Skip to content

First Observatory visit

Copy page

The verification checks passed. Now look at what your platform sees about your project.

Open the Observatory in your deployment — the /observatory route of your docs site or dashboard. It reads your registry + memory: the signals from every project wired into your platform. Nothing project-specific to set up; if your project’s signals have reached your platform (which they have, if Verify it worked passed), they show up here.

Want to see what the console looks like before you stand up your own? The public demo at tapestry-khaki.vercel.app/observatory runs on sample data — a read-only reference, not your data.

A new project’s first visit looks sparse — that’s correct, not a failure:

What it showsWhy it’s sparse on day one
Candidate inboxThe Observer cron runs every 6 hours. Your project hasn’t had a cycle yet, so the candidates list is empty or has only general findings.
Memory lensEmpty until you’ve written or accumulated memories. The auto-recall block in your session was empty for the same reason.
Architecture lensEmpty until your project starts running scripts/architecture_snapshot.py (optional — onboarding doesn’t add it; see Architecture snapshots).
Observer lensEmpty until the cron has interpreted enough signal from your project to surface findings.

Sparse first-visit is the norm. Patterns accumulate; come back after a few work sessions.

After a few days of normal use, the cockpit starts populating:

  • Corrections you make in chat become friction patterns the Observer surfaces.
  • Memory writes (your decisions, observer synthesis memos) populate the memory lens.
  • Architecture-snapshot runs (if you wire the optional scripts in) populate the architecture lens.
  • Recurring candidates across projects start composing under the cross-project lens (planned — currently surfaces as deduplicated findings in the candidate inbox).

What’s the right question to ask the cockpit

Section titled “What’s the right question to ask the cockpit”

The Observatory isn’t a status dashboard. It’s a trajectory dashboard. Ask trajectory questions:

  • Is coordination on this project getting smoother or rougher? (Friction recurrence, correction frequency)
  • Is anything stabilizing into a candidate worth promoting? (Candidate inbox + observer confidence)
  • Has memory been load-bearing, or have I been re-explaining things? (Memory miss rate vs reinforcement)
  • Has my project’s shape changed in ways I didn’t notice? (Architecture diffs)

State questions (“is the API up right now?”) belong in a regular operational dashboard, not here.