feat(46): live pulse-design interview — agent-guided chat, template registry, KaTeX Hamiltonian (UX1) - #82
Closed
kateebonner wants to merge 2 commits into
Closed
feat(46): live pulse-design interview — agent-guided chat, template registry, KaTeX Hamiltonian (UX1)#82kateebonner wants to merge 2 commits into
kateebonner wants to merge 2 commits into
Conversation
Member
|
I see this is in draft, just checking in if this wanted eyeballs on it, I brought it down and liked the start of what I saw! |
kateebonner
force-pushed
the
kate/46-interview
branch
from
July 5, 2026 02:57
32aaf73 to
0632ef6
Compare
…egistry, KaTeX Hamiltonian (UX1)
Amico guides a goal-directed interview (System Setup → Define Model →
Target → Problem) as a NATIVE opencode session: one long-lived turn,
questions as native question-tool calls, transcript recovery on
transport loss, strict-JSON résumé. The solve leg is deterministic — no
LLM in the critical path: envelope-gated template fill, amico-run spawn,
then the existing watcher/inspector/catalog chain.
Solve-template REGISTRY: adding a modality is content, not code (proven
by a synthetic-rydberg test). PLATFORM_FAMILIES makes every platform an
interview path with demo-repo provenance, per-family warm-start doctrine,
and catalog seeds; refusals name where the physics lives. Physicist
grounding: δ as a first-class slot, multi-select physics assembling a
live Ĥ(t) panel in locally-bundled KaTeX, soft feasibility hints,
team-vault priors. Chat layout per the fork-transcript contract:
vertical option rows, no stage rail, one bottom composer carrying all
free text ("Other"/"Custom response"/revisions), loader + honest
stage-aware activity docked by the input. Honors the locked Jul-2026 UX
decisions throughout.
Stacked on the template-identity and brand-accent commits; the interview
consumes both.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
kateebonner
force-pushed
the
kate/46-interview
branch
from
July 6, 2026 16:00
0632ef6 to
e12f28c
Compare
This was referenced Jul 6, 2026
kateebonner
changed the title
kateebonner
marked this pull request as ready for review
July 6, 2026 16:22
"Other" / "Custom response" now read as placeholder, not typed text — --vscode-input-placeholderForeground with --color-dim fallback, opacity pinned so the token's own alpha governs. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Top of the UX1 stack (3/3): #85 (template identity) → #86 (brand accent) → this. Diff shrinks as the base PRs merge.
Amico guides a goal-directed interview (System Setup → Define Model → Target → Problem) as a native opencode session — one long-lived turn, questions as native question-tool calls, transcript recovery on transport loss, strict-JSON résumé. The solve leg is deterministic (no LLM in the critical path): envelope-gated template fill → amico-run → the existing watcher/inspector/save-to-catalog chain.
Solve-template registry — adding a modality is content, not code (proven by a synthetic-rydberg registry test).
PLATFORM_FAMILIESmakes every platform an interview path: demo-repo provenance (harmoniqs/<system>-demo), per-family warm-start doctrine (fluxonium COLD-ONLY; rydberg baselines from J-P), catalog seeds; envelope refusals name where the physics lives instead of dead-ending.Physicist grounding (armonissima cross-check): δ anharmonicity as a first-class slot (positive-convention validated; the résumé always shows the δ that runs), physics as ONE multi-select question whose toggled terms assemble a live Ĥ(t) panel in locally-bundled KaTeX (CSP-safe; decoherence routes to the Lindbladian; unknown terms captured, never dropped), soft feasibility hints (under-driven gates, fast-gate leakage, coarse grids), team-vault priors in the agent guidance.
Chat layout (fork-transcript contract): vertical option rows with a recommended badge, no stage rail, one bottom composer carrying all free text ("Other"/"Custom response" per question, revisions on the résumé), loader + honest stage-aware activity docked by the input. Components assume no viewport height or column count — they lift into the opencode fork as-is; the webview driver is deliberate scaffolding the fork discards.
Honors the locked Jul-2026 UX research decisions: no solver UI, no phase labels, N as the primary knob, regularizers never asked, fidelity-first staging.
Verification: 162 extension tests pass (registry, envelope, kickoff contract, hints, Hamiltonian terms, activity extraction, estimates) · typecheck + build clean · each stack layer green in isolation.
🤖 Generated with Claude Code