Mirror summary
Mirror → Safety → Door → Reflection
Find one safer, clearer next step.
Crystal Door asks a few adaptive questions, reflects a possible pattern gently, checks whether action is safe, then offers one bounded Door you can start, change or decline. v0.10 adds login-based admin access and admin-only DB, email and AI connectivity tests.
Choose what brings you here.
Answer concrete, low-pressure prompts.
Route away from unsafe action when needed.
Receive one bounded next step.
Account structure
Sign in, create an account, or reset a lost password.
Saved Mirror sessions, Door choices and reflections are stored in the database when you are signed in. v0.10 sends reset links through the Comms API and gives approved admin users a private test panel.
Password recovery
Reset password
Before we begin
A clear boundary protects the product and the user.
Crystal Door can help with
- gentle self-reflection;
- spotting possible patterns;
- choosing one small next step;
- pausing when action is not safe yet.
Crystal Door cannot provide
- therapy or diagnosis;
- emergency support;
- legal or safeguarding advice;
- certainty about another person’s motives.
Orient
What brings you here today?
Choose the closest option. It does not need to be perfect.
Mirror question
Question
Answer in your own words. One sentence is enough.
Reflection
What happened with this Door?
Seed content
Door library
The first prototype includes curated Door templates across clarity, stabilisation, action, boundary, connection, exploration, repair and pause.
Database-backed history
Session history
Sign in to save sessions to the database. If you open the HTML directly, the fallback is local browser history only.
Admin only
Admin test panel
This panel is available only to logged-in users with admin access. It proves database, Comms/email and AI connectivity without exposing secrets.
Connectivity proof
System checks
Comms email
Send test email
Admin users
Grant or revoke admin
Overview
Counts and recent activity
About this build
What v0.10 is proving
This v0.10 version is still deliberately constrained. It does not use open-ended AI yet. It uses fixed question sets, rule-based safety/readiness routing and a curated Door library so the core experience can be tested safely. It now adds normal account login, clearer email validation, password recovery by Comms API email and PostgreSQL-backed persistence, login-based admin access and connectivity tests for database, Comms/email and Ollama AI.
Core loop
Welcome/consent → Mirror questions → Safety routing → Door recommendation → Reflection → History.
Next build candidates
Add email verification, email delivery monitoring, CSRF/rate limiting, admin editing, AI-generated Mirror summaries under strict templates, proper safety review, and later shared Doors.