJAX for real-time transcription from the call list without reloading the page
- Automatic retention configurable by element and period (GDPR)
- Automatic cleanup cron registered in the Dolibarr scheduled tasks system
- Based on Innovate Asterisk Softphone and JsSIP (open-source library for SIP over WebRTC)

---
⚙ COMPATIBILITY AND REQUIREMENTS

- 📦 Dolibarr 17.0 or higher (compatible with V17+)
- ⚙  PHP 7.4 – 8.3+
- 🗄  MySQL 5.7+ / MariaDB 10.2+
- 🌐 Browser with WebRTC support: Chrome, Firefox, Edge, Safari
- 📡 SIP server with WebSocket enabled (Asterisk 18+, FreeSWITCH, Kamailio, ...)
- 🔒 HTTPS mandatory on the Dolibarr server (browser requirement for microphone access)
- 🎬 ffmpeg (optional): required for stereo diarisation. Installation: `apt install ffmpeg` on Debian/Ubuntu, `dnf install ffmpeg` on RHEL/Rocky/AlmaLinux, or download from ffmpeg.org on Windows
- 🤖 OpenAI API key for AI features (transcription and softphone work without AI)
- 📄 GPL v3 licence — free software, you may use, modify and redistribute it freely

---
🖥 DEMO AVAILABLE

You can try DoliPhone on our demo installation:

🌐 https://dolibarr.igontek.com

Agent 1 access: username demo.200 / password demo.200
Agent 2 access: username demo.201 / password demo.201

To test the transcription system, you need to configure a ChatGPT API key. Remember to delete the data after testing.

---
✉ SUPPORT

Technical support by email: dolibarr@igontek.com
Developed and maintained by Igontek — https://igontek.com