OllaPy.

The project

OllaPy began with my interest in open-source models, especially those I could run locally on my own computer. As ChatGPT and chatbot interfaces were becoming familiar, I wanted to try building one myself: an interface connected to Ollama through its API. At the time, I did not have the front end I wanted for that local setup, so I made my own.

I worked on switching between models, configuring the context window and handling attachments alongside the conversation. Taking inspiration from other chatbots helped me explore which features I wanted and how I wanted to use them. The project was as much about designing an interface as connecting it to a model.

I used it for a while on a MacBook Pro M1 with 8 GB of memory. That hardware shaped the experience: I had to use small models, with limited capabilities for what I wanted to do, and the responses were slow. Trying the application myself made the trade-offs of running models locally very concrete.

As the tools around Ollama evolved, I no longer felt the same need for my own interface. OllaPy remains a learning project: it taught me to design a chatbot around the features and interactions I cared about. It was useful to me at that moment, and satisfying to build.