DEV Community: anusha

Building Voice AI Workflows with Branches Instead of One Giant Prompt

anusha — Fri, 26 Jun 2026 16:54:07 +0000

I built a Telnyx Conversational Workflows example for auto insurance claim intake.

The main idea is simple: when a voice agent has to do a lot, one big prompt can get hard to manage.

For intake calls, there are usually a bunch of paths to handle:

missing information
safety checks
fallback paths
priority follow-up
backend tool calls

This example models the call as a workflow instead of putting everything into one prompt.

The workflow handles the conversation path. A small Node.js/Express backend exposes tool endpoints for structured actions:

POST /tools/create-claim-intake
POST /tools/log-claim-intake-fallback
POST /tools/flag-priority-follow-up
GET /health

I like this pattern because it makes the flow easier to inspect. You can see the nodes, branches, and tool calls instead of trying to debug everything from prompt behavior.

Run it:

git clone https://github.com/team-telnyx/telnyx-code-examples.git
cd telnyx-code-examples/build-conversational-workflow-nodejs
cp .env.example .env
npm install
npm test
node server.js

Code example:

[https://github.com/team-telnyx/telnyx-code-examples/tree/main/build-conversational-workflow-nodejs]

Building a Reusable Voice Agent with Runtime Configuration

anusha — Fri, 26 Jun 2026 16:38:56 +0000

I built a small Telnyx Voice example around a pattern I keep running into: one voice agent, multiple business contexts.
For appointment scheduling, a dental office, medical clinic, and physical therapy practice often need the same basic call flow. The assistant still answers the call, collects appointment details, and confirms the next step.

What changes is the business context: name, hours, services, tone, and greeting.

Instead of creating a separate assistant for every business, this example uses one reusable base assistant and applies runtime instructions when the call starts.

The flow is:

Caller dials a Telnyx number.
The webhook checks which number was called.
The number maps to a business config.
The server renders the right instructions and greeting.
Telnyx starts the assistant with that runtime context.

That makes it easier to reuse one assistant pattern across multiple businesses without copying the whole assistant setup.

Run it:

git clone https://github.com/team-telnyx/telnyx-code-examples.git
cd telnyx-code-examples/provisional-telnyx-voice-api-agents-nodejs
npm install
cp .env.example .env
npm run dev

You can also preview a generated payload:
npm run preview -- smile-dental

Code example:
[https://github.com/team-telnyx/telnyx-code-examples/tree/main/provisional-telnyx-voice-api-agents-nodejs]

Outbound Voice Agent that Drops Out on Hold and Rejoins When a Human Answers

anusha — Fri, 26 Jun 2026 16:11:34 +0000

Outbound hold agent with Telnyx Voice

This Python example shows how to handle outbound calls where the agent has to wait on hold. It places an outbound call, uses an assistant for IVR/menu navigation, pauses during hold, and resumes when a human representative answers.

[https://github.com/team-telnyx/telnyx-code-examples/tree/main/outbound-hold-agent-python]

Check out the full example here.