Donna Swarm CallPilot

Voice-first executive assistant that books appointments by calling multiple providers in parallel, negotiating slots, and confirming the best option in real time.

What It Does

• Natural voice conversation (hands-free)
• Smart intake (symptoms, insurance, location, time preferences)
• Parallel swarm calling across multiple providers
• Live transcripts and status updates
• Decision engine that ranks best option
• Calendar conflict checks
• Persistent memory via Supermemory

Quick Demo Script (90 seconds)

1. "Donna, check my calendar for today."
2. "I've had a toothache for three days. I need a dentist."
3. "Yes, call them."
4. "Why did City Dental decline?"
5. "Book the best option."

Architecture (High Level)

Browser UI ↔ FastAPI WebSocket server ↔ OpenAI Realtime (reasoning) + ElevenLabs (voice) + Supermemory (memory) + Swarm orchestrator

Tech Stack

• Backend: Python 3.11, FastAPI, AsyncIO
• Reasoning: OpenAI GPT-4o Realtime
• Voice: ElevenLabs Realtime STT + TTS
• Memory: Supermemory
• Frontend: HTML/JS + TailwindCSS

Full details: TECH_STACK.md

Setup

1) Install Dependencies

pip install -r requirements.txt

2) Configure Environment

Create a .env file:

OPENAI_API_KEY=...
ELEVENLABS_API_KEY=...
SUPERMEMORY_API_KEY=...
SUPERMEMORY_USER_ID=demo_2026_02_08
ELEVENLABS_VOICE_ID=...
ELEVENLABS_TTS_MODEL=eleven_multilingual_v2

3) Ingest Demo Memory (Optional but Recommended)

python scripts/ingest_mock_data.py

4) Run Server

python -m uvicorn app.main:app --reload --host 0.0.0.0 --port 8001

Open the UI at: http://localhost:8001

Demo Mode (Deterministic)

Enable hardcoded demo flow for recordings:

export DEMO_MODE=1
python -m uvicorn app.main:app --reload --host 0.0.0.0 --port 8001

On Windows (PowerShell):

$env:DEMO_MODE="1"
python -m uvicorn app.main:app --reload --host 0.0.0.0 --port 8001

Repo Layout

app/               # FastAPI backend
data/              # Providers, timelines, demo memory
scripts/           # Ingestion + demo helpers
static/            # UI

Notes

• No PSTN integration (simulated providers only)
• ElevenLabs powers voice (STT + TTS)
• OpenAI used for reasoning and tool calling