Quickstart

The example bots live in examples/:

• echo: speak into the browser, hear yourself back. No API keys.
• voicebot: the full voice agent. STT → LLM → TTS with turn-taking and barge-in, plus long-term memory and tracing.
• voice/: the voice agent on each provider’s STT, LLM and TTS, wired explicitly in Go on the dev runner (go run ./examples/voice/voice-<provider>), with a browser client at http://localhost:7860.

Run them with Docker (no host setup) or with a local Go toolchain.

Run with Docker

jargo publishes a build base and a distroless runtime base image, so you can containerise a bot without installing any native dependencies on the host. The Deploy with Docker guide has a copyable two-stage Dockerfile for the example bots and the run command (-e DEEPGRAM_API_KEY=… etc., then open http://localhost:8080).

Run locally

Prerequisites

The default build is cgo-free: go build ./... needs no C toolchain. One native library is loaded at run time, through purego:

• ONNX Runtime: VAD and end-of-turn detection.

# Download a build for your platform, then point jargo at it:
export JARGO_ONNXRUNTIME_LIB=/path/to/libonnxruntime.so

Get it from the onnxruntime releases; the onnxruntime-linux-* archive contains lib/libonnxruntime.so. If the variable is unset, jargo looks for the library by its conventional name on the loader’s default search path.

Without the ONNX Runtime the voice bot still runs: it falls back to STT endpointing for turn-taking and loses barge-in. Everything else in the list below is optional. See Installation for RNNoise.

Echo bot, no keys

go run ./examples/echo           # then open http://localhost:8080

Voice bot

Set the provider API keys, then run:

export DEEPGRAM_API_KEY=...       # STT
export ANTHROPIC_API_KEY=...      # LLM
export ELEVENLABS_API_KEY=...     # TTS
go run ./examples/voicebot        # then open http://localhost:8080

The voicebot runs a fixed Deepgram + Anthropic + ElevenLabs stack. To try a different provider, run one of the per-provider examples under examples/voice, one self-contained file each, with the provider wired explicitly in Go:

go run ./examples/voice/voice-cartesia   # Cartesia STT + TTS, OpenAI LLM
go run ./examples/voice/voice-openai     # OpenAI STT + LLM + TTS
go run ./examples/voice/voice-groq       # Groq STT + LLM + TTS

Each runs on the dev runner: open http://localhost:7860 and connect. Each example’s doc comment lists the API keys it needs.

Next

• Your first bot: the same pipeline, built up line by line.
• Architecture: how the pieces fit together.
• Turn-taking: tuning end-of-turn detection and barge-in.