Local quickstart#

Use fake providers first. This verifies service wiring without making a phone call or spending provider credits.

Before you start#

A completed installation.
A local .env file that is not committed to Git.
Ports 8000 and 3000 available.

Start locally#

In .env, set the AI, speech-to-text and text-to-speech providers to their fake implementations. Keep the public base URL on localhost.
Start the runtime and dashboard:
```
.\scripts\dev.ps1 -Seed
```
Open http://localhost:3000.

Local development may omit the dashboard password while the service remains bound to localhost. Set production authentication before exposing it.

Verify#

In another terminal, check liveness:

Invoke-RestMethod http://localhost:8000/api/health

The request should return a healthy response. Open the dashboard and confirm that its overview loads runtime status. The production readiness endpoint may intentionally report blockers while fake providers and development secrets are configured; use it when preparing a live deployment.

If it does not work#

Symptom	Check	Recovery
The health request is refused	Whether both services actually started, and on the expected ports	Read the startup output; a port already in use fails quietly in one terminal and loudly in the other
Health answers but readiness does not	`/api/ready` names the blocker	See Readiness reports blockers
The dashboard loads but shows no data	Whether you are running against the fake providers this page configures	Fake providers produce no live calls; that is expected here
The runtime refuses to start after you change a setting	Whether a phone feature was enabled without its required values	See Troubleshooting