What is AI buddy?

AI Buddy is a personal assistant built on the Raspberry Pi 4 that uses computer vision and artificial intelligence to recognize users, greet them by name, and carry out intelligent conversations through text and voice, and it all runs locally, so it doesn’t require an internet connection.
It demonstrates how AI can be run locally without cloud services, integrating computer vision (CV), natural language processing (NLP), and text to speech (TTS) into one smooth system.

How does AI buddy work?

1. The camera starts and detects faces in real time using OpenCV.

2. Facial encodings are compared with pre-trained encodings (dataset) in a file called encodings.pkl.

3. If recognized, the assistant greets the user.

4. The user types a message or question into the terminal.

5. The text is processed by Ollama’s LLM through LangChain.

6. The AI generates a relevant response.

7. Piper TTS converts that response into natural speech and speaks it aloud.

This is a flow chart explaining how AI buddy works.

The picture above shows AI buddy detecting my face.

Here is a demo of AI buddy:

https://drive.google.com/file/d/171cGFHOft6Y06nTtU00MMYJv4_7aJ0b3/view?usp=sharing

How did I make AI buddy?

It took me a few weeks to make AI buddy. I used open CV for the facial recognition combined with a library called facial_recognition I used a tutorial from core electronics to first make the facial recognition part. Then, I used a tutorial from the Raspberry Pi foundation to run an LLM from ollama on the Raspberry Pi 4b (I used the gemma 3:1B model from ollama as it was a small, yet reliable model). Once I got both parts working, it was time to combine both of them to create AI buddy. I used Chat GPT to help me combine both projects into one single project.

Where can AI buddy be used in real life?

AI buddy can act as a companion to others, it can act as an assistant to answer simple questions and do a lot more. It can evolve into a home companion, smart receptionist, assistive healthcare agent, or an educational robot, depending on how it’s deployed.

Conclusion:

The AI Buddy successfully demonstrates computer vision, natural language processing, and text to speech being used in harmony, running locally without internet.
It can recognize people, respond naturally, and communicate via voice, making it ideal for personal assistants, robotics, or education projects.

Future versions could include:

• Offline speech recognition (Vosk or Whisper)

• Emotion detection from facial expressions

• Real-time context memory for deeper conversations