Project Preflight: Readiness as a Ritual
Onboarding is one of the most neglected parts of software. We usually treat it as a race to get the user clicking buttons, a series of tooltips and "Got it!" prompts. But when the "tool" is a complex AI, especially one designed for sensitive contexts like mental health, that approach feels reckless. This conviction led me to build Project Preflight.
The goal of Preflight isn't to just onboard a user; it's to prepare them for a partnership with AI. The Readme describes it as an "AI-readiness questionnaire," but it's really a slow, reflective, and conversational process. It's designed to warm users into a relationship with the technology, setting expectations and understanding their comfort levels before they ever start a real session.
Every Interaction is Research
Preflight is built on a simple but powerful mantra: every interaction is research, and every survey is a conversation.
Instead of a static form, the questions are driven by a versioned JSON file. This "Form DSL" means I can run experiments, change the flow, and adapt the survey without a full redeployment. The system autosaves progress and logs every interaction, turning the onboarding process itself into a valuable dataset.
After the initial questions, a carefully scripted LLM-powered chat begins. The prompt pipeline has firm rules: ask one question at a time, don't offer medical advice, stay polite. It's a gentle handshake, not an attempt to be a therapist. It gets the user accustomed to the conversational nature of the AI in a low-stakes environment.
The Conviction: Slowing Down to Build Trust
In a world that prizes instant gratification, deliberately slowing a user down feels counterintuitive. I'm absolutely risking losing people who just want to get to the "point." But that's the core philosophical stance of Preflight.
For a tool as potentially impactful as Sidekick, a rushed, unconsidered engagement with its AI could be more harmful than helpful. Misunderstandings, frustration, and unrealistic expectations are real dangers. So, this "readiness as a ritual" is worth the potential friction because it does two things:
- It prepares the user for the AI: It encourages them to think introspectively about what they want from the interaction and what their boundaries are.
- It prepares the AI for the user: It gives the Oceanheart ecosystem a real-time, evolving profile of the user's comfort levels, biases, and learning style. This allows for an unprecedented level of personalized, ethical guardrails.
The tech stack is a diplomatic summit—a SvelteKit frontend for a great UX, a FastAPI backend for the logic, and Passport for auth. But the real technology here is the methodology: using the first point of contact not for extraction, but for mutual understanding. It's the first step in treating the human-AI interaction with the gravity it deserves.