Call Attendant Next

A TypeScript/Next.js port of the callattendant Python project — an automated call screener and voicemail system for landlines using a USB modem.

Supported Modems

Only hardware modems are supported. Software modems (winmodems) will not work as they lack the voice mode capabilities required for call screening and audio playback.

Modem	Recommended Baud Rate
US Robotics USR5637	57600
MultiTech MT9234MU-CDC	115200 — provides best voice quality
ZOOM 3095	115200 — provides best voice quality

If you have a hardware modem not on this list and would like support added, open an issue at foureight84/CallAttendantNext.

Stack

Next.js 15 — frontend and API routes
ts-rest + zod — type-safe REST API contract
SQLite via drizzle-orm + libsql
serialport — serial communication with USB modem
Piper TTS — real-time speech synthesis for greetings
ffmpeg — voicemail encoding (MP3) and audio conversion
Docker — containerized deployment

What's New vs. the Original Python Project

Non-blocking call handling — async/await throughout, no blocking main thread
Easier debugging — structured log events streamed to the browser via SSE; Debug Console page
Piper TTS instead of WAV files — greetings are synthesized on demand from .txt scripts; no per-voice audio files to manage
ffmpeg voicemail encoding — recordings saved as MP3 (falls back to WAV if ffmpeg is unavailable); filenames match the Python pattern: {callLogId}_{number}_{name}_{MMDDyy_HHMM}.mp3 (e.g. 42_8005551234_JOHN_SMITH_032621_1423.mp3)
Updated Nomorobo scraping — adapted for their current website format
Improved serial port handling — faster modem detection, reduced call response time
Raspberry Pi GPIO LED support — toggle via ENABLE_GPIO=true in .env

Configuration (`.env`)

Copy .env.example to .env and set your values:

cp .env.example .env

On every startup, .env values are written to the database as defaults. This means:

First run: .env seeds all settings into the database
Subsequent runs: .env can override any setting back to a specific value
Settings changed via the web UI are stored in the database and take effect immediately, but will be overwritten on next restart if .env specifies that key

Only two keys are mandatory:

Key	Description
`SERIAL_PORT`	Path to your modem device (e.g. `/dev/ttyUSB0`, `/dev/tty.usbmodem*`, `COM3`)
`SERIAL_BAUD_RATE`	Modem baud rate — `115200` for ZOOM 3095 and MultiTech MT9234MU-CDC (best voice quality); `57600` for US Robotics USR5637

All other keys are optional and fall back to sensible defaults.

Piper TTS Setup

Required. Piper TTS is used to synthesize greeting messages in real time during calls — without it, greetings will not play.

Piper is a fast, lightweight C++ TTS engine optimized for low-latency local inference — well suited for real-time telephony use. While the library is no longer actively maintained, it remains fully functional with modern speech models as long as they are in ONNX format. You need:

The Piper binary
At least one voice model (.onnx + .onnx.json) in the piper-models/ directory

1. Download Piper Binary

Download the release for your platform from the rhasspy/piper releases page:

Platform	File
Linux x86_64	`piper_linux_x86_64.tar.gz`
Linux arm64 (Pi 4/5)	`piper_linux_aarch64.tar.gz`
macOS x86_64	`piper_macos_x64.tar.gz`
Windows	`piper_windows_amd64.zip`

Extract into a piper/ directory at the project root:

mkdir -p piper
tar -xzf piper_linux_aarch64.tar.gz --strip-components=1 -C piper

Set the path in .env:

PIPER_BINARY=./piper/piper
PIPER_MODELS_DIR=./piper-models

Docker: the Dockerfile downloads the Linux x86_64 binary automatically. You only need to supply models.

2. Download a Voice Model

Voice models are hosted on Hugging Face — rhasspy/piper-voices.

Each voice requires two files:

<model>.onnx — the model weights
<model>.onnx.json — metadata including sample rate

Example — US English female (medium quality):

mkdir -p piper-models

# Download model + metadata
wget -P piper-models https://huggingface.co/rhasspy/piper-voices/resolve/main/en/en_US/hfc_female/medium/en_US-hfc_female-medium.onnx
wget -P piper-models https://huggingface.co/rhasspy/piper-voices/resolve/main/en/en_US/hfc_female/medium/en_US-hfc_female-medium.onnx.json

Browse all available voices at: https://huggingface.co/rhasspy/piper-voices/tree/main

After downloading, select the model in Settings → Voice Model and click Play to preview.

Docker

Build and Run

# Copy and edit your .env
cp .env.example .env

# Create directories for persistent data
mkdir -p data messages logs

# Place your .onnx models in ./piper-models/
mkdir -p piper-models
# (download models as shown above)

# Start
docker compose up -d

The app will be available at http://localhost:3000.

Notes

The Dockerfile downloads the piper binary automatically during build
Mount your modem device via devices in docker-compose.yml (already configured for /dev/ttyUSB0). If your modem is on a different path, update both the devices entry and the SERIAL_PORT environment variable in docker-compose.yml
data/, messages/, and logs/ must exist before first run — Docker will create them automatically on start but as root-owned, which can cause permission issues
Pass all config via environment variables in docker-compose.yml — no .env file is loaded inside the container

Rebuild after code changes

docker compose up -d --build

Bare Metal (Raspberry Pi / Linux)

Prerequisites

# Node.js 22+
curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash -
sudo apt install -y nodejs

# ffmpeg and build tools
sudo apt install -y ffmpeg python3 make g++ libudev-dev

# Add your user to the dialout group for serial port access
sudo usermod -aG dialout $USER
# Log out and back in for this to take effect

Install and Build

git clone https://github.com/foureight84/CallAttendantNext callattendantnext
cd callattendantnext

npm install
npm run build

Configure

cp .env.example .env
# Edit .env — set SERIAL_PORT and SERIAL_BAUD_RATE at minimum
nano .env

# Create directories for persistent data
mkdir -p data messages logs

Download Piper and models as described in the Piper TTS Setup section above.

Run

npm start

The app will be available at http://localhost:3000 (or the port set via PORT in .env).

Run as a systemd Service

sudo nano /etc/systemd/system/callattendant.service

[Unit]
Description=Call Attendant Next
After=network.target

[Service]
Type=simple
User=pi
WorkingDirectory=/home/pi/callattendantnext
ExecStart=/usr/bin/npm start
Restart=on-failure
RestartSec=5

[Install]
WantedBy=multi-user.target

sudo systemctl daemon-reload
sudo systemctl enable callattendant
sudo systemctl start callattendant

Raspberry Pi GPIO LEDs (Optional)

Note: This feature has not been tested.

Set ENABLE_GPIO=true in .env to enable LED indicators on GPIO pins (requires the onoff package and appropriate wiring). See lib/modem/gpio.ts for pin assignments.

Migrating from the Python callattendant

If you have an existing callattendant Python installation, use the included migration script to import your call log, whitelist, blocklist, and voicemail recordings.

The schema is identical between both apps — migration is a direct copy. Voicemail WAV files are copied as-is and served without conversion.

1. Stop the new app (if running)

# systemd
sudo systemctl stop callattendant

# or just Ctrl-C if running in the foreground

2. Dry run — verify counts before committing

npx tsx scripts/migrate-from-python.ts \
  --old-db /path/to/callattendant/callattendant.db \
  --old-messages /path/to/callattendant/messages \
  --dry-run

Check that the row counts look correct before proceeding.

3. Run the migration

npx tsx scripts/migrate-from-python.ts \
  --old-db /path/to/callattendant/callattendant.db \
  --old-messages /path/to/callattendant/messages

The script will print a summary when done:

Whitelist : 12 / 12 rows
Blacklist : 47 / 47 rows
CallLog   : 1842 / 1842 rows
Message   : 38 / 38 rows
Files     : 38 copied, 0 already present
Done.

4. Start the new app

npm start

All historical call log entries, whitelist/blocklist entries, and voicemail recordings will be present.

Options

Flag	Default	Description
`--old-db`	(required)	Path to the Python app's `callattendant.db`
`--old-messages`	(optional)	Path to the Python app's `messages/` directory
`--new-db`	`./callattendant.db`	Path to the new app's database (Docker: `./data/callattendant.db`)
`--new-messages`	`./messages`	Path to the new app's messages directory
`--dry-run`	—	Read and count everything; write nothing

The script is safe to re-run — it uses INSERT OR IGNORE for all rows and skips files that already exist in the destination.

Note: Settings are not migrated. The Python app has no Settings table; the new app seeds settings from your .env on first startup.

Running the migration when both apps are in Docker

The new app's database and messages directory are bind-mounted to the host (./data/ and ./messages/ next to docker-compose.yml), so the migration script can reach them directly.

Step 1 — Stop the new app

docker compose down

Step 2 — Get the old database and messages onto the host

If the old Python callattendant also uses bind mounts, the files are already on the host — note their paths and skip to step 3.

If the old app uses named Docker volumes, copy the files out first:

# Find the old container name
docker ps -a

# Copy the database
docker cp <old-container>:/app/callattendant.db /tmp/old-callattendant.db

# Copy the messages directory
docker cp <old-container>:/app/messages /tmp/old-messages

Adjust /app/callattendant.db and /app/messages to match the actual paths inside the old container if they differ.

Step 3 — Run the migration inside a one-off container

Use the already-built callattendantnext image so Node.js and all dependencies are available — no need to install anything on the host:

# Dry run first
docker run --rm \
  -v /tmp/old-callattendant.db:/old/callattendant.db:ro \
  -v /tmp/old-messages:/old/messages:ro \
  -v ./data:/app/data \
  -v ./messages:/app/messages \
  callattendantnext-callattendant \
  node_modules/.bin/tsx scripts/migrate-from-python.ts \
    --old-db /old/callattendant.db \
    --old-messages /old/messages \
    --new-db /app/data/callattendant.db \
    --new-messages /app/messages \
    --dry-run

# Run for real (remove --dry-run)
docker run --rm \
  -v /tmp/old-callattendant.db:/old/callattendant.db:ro \
  -v /tmp/old-messages:/old/messages:ro \
  -v ./data:/app/data \
  -v ./messages:/app/messages \
  callattendantnext-callattendant \
  node_modules/.bin/tsx scripts/migrate-from-python.ts \
    --old-db /old/callattendant.db \
    --old-messages /old/messages \
    --new-db /app/data/callattendant.db \
    --new-messages /app/messages

The image name (callattendantnext-callattendant) is the default Docker Compose generates from the project directory name and service name. If yours differs, check with docker images.

If the old app's files are already bind-mounted on the host (e.g. at ./old-messages), pass the host paths directly — no docker cp step needed.

Step 4 — Start the new app

docker compose up -d

Greeting Scripts

Greeting text files live in public/audio/script/. Edit them to customize what is spoken. Do not rename these files — the filenames are hardcoded and the application will not find them if changed.

File	When used
`general_greeting.txt`	Played when a call is answered and sent to voicemail
`please_leave_message.txt`	Played after the greeting, just before recording begins
`blocked_greeting.txt`	Played to blocklisted callers before hanging up (when blocklist action is set to "play greeting")
`goodbye.txt`	Played when ending an interaction
`invalid_response.txt`	Played when an unrecognized input is received
`voice_mail_menu.txt`	Played to present the voicemail menu options

Name		Name	Last commit message	Last commit date
Latest commit History 48 Commits
app		app
components		components
lib		lib
public/audio/script		public/audio/script
screenshots		screenshots
scripts		scripts
.env.example		.env.example
.gitignore		.gitignore
Dockerfile		Dockerfile
README.md		README.md
docker-compose.yml		docker-compose.yml
instrumentation.ts		instrumentation.ts
next-env.d.ts		next-env.d.ts
next.config.ts		next.config.ts
package-lock.json		package-lock.json
package.json		package.json
postcss.config.cjs		postcss.config.cjs
server.ts		server.ts
tsconfig.json		tsconfig.json

Folders and files

Latest commit

History

Repository files navigation

Call Attendant Next

Supported Modems

Stack

What's New vs. the Original Python Project

Configuration (.env)

Piper TTS Setup

1. Download Piper Binary

2. Download a Voice Model

Docker

Build and Run

Notes

Rebuild after code changes

Bare Metal (Raspberry Pi / Linux)

Prerequisites

Install and Build

Configure

Run

Run as a systemd Service

Raspberry Pi GPIO LEDs (Optional)

Migrating from the Python callattendant

1. Stop the new app (if running)

2. Dry run — verify counts before committing

3. Run the migration

4. Start the new app

Options

Running the migration when both apps are in Docker

Greeting Scripts

Screenshots

Dashboard

Call Log

Phonebook

Blocklist

Voicemail

Settings

Debug Console

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases 3

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Configuration (`.env`)

Packages