Deployment

Docker Compose Services

docker-compose.yml defines:

Service	Command
`db`	PostgreSQL 16
`agent`	`${AGENT_START_CMD:-python agent_realtime.py start}`
`lead_fetcher`	`python lead_fetcher.py`
`scheduler`	`python scheduler.py`
`dashboard`	`streamlit run streamlit_app.py --server.port 8501 --server.address 0.0.0.0`
`web`	`python token_server.py`

docker compose up --build

Use the Compose service hostname:

DATABASE_URL=postgresql://ignitech:ignitech@db:5432/ignitech

Default Compose command starts OpenAI Realtime:

AGENT_START_CMD=python agent_realtime.py start

For Sarvam:

AGENT_START_CMD=python agent.py start

The dashboard is bound to localhost:

127.0.0.1:8502 -> container 8501

This keeps it private by default. Put a reverse proxy or tunnel in front of it if external access is needed.

The repository includes:

Recommended production steps:

Provision an EC2 instance with Docker and Docker Compose.
Clone the repository.
Create .env with production secrets.
Confirm security-group inbound rules for dashboard/webhook access as needed.
Run docker compose up --build -d.
Check logs with docker compose logs -f.
Verify python sip_setup.py list against the same LiveKit project before live calling.

.env contains real LiveKit, Vobiz, database, and model credentials.
LIVEKIT_SIP_TRUNK_ID is the outbound trunk ID.
CALL_WHITELIST_NUMBERS is empty only when live calling is intended.
CALL_ALLOWED_WEEKDAYS, CALL_HOURS_START, and CALL_HOURS_END match calling policy.
MAX_CONCURRENT_CALLS is set to an approved value.
Dashboard is protected by network rules or auth controls.
Facebook webhook URL is public and uses HTTPS in production.
Database volume backups are configured.
Logs do not expose API keys or sensitive qualification data.