Indus Transports Auto Dialer

Professional Windows auto dialer — Google Voice, AMD, predictive pacing.

Multi-line dialing · AMD · CRM · Excel lists · Predictive pacing

Features · Quick Start · Documentation · Screenshots · Contributors · Contributing

---

Professional Windows desktop dialer for INDUS TRANSPORTS LLC. Agents use a simple branded app while Google Voice runs in the background on each line.

🖼 Screenshots

Captured from the live application UI (light and dark themes).

Light mode

Sign-in	Dialer
Live calls	Call logs
CRM	Settings
Administration

Dark mode

Sign-in	Dialer
Live calls	Call logs
CRM	Settings
Administration

---

Documentation

Doc	Purpose
CONTRIBUTING.md	Dev setup and PR guidelines
SECURITY.md	Vulnerability reporting
CODE_OF_CONDUCT.md	Community standards
docs/RUNBOOK.md	Operator runbook
docs/AMD_TESTING.md	AMD / pacing test guide
CLIENT.md	End-user install and daily use
CLIENT_DELIVERY.md	How to package and deliver to clients

Who uses what

Role	Who	What they can do
Administrator	You (owner)	First account at install, Administration tab, add Google Voice lines, connect accounts, create client users
Agent	Your clients / dialers	Sign in with credentials you create, load lists, start dialing, listen to lines, view logs and CRM

Typical setup (your PC = administrator)

1. Install the app and create the administrator account (once).
2. In Settings, add each Google Voice line (email + password) — the app signs in automatically in the background.
3. In Administration → Export client package…, create a folder for each client PC (agent login + voice profiles + config).
4. On the client PC, install the app, copy the package contents into the app folder, run once — they see Agent sign-in only (no admin setup wizard).
5. Give each client only their email and password from the export step.

Agents do not see Google Voice settings, cannot add voice accounts, and cannot manage users. On a client workstation, even an admin password will not grant administrator features.

Client PC install (agent-only)

Use this when you configure everything on your machine and deliver a ready folder to the client.

On your PC (administrator)

1. Complete Google Voice setup in Settings.
2. Administration → Export client package…
3. Enter the client’s name, login email, and password (8+ characters).
4. Choose a save location (e.g. Desktop). You get a folder like IndusTransports_AutoDialer_Client with:
   • dialer_config.json (deployment_mode: client)
   • logs/crm.sqlite3 (single agent account)
   • data/gv_accounts.json and chrome_profiles/ (signed-in voice lines)
   • CLIENT_SETUP.txt (instructions and credentials summary)

CLI alternative:

python scripts/prepare_client_install.py --name "Jane Agent" --email jane@example.com --password "TheirPassword8"

On the client PC

1. Install the same app (Python source or built EXE).
2. Copy everything from the export folder into the app directory (merge logs, data, chrome_profiles, replace dialer_config.json).
3. Run the app — first screen is Agent sign-in only.
4. Client signs in with the email/password you set during export.

If the client sees “Workstation not configured”, the logs folder from the package was not copied correctly.

Features

• Light, client-ready interface
• Hidden Google Voice browser per line (QWebEngineView)
• Automatic sign-in using saved email/password and persistent profiles
• Live browser panel on any line — hear and talk through the embedded Google Voice call
• Test call on every line — place one manual call without starting a campaign
• Live status: Waiting, Ringing, On call, Voicemail
• Excel contact import, call timeout, voicemail auto-hangup
• Local CRM and call history
• Windows EXE build script

Requirements

• Windows 10 or 11
• Python 3.10+ (for development)
• Google Voice account per line
• Microphone/speakers allowed when prompted

pip install -r requirements.txt

Run

python autodialer_gui.py

Google Voice lines (administrator)

Add a line

1. Settings → Add account
2. Fill in Display name, Google email, and Password
3. Leave Sign in automatically in background selected (recommended)
4. The app signs in headless — no window unless Google requires CAPTCHA/2FA
5. If verification is needed: Settings → select the line → Connect account, finish in the browser, then close

Passwords are stored only in data/gv_accounts.json on this PC (not committed to git).

Duplicate a line (no second login)

Duplicate copies the entire browser profile from the source line. If the original was already signed in, the copy is ready immediately — no password or login step.

Connect manually

Connect account opens the sign-in window when automatic login fails or Google asks for 2FA.

Session files

• Profiles: chrome_profiles/<profile_name>/
• Ready flag: chrome_profiles/<profile_name>/.gv_session_ok

Dialing workflow (agent or admin)

1. Dialer → choose Excel file → Load contacts (or Sample list for testing)
2. Set Lines at once, Call timeout, Voicemail hangup
3. Confirm each line shows Ready on Live Calls
4. Start dialing
5. On Live Calls:
   • On call — the answered line highlights and opens its embedded browser panel so the agent can talk
   • Voicemail — app hangs up and moves on automatically
   • Next number / End call — manual control

Confirmed call-state pipeline

The dialer follows a VICIdial-style state pipeline:

IDLE
-> DIALING
-> RINGING
-> ANSWERED_PENDING
-> CONNECTED / VOICEMAIL / BUSY / NO_ANSWER / FAILED
-> LOG_RESULT
-> NEXT_LEAD

Important safety rules in the current implementation:

• RINGING is never treated as VOICEMAIL, FAILED, or ENDED.
• Voicemail detection is allowed only after answer evidence such as a call timer or enabled in-call controls.
• The app does not hang up during ringing before the configured call timeout.
• Human/live answer is promoted only after the local decision engine classifies the call as human/live, not merely because the outbound call is ringing.
• Each final state is logged once, then the slot cools down and moves to the next lead.

Google Voice dialing uses DOM/BOM control inside the embedded browser. The automation refuses stale call buttons whose label contains a different phone number than the current lead. If Google Voice already opened a call panel, the recovery path starts call-state polling instead of reloading the page.

Dry-run and approved live testing

For scheduler testing without placing calls, set this in dialer_config.json:

"dry_run_mode": true

Dry-run simulates DIALING -> RINGING -> NO_ANSWER and lets you verify queueing, logging, retries, cooldown, and UI behavior without touching Google Voice. Keep it false for real calling.

Live smoke tests must use only owner-approved test or CRM numbers:

python scripts/live_call_smoke.py --from-crm --crm-limit 2 --call-timeout 45

Do not mass dial or call random numbers for verification.

Live line panel

On Live Calls, click Listen on any line to open its embedded Google Voice view. When a person answers, the app automatically switches to Live Calls, highlights the picked-up line, and opens that same panel. Use the panel to talk through the computer mic/speakers, then choose Release slot / next or End call.

Reliability (watchdog, retries, logging)

For long campaigns (many numbers, several lines):

• Watchdog — restarts a stuck or high-memory line automatically (logs/dialer.log).
• Retries — failed dials retry up to 3 times (5s / 15s / 45s backoff) before FAILED in logs.
• Memory — HTTP cache cleared when starting/stopping a campaign; lines recycled after ~75 calls or high WebEngine RAM.

Tune in dialer_config.json — see docs/RUNBOOK.md.

First-time setup: copy dialer_config.example.json to dialer_config.json (defaults: 1 line, 6s cooldown). Local config is not tracked in git. On the Dialer tab, use Apply stable mode for safe settings on 8 GB PCs.

Dev cycle (test + optional push): .\scripts\dev_cycle.ps1 or .\scripts\dev_cycle.ps1 -Push

See docs/GV_DIAL_RESEARCH.md for Google Voice web dialing notes.

Load testing: docs/LOAD_TEST.md and python scripts/generate_load_test_list.py.

How call detection works

The app reads Google Voice page state and fuses it with optional local AI audio features.

Status	Meaning
Ringing	Outbound ring
On call	Person answered (timer or hold/mute controls)
Voicemail	Greeting / beep detected → auto hangup after configured seconds
Waiting	Idle, ready for next number

Live detection and audio

The app now fuses Google Voice page evidence with local Windows audio features. It does not use Twilio, paid APIs, cloud calling APIs, or fake demo detection.

Audio capture tries Windows WASAPI loopback first. If that is unavailable, it falls back to local capture devices such as Stereo Mix or VB-Cable. If no backend works, the live cards show AI Audio: NO BACKEND and the dialer continues with DOM-only detection.

Status	Meaning
Ringing	Outbound ring
On call	Human pickup confirmed; short hello/background noise wins over voicemail
Checking answer	Answer evidence exists, waiting for human vs voicemail confirmation
Voicemail	Confirmed only after answer evidence, 7+ seconds, high confidence, stable signals
Busy	Busy tone cadence detected
No answer	Ring timeout reached
Ended by operator	Manual hangup/release
Waiting	Idle, ready for next number

Audio device test:

python scripts/audio_device_test.py
python scripts/audio_device_test.py --device 14 --seconds 2

Live smoke test:

python scripts/live_call_smoke.py
python scripts/live_call_smoke.py --from-crm --crm-limit 2

With no numbers passed, the smoke script loads fresh CRM contacts and caps the run to the configured Google Voice account count. It prints [CALL DEBUG] blocks and writes a JSON report under logs/.

Final call logs include one row per call with final outcome, detection reason, confidence, and state history.

Pipeline details and dataset layout: docs/DETECTION_PIPELINE.md.

Administration (you only)

• Add user — creates agent accounts for clients
• Reset password / Activate / deactivate / Delete user
• Admins see all call logs; agents see only their own

Build EXE (client-ready)

Double-click Build Auto Dialer.bat or run:

python build.py

Output: dist/IndusTransports_AutoDialer.exe

See CLIENT_DELIVERY.md for the full administrator → client handoff checklist.

Data on disk (do not share)

Path	Purpose
data/gv_accounts.json	Voice line emails/passwords
chrome_profiles/	Google sign-in sessions
logs/crm.sqlite3	Users, CRM, call history

Operations runbook

Full troubleshooting: docs/RUNBOOK.md (stuck slots, client install, GV login, config keys).

Troubleshooting

JavaScript confirm: "Are you sure you want to leave this page?"

This appears when Google Voice thinks the app is navigating away from an active call panel. The controller auto-cancels that confirm and, before any dialpad recovery reload, checks whether a call panel is already active. If active, it does not reload; it switches into call-state polling.

If this message still appears repeatedly, collect:

• the last 50 lines from the on-screen log,
• logs/dialer.log,
• whether dry_run_mode is true or false,
• the visible Google Voice button text/number in the monitor.

“Google Voice is not ready”

1. Settings → Connect account for that line
2. Or wait ~90s after adding an account with automatic sign-in
3. Check Live Calls shows Ready

Blank sign-in window

1. Click Reload in the connect dialog
2. Restart the app (GPU fallback is enabled automatically)
3. Install: pip install PyQt6-WebEngine

Client cannot change voice settings

Expected — only administrators configure Google Voice. Clients use Dialer and Live Calls only.

Compliance and Responsible Use

This auto dialer is intended for lawful, permission-based business calling only. Users are responsible for consent, Do Not Call compliance, caller ID rules, recording disclosure where required, Google Voice terms, carrier limits, and anti-spam regulations.

Do not use this software for harassment, unsolicited spam calling, or deceptive outreach. Use verified contact lists, permission-based campaigns, and proper business identification.

The maintainers are not responsible for misuse of this software.

Status: Client-ready · Windows desktop · demo/sample contacts for testing

---

Contributors

INDUS TRANSPORTS LLC	Product owner and primary development
Cursor	AI-assisted development — planning, implementation, and documentation

Screenshots are regenerated from the live UI with python scripts/capture_readme_screenshots.py.

Support

WhatsApp: +923079670503