Chatbot Zero - AI on Raspberry Pi zero

Description

This is a pocket-sized AI chatbot device built using a Raspberry Pi Zero 2w. Just press the button, speak, and it talks back like a futuristic walkie-talkie with a mind of its own. Built-in screen, RGB led, microphone & speaker. Push-to-talk button. Fully portable and USB rechargeable. AI responses generated using OpenAI’s API or other cloud services, configurable. A TypeScript server handles the logic and message flow on RPI, and it's extendable: run your own tools or scripts in response to AI commands.

Details

Subsequent Goals:

* ✅ [Hardware] Use mouse buttons instead to improve button comfort
* ✅ [Hardware] Optimize the HAT and enclosure structure, with the speaker outlet facing upward
* ✅ [Hardware] PCB layout redesign and logo for the HAT
* [Hardware] Support RPI camera input
* ✅ [Software] Integrate the tool with the API
* ✅ [Software] Enable the AI assistant to adjust the volume autonomously
* ✅ [Software] Reset the conversation history if there is no speech for five minutes
* ✅ [Software] Support custom tools (open light demo)
* [Software] Support third party MCP server
* ✅ [Software] Support Local LLM (ollama)
* ✅ [Software] Support LLM Thinking
* ✅ [Software] Support ASR and TTS locally
* [Software] Support speaker identity recognition

Components

1 × Raspberry Pi zero 2w

1 × Whisplay HAT (released on Oct 9, 2025) 1.69inch LCD screen, wm8960 sound chip with on-board speaker and mic, RGB led, GPIO Button

1 × PiSugar 3 1200mAh, battery, UPS

1 × 3D resin printed enclosure

Project Logs

Collapse

Offline Version Tutorial (RPi 5B)
Jdaie • 10/30/2025 at 09:09 • 0 comments
Step by step tutorial
Jdaie • 10/15/2025 at 01:21 • 0 comments
Whisplay HAT is now available!
Jdaie • 10/09/2025 at 12:00 • 0 comments

I'm thrilled to share that the Whisplay HAT is available today!
Link: https://www.pisugar.com/products/whisplay-hat-for-pi-zero-2w-audio-display
Chat With Local Ollama
Jdaie • 08/16/2025 at 12:51 • 0 comments

The chatbot now support local ollama!
This demo uses the deepseek-r1:1.5b model to generate responses. Since this is a reasoning model, I added a "thinking" display in the chatbot UI. You can choose to turn off the "thinking," which makes the responses faster, but the quality may noticeably decrease.
I also tried running Ollama on the Zero2W with a smaller model, smollm2:135m-instruct-q4_K_S, but after answering just one question, the Zero2W became extremely sluggish — its limited hardware performance is the objective bottleneck.
Running LLM locally on RPI zero 2w
Jdaie • 08/04/2025 at 03:18 • 0 comments

I tried running Ollama on the Zero 2W, and the output speed was actually quite decent. I also ran a chatbot program at the same time — while it did slow things down a bit, it shows that the Zero 2W does have potential! That said, the core temperature quickly rose to 68°C, so it’s clearly struggling a bit.
Install ollama following this tutorial: https://github.com/Gilzone/Installing-a-LLM-on-Raspberry-Pi-Zero-2-W
Logo for Whisplay HAT
Jdaie • 07/20/2025 at 02:06 • 0 comments

Today I received the HAT printed with the Whisplay logo. The logo features a screen-faced robot with a cartoon expression that gives off a strong electronic vibe. Interestingly, its ears are actually speakers!
You can just the waitlist via the link below:
https://www.pisugar.com/pages/subscribe-whisplay-hat
switch to mouse button
Jdaie • 06/24/2025 at 02:33 • 0 comments

I want to make this device work like a walkie-talkie, where you press and hold the button to talk, so the button will be used quite frequently. That’s why I specifically chose this mouse button. It has excellent tactile feedback — you can easily press it with a light touch. But compared to a regular button, it’s definitely a big guy!