Wiki: MCP tool surface audit (#335) - Architecture §10, Decision Log from main, ledger
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WT9mcMR4C42maYytEC8CRt
Wiki: deck widget versions (#305), battle logs v2 (#315), Deep Think (#316) - Architecture §15b on, Decision Log from main, ledger
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015oYErNsTyzvYNppwoipzdY
Wiki: Deck-E harness v2 (#293) and follow-ups (#300-#302, #304, #312-#314) - Architecture §15b on, Decision Log from main, ledger
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015oYErNsTyzvYNppwoipzdY
Wiki: scanner OCR digit drop (#297) - Decision Log from main, Architecture confidence rule, ledger
The Decision Log regeneration from main also publishes three decisions other
sessions merged today (Deck-E sonar, deck odds, set_progress).
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WT9mcMR4C42maYytEC8CRt
Wiki: battle simulator (@deckpal/sim) and deck_odds — architecture, MCP tools, decision log, contribution record
Architecture gains a section on packages/sim: engine as sole judge of
legality, state as data, one decision shape, seeded streams, cards keyed
by printed text, a search CPU that never reads hidden information, the
paired runner and honest report, and where it runs (POST /decks/simulate
inside the /api function, 25 s budget). MCP-Setup lists check_deck,
deck_odds and simulate_battles and states 27 tools once both ship.
Decision Log regenerated with the three 2026-10-10 decisions.
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Wiki: Deck-E battle-log fix (#291) — decision log, contribution record
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Record the full SEO pass; sync Architecture and the Decision Log
- Contribution Record: PR #277 (middleware-served prerenders, per-page
metadata, catalog sitemaps) and the ARCHITECTURE note (PR #278).
- Architecture: synced from the repo's ARCHITECTURE.md on main.
- Decision Log: regenerated from decisions/ (pnpm decisions:wiki).
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WT9mcMR4C42maYytEC8CRt
Record the landing rebuild; sync Architecture and the Decision Log
- Contribution Record: the 2026-10-04 landing rebuild (PR #275, #276), and
the 2026-09-28 Deck-E chat overhaul line that session never pushed.
- Architecture: synced from the repo's ARCHITECTURE.md on main (the page
had fallen behind; the unpushed 2026-09-28 edit was stale by then).
- Decision Log: regenerated from decisions/ (pnpm decisions:wiki).
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WT9mcMR4C42maYytEC8CRt
docs: record verified Deck-E approval and quantity fixes
Document chat apply/preview intent, preserve MCP behavior, and record the measured production outcomes, actual cost and verification limits for PRs 198–200.
On-Behalf-Of: @cheyras
Co-Authored-By: GPT-5.6 Sol <noreply@anthropic.com>
Co-Authored-By: GPT-5.6 Terra <noreply@anthropic.com>
Co-Authored-By: GPT-5.6 Luna <noreply@anthropic.com>
Document administration tiers, feature controls and AI usage
Synchronize the owner guide, architecture, security boundaries, UI conventions and decision record with app commit 8f44c4b6. Live release verification remains separate.
On-Behalf-Of: @cheyras
Co-Authored-By: GPT-6 <noreply@anthropic.com>
Document reusable admin tables and verified preview status
Record the shared DataTable contract, six administration consumers, full-result filter semantics, desktop/mobile behavior and completed isolated browser acceptance. Preserve historical decisions and release records.
On-Behalf-Of: @cheyras
Co-Authored-By: Codex <noreply@openai.com>
Record bounded limiter implementation and request ordering
Synchronize final source architecture and decision records for PR188 while retaining the production release prerequisite.
On-Behalf-Of: @cheyras
Co-Authored-By: Codex in-session worker <noreply@anthropic.com>
Document administration, roles and configurable AI credits
Add the owner guide and synchronize architecture, schema, frontend, MCP and decision records for DeckPal PR188. Production schema and payment verification remain pending.
On-Behalf-Of: @cheyras
Co-Authored-By: Codex in-session worker <noreply@anthropic.com>
docs: record permanent CI boundaries and PR 184 validation
Mirror the architecture and decision record and record the completed database/browser/test cleanup work with final review and CI evidence.
On-Behalf-Of: @cheyras
Co-Authored-By: Codex in-session worker <noreply@anthropic.com>
docs: explain bounded price-history continuation
On-Behalf-Of: @cheyras
Co-Authored-By: Codex in-session worker <noreply@anthropic.com>
docs: record reviewed calendar-date and query-test repairs
On-Behalf-Of: @cheyras
Co-Authored-By: Codex in-session worker <noreply@anthropic.com>
docs: record agent history and upcoming-set completion
Synchronize the shared tool surface and verified product behavior with the repository.
On-Behalf-Of: @cheyras
Co-Authored-By: GLM-5.2 <noreply@anthropic.com>
Co-Authored-By: Codex <noreply@anthropic.com>
docs: sync repository hygiene and security contracts
On-Behalf-Of: @cheyras
Co-Authored-By: GPT-6 Astra <noreply@anthropic.com>
Sync wiki: pay-what-you-want billing (feat/pwyw-billing)
Decision-Log: the full entry -- no entitlement column, $0 as a preset
rather than an escape hatch, why the Stripe customer id is guarded by the
database rather than by the API, and why 053 backfills existing accounts
with onboarded_at stamped.
Architecture: new section 16, billing as a CACHE in front of Stripe plus
the schedule for one question. Canonical copy stays in the repo.
Contribution-Record: one ledger line, including what is named-not-done
(unverified against real Stripe -- no key exists here, B9).
NOT PUSHED, deliberately: `feat/pwyw-billing` is not merged, so these
pages describe behaviour that is not live yet. Push them with the merge.
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Purge 2a: annotate pre-rewrite commit SHAs cited in Architecture, Contribution-Record, Decision-Log
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MExGVHFfAsqf2HDAhG2wAa
Agentic pass: Decision Log entry, Architecture flows, Contribution ledger
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
SSRF hardening (issue #96): destination control on the image fetch
Architecture: what the MISS path now checks and what it does not.
Decision Log + Contribution Record: synced with the repo entry.
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Architecture + Frontend Research + Decision Log + Contribution Record: first paint has a deadline now (#75)
The grey screen that survived #87 was an unbounded `supabase.auth.getSession()`,
not a bundle: it refreshes over the network inside the 90s expiry margin and
auth-js puts no timeout on that fetch, so a stalled socket held `beforeLoad`,
AuthGuard and every api.ts request open forever.
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Deck-E narration leak filter: the tag list is derived now (issue #90)
Architecture: the "anchored on our seven tool names" sentence was describing a
list that had been wrong since `journey` and `escort` were added. Rewritten to
say the alternation is derived from `COSMETIC_TOOLS`, what pins it, and why the
23 data + 4 deep tools are deliberately out of scope.
Decision Log + Contribution Record: the corresponding entries.
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Card art: direct object URLs, cloud warm path, and the 2026-08-26 measurements
Architecture: the SPA no longer goes through the image function for art that is
already in the bucket; records why a HIT was never free (invocation + Storage
probe before the 302) and that the proxied path remains the fallback. Also
corrects the "no manifest row is a FAIL" description, which was a permanent
placeholder for 585 cards.
Data-Layer 5.3a: the cloud tier had no bulk warmer at all -- the disk warmers
cannot reach it and storage:backfill only mirrors a disk cache -- so the bucket
held only what someone had looked at (18,840 of 21,066 cards missing).
warm:cloud is that path now.
Frontend-Research C.2a: the SW route matches two origins, the image cache is
deckpal-img-v2, and the rename is load-bearing because v1 held opaque responses
the browser pads against the origin quota.
Decision Log + Contribution Record: the full entry with before/after figures.
Both the Decision Log and Data-Layer note that pkmn.gg is ruled out as a source
on the owner's decision, so warm:pkmn is not the route for the remaining 592
cards and should not be reached for on the next gap either.
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VqQ2pWdA2FAhXPjsooPy7K
Sync the wiki, and correct a page that described branch work as shipped
Brings Architecture, Decision Log and the Contribution Record in line with
the 2026-08-22 work on `decke-agent` (PR #74, open): the approval
`signature` bug that broke every approved write once
DECKE_APPROVAL_SECRET was set, the round trip extracted into a testable
module, the narration transform's move, search_cards no longer grounding
a guessed set id, and the chat model switch. Twelve DECISIONS entries
appended verbatim -- this page is a snapshot, not a rewrite.
The correction that mattered most was to our OWN previous sync. MCP Setup
said the 23 tools "are defined once, in the packages/agent-tools workspace
package". That package does not exist on origin/main. It is branch-only
work from PR #74, and the 2026-08-21 sync wrote it up as current
behaviour -- so a reader following that page would have gone looking in a
directory that is not there, and Deck-E in production has its own separate
tool set rather than a share of the MCP registry. The page now says what
is true today and names the pending PR without claiming it.
Everything new is marked as branch-only for the same reason. Nothing here
describes the collection-write path as working, because it does not: the
model will not call `log_cards` at all, which is recorded plainly rather
than omitted.
Two older pages get a banner instead of a rewrite. Data Layer and
Frontend Research §D sized a single-user Raspberry Pi deployment that the
project abandoned for Vercel + Supabase the same week they were finished.
Their measurements are captioned with the command that produced them, so
they are left as measured and the banner says which sections are still
evidence and which describe a plan that was dropped. A dated measurement
rewritten to a current number stops being evidence.
Also: removed an orphaned mid-file footer from Decision Log and the
Contribution Record, each left behind by an earlier sync that appended
below it, so both pages now carry exactly one and a reader scanning for
the date gets the right one. TypeScript's entry corrected to 7.0.2.
Home was checked against the files on disk and the AGENTS.md inventory
and left alone. Dex Data, Prior Art, Project Brief and UI Spec were
checked and deliberately not touched -- Project Brief is historical by
designation, and UI Spec already carries its own supersession notice.
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Sync wiki with Deck-E rev 2: agent-tools extraction, entitlement/meter, data access, deep tier
Brings Architecture, MCP Setup, Decision Log and the Contribution Record in
line with the deckpal repo's Deck-E PRs 1-4 and 7 (the browser tool wire fix,
server-side entitlement + spend meter on /api/chat, the packages/agent-tools
extraction with two front-ends, Deck-E's lazy-RLS data access, and the
metered deep tier). Also corrects two staleness gaps found while syncing:
the MCP tool count (21 -> 23) and the Deck-E route's dev-only status, both
of which had drifted from the canonical repo docs before this session.
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Deck-E: parentage over analysis, and the lighting was never occlusion
Five defects closed in the three.js character port, and in four of them the
recorded diagnosis had to be discarded before the real cause showed up.
Three placement errors turned out to be the same mistake: a relationship the
.blend expresses as PARENTAGE had been reimplemented analytically.
- the lid pivot is a matrix pair, not an angle pair. DeckBox_Lid.location is
keyed and reaches (0, 0.152263, -0.117046) at the full gape; leaving it at
rest put the lid 0.313 BU out of place, which is why card_stash was the
worst-scoring frame in the sweep.
- Eye_Rig is VERTEX_3-parented to the MORPHED lid, so it tracks the shape
keys. Reproduced to 1e-6; the eyes no longer draw on the inside of the
open lid.
- the brow sockets hang off Eye_Rig, not the lid (this page said otherwise
and was wrong), and carry a keyed follow-through of their own. Worst brow
error 0.36 BU -> 0.006.
The lighting residual was not occlusion. Filed for weeks as "shadows 24-37%
too bright", it was an unported EEVEE firefly clamp, and it has to be applied
to the HDRI texels rather than to the finished IBL lookup. Bucketing the
residual by surface NORMAL rather than by pixel is what showed it: the
up-facing lid top was +44% against the front face's +5%, a shape no occlusion
term can produce.
Also recorded: never quantize the glb (it parks a de-quantisation transform on
nodes the rider system overwrites), the SDF atlas must stay 16-bit (the
shader's edge band is narrower than an 8-bit step), gating a dev route needs
its import() guarded and not just its beforeLoad (the PWA was precaching
945 kB of three.js for every production user), and travel_far's flight is
baked into the .blend at 2x speed where the port is the correct one.
Parity coverage went from 8 stills to 14 stills plus 3 clips compared
frame-by-frame, and the harness was made deterministic after it was found to
vary by up to 0.045 IoU between identical runs.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EFYdqPpfFcY8z8sKzggnZJ
Document the Deck-E three.js character runtime
Architecture gains §15: the subsystem, the decision to drive 47 normalised pose
channels rather than AnimationMixer clips, and the traps worth knowing before
touching it. Frontend Research gains section F on the stack choices — vanilla
three.js over r3f, the AgX colour recipe for matching Blender, why PMREM is a
correctness fix and not just an aliasing one, and why Draco cannot be used with
morph targets.
Also records that the separate character wiki is now stale in eight measured
places: its own regeneration chain died on 2026-08-17, so the .blend is the
authority.
On-Behalf-Of: @cheyras
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Rename the product to DeckPal
Follows the repo rename: DeckScout → DeckPal, deckscout.io → deckpal.app,
and the MCP server rotom-mcp → deckpal-mcp.