Selfology combines structured self-reflection programmes with an experimental AI component. My aim was a personal coach that could use someone’s answers and previous conversations as context.
Questions and conversation
You choose a topic and work through a sequence of questions. Your answers give you material to reflect on and revisit.
In the AI version, that material is used alongside a profile and conversation history. The idea was to carry the work from the programmes into further conversation about what matters to each person.
The role of vector memory
I used Qdrant as a vector database for the memory layer. It helps assemble context from the user’s stored information for the AI assistant.
My contribution and the outcome
I worked on the programme structure, question flows and software connecting responses, memory and conversation.
The prototype works, but the version with vector memory was slower than I had hoped. I don’t have enough user data to fine-tune the model. I see the outcome as partial: the programmes and conversation are connected through memory, but the project did not produce the personal coach I had set out to build.