Skip to main content

Day 9: Use OpenAI as Your EchoKit LLM Provider | The First 30 Days with EchoKit

· 4 min read

(And today, you’ll see how easy it is to use OpenAI as your LLM provider.)

Hey everyone, and welcome back! We've covered a ton of ground over the past two weeks in "The First 30 Days with EchoKit." Seriously, look how much we've accomplished:

If you remember, everything inside EchoKit runs through that simple yet incredibly powerful pipeline: ASR → LLM → TTS so far.

Each piece plays a crucial part in the voice AI loop:

  • ASR (The Ears): Converts your spoken words into text.
  • LLM (The Brain): Interprets that text, thinks about it, and decides what the perfect response should be.
  • TTS (The Mouth): Turns the final text answer back into speech.

Last week, we were all about replacing Whisper and swapping out the "ears." For the next few days, we're putting the spotlight squarely on the middle piece: the LLM.

And today, we’re starting with the most common and powerful choice out there—OpenAI!

⭐ What Exactly Does the LLM Do in the EchoKit Server? (It's the Mastermind!)​

The LLM is, quite literally, the mastermind of your entire setup. It's the engine that:

  • Instantly grasps what the user actually wants.
  • Processes all the conversational history (context).
  • Generates those helpful, natural, and human-like responses.
  • Controls how your EchoKit behaves during a conversation.
  • And, yes, it calls the necessary MCP servers to get things done!

EchoKit proudly supports any provider that uses an OpenAI-compatible LLM API.

Step 1 — Get Your Key Ready​

Open up your trusted config.toml file and find the [llm] section. Replace it with this block:

[llm]
llm_chat_url = "https://api.openai.com/v1/chat/completions"
api_key = "YOUR_OPENAI_KEY" # Don't forget to replace this!
model = "gpt-5-mini-2025-08-07" # Choose your favorite model here (e.g., gpt-3.5-turbo)
history = 5

Here's the quick rundown on those settings, just so you know what you're tuning:

  • [llm]: We're configuring the Large Language Model section.
  • llm_chat_url: OpenAI’s chat completions endpoint.
  • api_key: Get your key from the OpenAI API platform.
  • model: Which OpenAI model should power your EchoKit's thoughts? Up to you!
  • history: How many previous turns of the conversation should your EchoKit remember for context?

Step 2 — Time for a Quick Reboot!​

Whether you’re running your EchoKit server via Docker or from the Rust code, go ahead and restart it right now. That’s it! You're completely done with the server configuration. Told you it was easy!

Step 3 — Connect the New Brain to Your Device​

The grand finale! Time to link up your physical EchoKit device to the server with its shiny new OpenAI brain:

  1. Head over to https://echokit.dev/setup/ and reconnect the server if you need to.
  2. Pro Tip: If you only changed your LLM configuration and nothing else (URL, WiFi), you can just hit the RST button on your EchoKit device. It will restart and sync the new settings instantly!
  3. If your server URL or WiFi setup changed, you'll need to reconfigure them through the setup page, just like you did on Day 1.

Next, press that K0 button and start speaking. Every clever thing your EchoKit says back to you is now being powered by OpenAI!

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

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

Want to get your own EchoKit device?