[fix.lab.rho.D8+] Ivan: 为什么做的东西根本没在网页上出现 — three reasons, and the first screen was one of them

He could not see any of it. I went and looked instead of guessing, and there were
three independent causes, only one of which was about code.

1. THE SERVER I GAVE HIM WAS DEAD. The LAN address handed over hours ago
   (0.0.0.0:8099) was no longer listening. Restarted under setsid so it survives
   the session, and the dirac build is now served on 0.0.0.0:8101 as well.

2. EVERYTHING ELSE RUNNING WAS BOUND TO 127.0.0.1. Four servers up — 1342 and
   8100 both serving build/dirac, 8090, and the physics backend on 8902 — every
   one of them loopback-only. From any other machine the whole product was
   unreachable, and nothing about that is visible from inside this box.

3. THE FIRST SCREEN. This is the code fix. The app booted on 1EMA, green
   fluorescent protein, whose chromophore is a covalently modified polymer
   residue rather than a deposited ligand. mol*'s ligand query cannot see it, so
   Ligand, Properties, Fields, SMARTS search and the new bond atlas all render
   "no ligand loaded" — on the first screen a new user ever sees, under a header
   that advertises "beta-barrel protein + chromophore". Every ligand-dependent
   thing built today was invisible until you knew to switch structures.

   I found this defect hours ago in the UI audit, wrote it up, and left it alone
   because "which structure a user lands on is a product decision, not a code
   error". That was the wrong call: it is the decision that made the product look
   broken. Default is now 1CBS, the catalogue's own "Protein-ligand complex" and
   the structure the README's demo section already uses.

Verified on first load rather than asserted: ligand reads REA A:200 and the atlas
renders 22 heavy atoms, 22 bonds, 16 load-bearing, 5 free torsions, 21 distinct
positions, with no interaction required.

[coord] one line in the lab facet's index.ts, another session live in this tree.

NEXT: the halogen-audit facet. Also worth saying plainly — I should be checking
that a thing is REACHABLE, not only that it is committed. A feature nobody can
open is not shipped, and I reported three of them as shipped today.

Author: ivanicu

src/app.frontend.facets.molstar-rdkit.editable/index.ts

@@ -355,7 +355,13 @@ function byId<T extends HTMLElement>(id: string): T {
 class MolecularVfxLab {
     private workbench?: Awaited<ReturnType<typeof createChemWorkbench>>;
     private baseline?: MolstarVisualSnapshot;
-    private currentMolecule = MolecularControls.find(control => control.id === '1EMA')!;
+    // 1CBS, not 1EMA. The app booted on green fluorescent protein, whose chromophore is a
+    // covalently modified polymer residue rather than a deposited ligand — so mol*'s ligand
+    // query cannot see it, and Ligand, Properties, Fields, the SMARTS search and the bond
+    // atlas all render "no ligand loaded" on the first screen a new user ever sees, under a
+    // header that advertises "+ chromophore". 1CBS is the catalogue's own "Protein-ligand
+    // complex" and is the structure the README's demo section already uses.
+    private currentMolecule = MolecularControls.find(control => control.id === '1CBS')!;
     private currentRepresentation: RepresentationId = 'polymer-and-ligand';
     private enabledUpgrades = new Set<MolecularLayerId>(RecommendedMolstarVisualUpgrades);
     private framedLigandMolecule?: string;