# Deepgram Go Voice Agent

> Use when writing or reviewing Go code in this repo that runs a Deepgram Voice Agent session over WebSockets, including runtime settings, prompt updates, speak updates, injected messages, and event handling. Route standalone STT/TTS work to deepgram-go-speech-to-text or deepgram-go-text-to-speech.

- Skill: `deepgram/deepgram-go-voice-agent` (Agent Skill)
- Install (CLI): `npx skillmds@latest add deepgram/deepgram-go-voice-agent`
- Raw SKILL.md: https://api.skillmd.com/api/skills/deepgram/deepgram-go-voice-agent/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: deepgram (https://skillmd.com/u/deepgram)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/deepgram/deepgram-go-voice-agent

---


# Using Deepgram Voice Agent from the Go SDK

## When to use this product

Use this skill for live Voice Agent runtime flows in `pkg/client/agent`.

- open an agent WebSocket session
- send initial settings
- stream audio
- react to agent events and function-call messages
- update prompt/speak settings during the session

Use a different skill when:

- you only need STT (`deepgram-go-speech-to-text`)
- you only need TTS (`deepgram-go-text-to-speech`)
- you need management/admin endpoints (`deepgram-go-management-api`)

## Authentication

Set `DEEPGRAM_API_KEY` before opening the agent connection.

```bash
export DEEPGRAM_API_KEY="your_api_key"
```

## Quick start

```go
package main

import (
	"context"
	"fmt"
	"log"

	agentws "github.com/deepgram/deepgram-go-sdk/v3/pkg/api/agent/v1/websocket"
	agent "github.com/deepgram/deepgram-go-sdk/v3/pkg/client/agent"
)

func main() {
	if err := run(); err != nil {
		log.Fatal(err)
	}
}

func run() error {
	ctx := context.Background()
	settings := agent.NewSettingsConfigurationOptions()
	handler := agentws.NewDefaultChanHandler()

	conn, err := agent.NewWSUsingChanWithDefaults(ctx, settings, handler)
	if err != nil {
		return err
	}
	defer conn.Stop()

	if ok := conn.Connect(); !ok {
		return fmt.Errorf("connect failed")
	}

	conn.Start()

	// The handler receives Welcome, ConversationText, FunctionCallRequest, and audio events.
	// Stream audio frames, watch agent events, and respond to function calls as needed.
	return nil
}
```

## Key parameters

- constructors
	- `agent.NewSettingsConfigurationOptions()`
	- `agent.NewWSUsingChanWithDefaults(...)`
	- `agent.NewWSUsingChan(...)`
- runtime methods
	- `Connect`, `Start`, `ProcessMessage`, `Stream`, `Write`, `KeepAlive`
	- reconnect helpers like `AttemptReconnect` and error handling helpers in the WS client
- message and event payloads in `pkg/api/agent/v1/websocket/interfaces/types.go`
  - `UpdatePrompt`
  - `UpdateSpeak`
  - `InjectAgentMessage`
  - `InjectUserMessage`
  - `FunctionCallResponse`
  - server events such as `WelcomeResponse`, `ConversationTextResponse`, `FunctionCallRequestResponse`, `AgentStartedSpeakingResponse`, `AgentAudioDoneResponse`

## API reference (layered)

1. In-repo reference
   - `README.md`
   - `pkg/client/agent/client.go`
   - `pkg/client/agent/v1/websocket/client_channel.go`
   - `pkg/client/agent/v1/websocket/new_using_chan.go`
   - `pkg/client/interfaces/v1/types-agent.go`
   - `pkg/api/agent/v1/websocket/interfaces/types.go`
2. OpenAPI
   - `https://developers.deepgram.com/openapi.yaml`
3. AsyncAPI
   - `https://developers.deepgram.com/asyncapi.yaml`
4. Context7
   - `/llmstxt/developers_deepgram_llms_txt`
5. Product docs
   - `https://developers.deepgram.com/reference/voice-agent/voice-agent`
   - `https://developers.deepgram.com/docs/voice-agent`
   - `https://developers.deepgram.com/docs/configure-voice-agent`
   - `https://developers.deepgram.com/docs/voice-agent-message-flow`

## Gotchas

1. This repo exposes live Voice Agent runtime over WebSockets, not a persisted configuration-management surface.
2. Keep audio streaming, event handling, and any function-call response loop running concurrently.
3. Follow the example session setup instead of inventing your own event names; the repo already defines concrete message structs.
4. Channel-based constructors require an `AgentMessageChan` handler; events are routed there rather than pulled from the client.
5. Use `defer conn.Stop()`, and remember that `Connect()` returns `bool` rather than `error`.

## Example files in this repo

- `examples/agent/websocket/simple/main.go`
- `examples/agent/websocket/no_mic/main.go`
- `examples/agent/websocket/arbitrary_keys/main.go`
- `tests/unit_test/agent/agent_speak_test.go`

## Central product skills

For cross-language Deepgram product knowledge — the consolidated API reference, documentation finder, focused runnable recipes, third-party integration examples, and MCP setup — install the central skills:

```bash
npx skills add deepgram/skills
```

This SDK ships language-idiomatic code skills; `deepgram/skills` ships cross-language product knowledge (see `api`, `docs`, `recipes`, `examples`, `starters`, `setup-mcp`).