🇺🇸💭📞 American Dream Phone 🇺🇸💭📞

+ =

Democracy Dies in Silence

See a demo of it in action here.

The first thing you need to know is that in the 90's, there was a board game called Dream Phone. This is a real thing that exists. With that whimsy in mind, American Dream Phone is a Pipecat AI Voice Agent to help you call your political representatives. Some may call it 'revenge of the robo-calls', I call it civic engagement.

dependencies

• Python 3.12+
• Node.js 22+
• uv
• API keys. See services.

setup

cp env.example .env

Add API keys as needed. See services.

# Backend
uv sync

# Frontend
cd frontend && npm install

run locally

You need two terminals:

# Terminal 1 — API server
uv run python server.py

# Terminal 2 — Frontend dev server
cd frontend && npm run dev

Open http://localhost:5173 in your browser. The frontend proxies API requests to the backend at localhost:7860.

headless / curl testing

If you don't need the frontend, you can still run the bot directly:

uv run python bot.py -t daily

test (webrtc)

curl -X POST http://localhost:7860/start \
  -H "Content-Type: application/json" \
  -d '{"createDailyRoom": true, "body": {"testInPrebuilt": true}}'

The response includes a dailyRoom URL. Open it in your browser to talk to and test the bot.

phone call

curl -X POST http://localhost:7860/start \
  -H "Content-Type: application/json" \
  -d '{"createDailyRoom": true, "dailyRoomProperties": {"enable_dialout": true}, "body": {"dialout_settings": [{"phoneNumber": "+15551234567"}]}}'

This actually calls the phone number. (Debug pro tip — open the dailyRoom URL and listen in on the conversation.)

deploy

build the frontend

cd frontend && npm run build

This outputs static files to frontend/dist/. Serve them however you like (nginx, Cloudflare Pages, etc.) and point API requests (/start, /preview, /upload-voice, /clone-voice, /representatives) at the backend.

run the backend

uv run python server.py --host 0.0.0.0 --port 7860

environment variables

All API keys go in .env. See env.example for the full list. Required:

Variable	Service
DAILY_API_KEY	Daily (transport + dialout)
ANTHROPIC_API_KEY	Claude (LLM)
DEEPGRAM_API_KEY	Deepgram (STT)
CARTESIA_API_KEY	Cartesia (TTS + voice cloning)

API endpoints

Route	Method	Description
/start	POST	Create Daily room, launch bot, start call
/preview	POST	LLM-generated preview of what the bot will say
/upload-voice	POST	Save voice recording for cloning
/clone-voice	POST	Clone voice via Cartesia API
/representatives	GET	Look up reps by address (placeholder)

project structure

bot.py          — Bot logic (IVR navigation, voicemail, human conversation)
server.py       — FastAPI server (all HTTP endpoints)
prompts/        — System prompts and message templates
frontend/       — Vite + vanilla JS web UI
test_calls.sh   — CLI test harness

services

By using pipecat, all services (transport, LLM, STT, and TTS) are all changeable. These are just the ones I started with.

• "Orchestration"
  • Pipecat - Python framework; the glue that makes it all possible.
• Transport
  • Daily - webrtc transport docs
  • The domain needs to have dialout enabled and a phone number purchased.
• LLM
  • Anthropic Claude
• STT
  • Deepgram
• TTS
  • Cartesia

testing

With the bot server running (uv run python server.py), use the test harness:

# Browser test — talk to the bot directly, verify IVR detection + conversation
./test_calls.sh webrtc

# Phone test — calls TEST_PHONE_NUMBER, verify IVR nav + voicemail/human routing
TEST_PHONE_NUMBER=+15551234567 ./test_calls.sh dialout

# Run both
TEST_PHONE_NUMBER=+15551234567 ./test_calls.sh all

Each test starts a session and returns a Daily room URL. Open it in your browser to watch both the bot and telephony participant in real time.

TODO list / notes to self / ideas / creedthoughts

• use some API to auto get representative phone numbers like https://5calls.org/representatives-api/ (maybe an MCP around this?)

credits

• huge shout out to this pipecat example
• E l e c t r o n i c Dream Phone