Skip to main content

Day 10: Using OpenRouter as Your EchoKit LLM Provider | The First 30 Days with EchoKit

· 3 min read

Over the past two weeks, we’ve explored many moving parts inside the ASR → LLM → TTS pipeline. We’ve changed the welcome voice, updated the boot screen, switched between multiple ASR providers, and learned how to run the EchoKit server both via Docker and from source.

This week, we shifted our focus to the LLM, the part of the pipeline that interprets what you say and decides how EchoKit should respond.

Yesterday, we used OpenAI as the LLM provider. Today, we’re going to try something more flexible — OpenRouter.

What Is OpenRouter?​

OpenRouter is a unified API gateway that gives you access to many different LLMs without changing your code structure. It’s fully OpenAI-API compatible in the context of text generation models, which means EchoKit can work with it right away.

Some reasons I like OpenRouter:

  • You can choose from a wide selection of open source LLMs: Qwen, Llama, DeepSeek, Mistral, etc.
  • Switching models doesn’t require code changes — just update the model name.
  • Often more cost-effective and more customizable.
  • Great for exploring different personalities and response styles for EchoKit.

How to Use OpenRouter as Your LLM Provider​

1. Get Your OpenRouter API Key​

Go to your OpenRouter dashboard and generate an API key. Keep it private — it works just like an OpenAI API key.

2. Update config.toml​

Open your EchoKit server configuration file and locate LLM:

[llm]
provider = "openrouter"
chat_url = "https://openrouter.ai/api/v1/chat/completions"
api_key = "YOUR_API_KEY_HERE"
model = "qwen/qwen3-14b"
history = 5

You can replace the model with any supported model on OpenRouter.

3. Restart Your EchoKit Server​

If you’re running from the Rust source code, after saving the updated config.toml:

# Enable debug logging
export RUST_LOG=debug

# Run the EchoKit server in the background
nohup target/release/echokit_server &

Or using Docker:

docker run --rm \
  -p 8080:8080 \
  -v $(pwd)/config.toml:/app/config.toml \
  secondstate/echokit:latest-server-vad &

Then return to the setup page, pair the device if needed, and EchoKit will now respond using OpenRouter. That's it.

Connecting EchoKit to OpenRouter feels like I unlocked a new layer of creativity. OpenAI gives you a clean and reliable default, but OpenRouter opens the door to experimenting with different model behaviors, tones, and personalities — all without changing your application logic.

If you enjoy tweaking, tuning, and exploring how different models shape your EchoKit’s “brain”, OpenRouter is one of the best tools for that.

If you want to share your experience or see what others are building with EchoKit + OpenRouter:

  • Join the EchoKit Discord
  • Or share your latency tests, setups, and experiments — we love seeing them

Want to get your own EchoKit device?