EchoKit is an open-source toolkit designed to help developers build real-world AI applications using Rust and ESP32. It handles the full pipeline: Voice Activity Detection (VAD), ASR, LLM orchestration, and TTS.
GitHub: https://github.com/second-state/echokit_server
We just released our November updates for both the hardware firmware and the agent server. Here is what we shipped.
Server Update: Dynamic Personas Hardcoding prompts is fine for a demo, but rigid for an agent. We added Dynamic Prompt Loading. You can now configure the server to fetch system prompts from a URL.
Why it matters: You can update your AI's behavior, knowledge base, or personality remotely without touching the server binary.
Firmware Update: Quality of Life For the ESP32 side, we focused on usability:
Unified Provisioning: Wi-Fi credentials and Server URLs are now configured in a single step.
Physical Controls: We mapped the hardware buttons (K1 and K2) to volume control, allowing immediate audio adjustments.
MCP Feedback: When the AI uses the Model Context Protocol (MCP) to perform a search or an action, the device now verbally notifies the user ("Please wait...") so they know the agent is "thinking".
Get Started You can flash the new firmware via the ESP32 Launchpad or build the Rust server from source.
Docs: https://echokit.dev/docs/
Update blog: https://echokit.dev/docs/dev/server-firmware-updates-nov/
Learn more by watching EchoKit demos at Open Source Conferences: https://www.secondstate.io/articles/ossummit-korea-and-kubecon-na-2025/
Top comments (0)