Pure-Python MSF 7.00 PDB extractor (no deps, stdlib only). Reads
refs/acclient.pdb directly:
- DBI stream (3) -> symbol record stream index + section header
stream index
- Section headers stream (9) -> per-segment image VA bases
- Symbol record stream (8) -> S_PUB32 records with image VAs
- TPI stream (2) -> LF_CLASS / LF_STRUCTURE named records (not
forward-declared), with size leaf + name
Includes a best-effort MSVC C++ demangler so symbols.json is
grep-friendly:
?EnchantAttribute@CEnchantmentRegistry@@QBEHKAAK@Z
-> CEnchantmentRegistry::EnchantAttribute
Both demangled `name` + raw `mangled` emitted per entry so callers
can choose. Operator overloads, vtables, and other special forms
where a partial demangle would be misleading are kept mangled.
Outputs committed to docs/research/named-retail/:
- symbols.json (2.9 MB) — 18,366 named public function symbols
- types.json (506 KB) — 5,371 unique named class/struct records
Spot check (matches discovery agent's earlier finding):
CEnchantmentRegistry::EnchantAttribute -> 0x00594570 ✓
Updated docs/research/acclient_function_map.md header preamble to
direct readers at the new symbols.json as the authoritative name
source; the hand-curated table stays as the cross-port (ACE/ACME)
index. Several addresses there are wrong vs the PDB and will be
swept in the issue #9 close (Phase E).
Closes#8 (filed in Phase D's commit). Foundation for the address
sweep + name-driven workflows from here on.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Move the high-value retail RE artifacts from refs/ (per-developer
download cache, gitignored) into committed paths so subagents +
post-compaction sessions inherit them without round-tripping:
- docs/research/named-retail/acclient_2013_pseudo_c.txt (62 MB,
Binary Ninja named pseudo-C, 99.6% function-name recovery —
18,366 named functions out of 18,598 public symbols)
- docs/research/named-retail/acclient.h (1.7 MB / 70,719 lines,
IDA-decompiled retail struct definitions verbatim — Attribute,
SecondaryAttribute, AttributeCache, Attribute2ndTable, SkillFormula,
Enchantment, CEnchantmentRegistry with _mult_list/_add_list/_vitae,
CSpellBook, MotionState, RawMotionState, MoveToStatePack, CACQualities,
CPhysicsObj — every retail object-model layout we'd otherwise have
to guess at)
- docs/research/named-retail/acclient.c (46 MB, secondary named
decomp — IDA full-binary export with mixed FUN_/named functions
plus named struct fields the chunked Ghidra output lacks)
- docs/research/data/spells.csv (3,956 spells × 35 cols including
Family for buff stacking — issue #6 unblocked)
actestclient-master vendored at references/actestclient/ (extracted
from refs/actestclient-master-2019-01-10.zip; contains the canonical
machine-readable wire-schema messages.xml). Covered by existing
references/ gitignore — per-developer reference, not committed.
Repo precedent for committing decompiled retail content was set at
commit 4d36756 (18 MB Ghidra chunks). This adds ~110 MB more of the
same qualitative content. Ripgrep handles it in <1s.
Foundation for the named-retail workflow change in CLAUDE.md (next
commit). Plan at C:/Users/erikn/.claude/plans/ticklish-conjuring-cake.md
Phase A.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Visual-verified — Vitals window now shows three bars (HP/Stam/Mana)
with live values. Closes ISSUES.md #5; ~95% reading on Stam/Mana
traced to active buff multipliers, filed as #6.
Why the rewrite
The first attempt (commit d42bf57) routed PlayerDescription (0x0013)
through AppraiseInfoParser, trusting a misleading xmldoc claim.
Live diagnostics proved the format is wrong — ACE source
(GameEventPlayerDescription.WriteEventBody) hand-writes a body
distinct from IdentifyObjectResponse's AppraiseInfo: property
hashtables gated on DescriptionPropertyFlag, vector-flag-gated
attribute / skill / spell blocks, then a long options + inventory
trailer. Vitals only arrive via the attribute block at login.
Holtburger's events.rs:220-625 has the canonical client-side
unpacker; this commit ports the early-section walker through spells.
What landed
PlayerDescriptionParser.cs (new — 350 LOC):
Walks propertyFlags + weenieType, then property hashtables
(Int32/Int64/Bool/Double/String/Did/Iid) + Position table —
each gated on a property flag bit, header is `u16 count, u16
buckets`. Then vectorFlags + has_health + the attribute block
(primary attrs 1..6 = 12 B each, vitals 7..9 = 16 B with
`current`), then optional Skill + Spell tables. Stops cleanly
before the options/shortcuts/hotbars/inventory trailer (filed
as #7 — heuristic alignment search needed for gameplay_options).
PrivateUpdateVital.cs (new — 95 LOC):
Wire parsers for the GameMessage opcodes 0x02E7 (full snapshot)
and 0x02E9 (current-only delta), per holtburger UpdateVital +
UpdateVitalCurrent. WorldSession dispatches each to a session-
level event the GameWindow forwards into LocalPlayerState.
LocalPlayerState (full redesign):
VitalKind (Health/Stamina/Mana) + AttributeKind (six primary).
VitalSnapshot stores ranks/start/xp/current; AttributeSnapshot
stores ranks/start/xp with `Current = ranks+start` per
holtburger. GetMaxApprox computes the retail formula
vital.(ranks+start) + attribute_contribution
where the contribution is hardcoded from retail's
SecondaryAttributeTable: Endurance/2 for Health, Endurance for
Stamina, Self for Mana. Enchantment buffs not yet folded in
(filed as #6). VitalIdToKind now accepts both ID systems
(1..6 wire, 7..9 PD attribute block); AttributeIdToKind covers
primary attrs 1..6.
GameEventWiring:
PlayerDescription handler. Walks parsed.Attributes, routes
primary attrs (id 1..6) to OnAttributeUpdate and vitals
(id 7..9) to OnVitalUpdate. Player's full learned spellbook
also lands here. ACDREAM_DUMP_VITALS=1 traces every PD attribute
+ every PrivateUpdateVital(Current) opcode for diagnostics.
WorldSession:
Dispatch chain re-ordered — the diagnostic else-if for
ACDREAM_DUMP_OPCODES=1 was originally placed before
GameEventEnvelope.Opcode, which silently intercepted 0xF7B0 and
broke UpdateHealth dispatch when the env var was set. Moved to
the very end of the chain so it only fires for genuinely
unhandled opcodes. (Diagnostic-only regression; production
launches without the env var were unaffected.)
Test deltas
Added:
- PlayerDescriptionParserTests (6 — empty header, full attribute
block, partial flags, post-property-table walk, spell table)
- PrivateUpdateVitalTests (7 — fixture round-trip, vital ID
coverage, opcode rejection, truncation)
- LocalPlayerStateTests rewritten (20 — VitalIdToKind +
AttributeIdToKind theories, Endurance/Self formula coverage,
delta semantics, change events)
- GameEventWiringTests for PlayerDescription dispatch (2 —
end-to-end populate + spellbook feed)
Updated:
- VitalsVMTests rephrased onto the new OnVitalUpdate API.
Total: 765 → 817 tests passing.
Diagnostics
ACDREAM_DUMP_VITALS=1 — log every PD attribute extracted,
every 0x02E7/0x02E9 dispatch.
ACDREAM_DUMP_OPCODES=1 — log first occurrence of any unhandled
GameMessage opcode (now correctly placed at end of chain).
Visual verify
$env:ACDREAM_DEVTOOLS = "1"
dotnet run --project src\AcDream.App\AcDream.App.csproj -c Debug
Vitals window shows three bars; HP at 100%, Stam/Mana at ~95%
(the gap is buff enchantments — filed as #6 with the holtburger
multiplier+additive aggregator pattern as the reference for the
fix).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes ISSUES.md #5. The Vitals devtools window now draws three bars
(HP / Stamina / Mana) once the server sends the first PlayerDescription
(0x0013), instead of HP only. Built test-first per CLAUDE.md TDD rule —
16 new tests went red before the implementation went in.
New AcDream.Core.Player.LocalPlayerState (cache):
- {CurrentStamina, MaxStamina, CurrentMana, MaxMana} as uint? — null
until first received.
- StaminaPercent / ManaPercent: 0..1 fraction or null when either
field is missing or max is zero. Clamps to 1.0 if current > max
(server can briefly report this during buff transitions).
- OnPlayerDescription preserves any previously known good value when
an incoming field is null — partial profiles don't wipe state.
- Changed event for future subscribers.
GameEventWiring.WireAll:
- New optional 6th parameter: LocalPlayerState? localPlayer = null.
Existing 5-arg call sites still work; without the parameter the new
PlayerDescription handler still parses + feeds the spellbook but
skips the cache update.
- PlayerDescription (0x0013) shares AppraiseInfo wire format with
IdentifyObjectResponse (0x00C9) per AppraiseInfoParser docstring,
so the new handler reuses the existing parser and pulls
CreatureProfile.{Stamina, StaminaMax, Mana, ManaMax}.
- Player's full learned spellbook also lands here (previously only
item-scoped Identify responses fed the spellbook).
VitalsVM:
- Constructor adds optional LocalPlayerState? parameter (default null
keeps every existing caller compiling).
- StaminaPercent / ManaPercent now read through to LocalPlayerState
every access — no VM-side caching, so a server-side delta to the
cache surfaces next frame without any explicit refresh.
GameWindow:
- Public readonly LocalPlayer field alongside Combat / Chat / Items /
SpellBook so plugins + future panels can bind directly.
- WireAll call updated to pass LocalPlayer.
- VitalsVM construction passes LocalPlayer so the existing
VitalsPanel automatically picks up the two new bars.
Test counts:
- AcDream.Core.Tests: 550 → 561 (+11 LocalPlayerStateTests)
- AcDream.UI.Abstractions.Tests: 23 → 26 (+3 VitalsVM through-cache)
- AcDream.Core.Net.Tests: 192 → 194 (+2 PlayerDescription wiring)
- Total: 765 → 781
Build: 0 warnings, 0 errors.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Filed as the one explicit post-D.2a follow-up. VitalsVM returns float?
null for Stamina/Mana because absolute values only arrive in
PlayerDescription (0x0013) today and we parse-then-discard. A small
LocalPlayerState Core class that retains the parsed fields unblocks
two more progress bars in the existing Vitals window — no new wire
work needed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes Phase D.2a. Launch with ACDREAM_DEVTOOLS=1 now shows a live
ImGui "Vitals" window whose HP bar reads CombatState.GetHealthPercent
for the local player. Without the env var the branches are dead code,
no ImGui context is created, and behaviour is identical to before.
GameWindow hunks:
- fields: _imguiBootstrap / _panelHost / _vitalsVm + DevToolsEnabled
- init (OnLoad): construct bootstrap + host, register VitalsPanel
- GUID push: _vitalsVm?.SetLocalPlayerGuid(chosen.Id) at live-connect
- frame begin: _imguiBootstrap.BeginFrame(dt) after GL clear
- frame end: _panelHost.RenderAll(ctx) + _imguiBootstrap.Render() after debug overlay
- input gating: skip WASD when ImGui.GetIO().WantCaptureKeyboard
Backend pivot: Hexa.NET.ImGui → ImGui.NET + Silk.NET.OpenGL.Extensions.ImGui.
First-light integration with the Hexa backend crashed 0xC0000005 inside
Hexa.NET.ImGui.Backends.OpenGL3.ImGuiImplOpenGL3.InitNative. Root cause:
Hexa's native OpenGL3 backend resolves GL function pointers via GLFW or
SDL internally; with Silk.NET (which uses neither) the pointers are null
and the native code crashes on first use. The mitigation path was
already planned — the design doc's Risk section called a pivot to
ImGui.NET a "one-morning operation" — and that's exactly what happened.
- Packages: Hexa.NET.ImGui 2.2.9 + Hexa.NET.ImGui.Backends 1.0.18
→ ImGui.NET 1.91.6.1 + Silk.NET.OpenGL.Extensions.ImGui 2.23.0
- ImGuiBootstrapper: was static Initialize(gl)+Shutdown() wrapping
Hexa's OpenGL3 init; now an IDisposable wrapping Silk.NET's
ImGuiController instance which handles GL backend init + input
subscription in one go.
- SilkInputBridge.cs deleted (~190 LOC): ImGuiController subscribes
IKeyboard / IMouse events itself, we don't need a bespoke bridge.
- ImGuiPanelRenderer: ImGuiNET.ImGui.* calls instead of
Hexa.NET.ImGui.ImGui.*. Widget surface unchanged.
Boundary discipline is preserved — no panel imports ImGuiNET; only
ImGuiPanelRenderer does. The D.2b custom toolkit will implement the
same IPanelRenderer contract without touching panel code.
Out of scope (tracked for follow-up):
- Stam/Mana currently return float? null (VitalsVM). Absolute values
need LocalPlayerState + PlayerDescription (0x0013) parsing to be
stored rather than discarded — filed as a post-D.2a issue.
- Mouse-capture gating (WorldMouseFallThrough-style click-through
tests) — not needed until we add clickable inventory items.
Roadmap + memory + architecture doc + UI framework plan updated in the
same commit per CLAUDE.md roadmap-discipline rules. 753 tests pass
(550 Core + 192 Core.Net + 11 new UI.Abstractions), 0 build warnings.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Introduces `docs/ISSUES.md` as the tactical rolling list of known bugs +
small deferred features. Scope is strict: anything fitting in one-to-two
commits lives here; anything larger gets promoted to a Phase in the
roadmap. Complement, not replacement, of the strategic roadmap.
Schema per issue:
- Sequential integer ID (#1, #2, ...)
- Status (OPEN / IN-PROGRESS / DONE)
- Severity (HIGH / MEDIUM / LOW)
- Filed date, component, description, root cause, files, research,
acceptance
- DONE items move to "Recently closed" at the bottom with commit SHA
Four seeded issues from the 2026-04-24 sky-debug session:
#1 — Rain falls only to horizon, not to player (weather/particles)
#2 — Lightning visual not wired (dat-baked PES triggers; research done)
#3 — Client clock drifts from retail after ~10 min (periodic TimeSync)
#4 — Sky horizon-glow disabled (fog-mix skipped on sky meshes)
CLAUDE.md adds an "Issue tracking" section right after roadmap
discipline with the four maintenance rules: (1) scan OPEN at session
start, (2) add/close at session end, (3) reference IDs in commit
messages, (4) promote to a Phase if an issue grows. Calls out the
strategic-vs-tactical split so future sessions know when to reach for
the roadmap vs the issues file.
README.md gets one line pointing at `docs/ISSUES.md` below the existing
roadmap link for first-class discoverability.
No code changes; doc-only.
Landed the UI framework design in 2026-04-24-ui-framework.md yesterday;
this commit propagates the decisions across the documents that future
sessions touch first, so the three-layer pattern is discoverable without
re-reading the full plan.
Changes:
* NEW memory/project_ui_architecture.md — evergreen crib-sheet:
three-layer diagram, AcDream.UI.Abstractions contract, D.2a/D.2b
split, module layout, hard rules, why staged not pure-custom.
* CLAUDE.md: new paragraph describing the three-layer UI split, naming
AcDream.UI.Abstractions as the plugin-facing contract, pointing at
the full plan + memory crib.
* docs/architecture/acdream-architecture.md: new "UI Architecture"
companion-stack diagram after Layer 0-5 (doesn't renumber the main
stack), plus step 6a "UI tick" in Per-Frame Update Order.
* docs/plans/2026-04-11-roadmap.md Phase D tightened:
- D.2 split explicitly into D.2a (Hexa.NET.ImGui scaffold + abstraction
layer) and D.2b (custom retail-look backend, implements same contracts).
- D.3 AcFont / D.4 dat sprites / D.7 cursor flagged as D.2b dependencies.
- D.5 core panels / D.6 HUD flagged as abstraction-layer deliverables
— ship with D.2a, reskinned by D.2b.
- D.8 Sound marked superseded (shipped as Phase E.2).
- F.5 core panels + H.1 chat-window cross-references updated to say
they target AcDream.UI.Abstractions, unblocked by D.2a.
- Shipped-phases table untouched.
* docs/research/retail-ui/00-master-synthesis.md: scope note at top
clarifies the Keystone research is the D.2b (custom backend)
foundation, NOT where D.2a starts.
* ~/.claude/.../memory/MEMORY.md: one-line index entry pointing at the
new project_ui_architecture.md (so session auto-load surfaces it).
Zero code changes; doc-only. dotnet build stays green. All verification
greps pass (see plan file for exact checks).
Two-stage rollout, one stable abstraction layer:
1. Short-term: Hexa.NET.ImGui as the backend. Wire up in days, iterate
game logic (chat, inventory, vitals) in weeks. Looks like a debugger,
acceptable while we prove the interaction logic end-to-end.
2. `AcDream.UI.Abstractions` — ViewModels + Commands + `IPanel` /
`IPanelRenderer` interfaces. Backend-agnostic. Plugin API targets
this layer; plugins never see ImGui.
3. Long-term: custom retail-look backend using dat assets. Swap panel
by panel. ImGui stays forever as the `ACDREAM_DEVTOOLS=1` overlay.
The new doc (`2026-04-24-ui-framework.md`) captures:
- Full design of the three-layer split
- Why Hexa.NET.ImGui over ImGui.NET + Silk.NET.OpenGL.Extensions.ImGui
(AOT readiness, tracks upstream ImGui faster, cleaner native-lib
bundling)
- Alternatives considered and ruled out (Myra, Avalonia, NoesisGUI,
RmlUi, pure custom from day one)
- Implementation order (Sprint 1 vitals HUD → Sprint 2 interaction
panels → Sprint 3 plugin API → Sprint 4+ more panels → later
custom retail-look)
- Risks + mitigations and open questions deferred to implementation
Roadmap Phase D updated with a pointer to the new plan so future
sessions start from the latest strategy, not the original
all-custom-from-day-one Phase D description.
No code changes yet. Ready to start Sprint 1 when approved.
Iteration on the sky rendering pipeline to restore stars/moon visibility
at night and fix washed-out grey daytime clouds. Key fixes:
* sky.frag: disable fog-mix on sky meshes. Retail's keyframe FogEnd
(0..400m at midnight, up to 2400m during day) is calibrated for
terrain; sky meshes are authored at radii 1050-14271m which sits
past FogEnd universally, causing every sky pixel to saturate to
fogColor (dark navy). Stars, moon, dome texture all got
obliterated. The horizon-glow trade-off is noted in the shader
comment; research item to find retail's sky-specific fog range
later.
* SkyRenderer + sky.frag: promote rep.Luminosity into uEmissive so the
vertex lighting saturates properly for bright keyframes. Retail's
FUN_0059da60 non-luminous path writes rep.Luminosity into
material.Emissive via the cache +0x3c slot; we were instead using
it as a post-fragment multiply which could only dim, never brighten.
Net effect: daytime clouds now render saturated white, dome dims
correctly at night (rep.Luminosity=0.11 → Emissive=0.11), stars
and moon unchanged.
* terrain.vert: MIN_FACTOR 0.08 -> 0.0 per retail FUN_00532440 decompile
(DAT_00796344 ambient-floor = 0.0). Back-lit terrain now falls to
pure ambient rather than getting an 8% sun floor.
New research / tooling (no runtime impact):
* docs/research/2026-04-24-lambert-brightness-split.md — retail's
ambient-brightness formula pinned from PE .rdata read + live
RetailTimeProbe capture: effAmbBright = AmbBright + |sunDir| * 0.2
where scale constant 0x0079a1e8 = 0.2f exactly.
* docs/research/2026-04-23-lightning-real.md — research note on the
dat-baked PhysicsScript-driven lightning path (Rainy DayGroup has
explicit PES-triggered flash SkyObjects with 5ms time windows).
* Corrections stapled to sky-decompile-hunt-{B,C}.md: DAT_00842778 is
DirColor, DAT_0084277c is AmbColor (the hunt docs had the swap
backwards).
* tools/RetailTimeProbe/Program.cs: extended with pid=NNNN selector,
sky global probe (DirColor/AmbColor/AmbBright/sunDir/cache.amb),
and the 0x0079a1e8 scale-factor readout.
* tools/SkyObjectInspect/: throwaway dat-inspector built by the Opus
deep-dive agent. Identified GfxObj 0x010015EF as the stars layer
(A8R8G8B8 128x128 texture, 4% bright-pixel ratio).
* src/AcDream.App/Rendering/TextureCache.cs: per-texture alpha
histogram dump under ACDREAM_DUMP_SKY=1 for diagnosing "are the
clouds decoded with proper alpha" type questions.
README: rewrite to reflect current state (playable pre-alpha rendering
Dereth with animated characters, day-night cycle, weather, etc.)
instead of the stale "Phase 0 dat inventory only" description.
All 742 tests green.
User-observed regression 2026-04-23: acdream spawned rain particles
when retail showed no rain at the same server tick. Root cause: my
Phase 3e shortcut mapped DayGroup.Name = "Rainy" → WeatherKind.Rain →
rain particle emitter. That's not what retail does.
Parallel decompile research confirms:
- Agent A (2026-04-23-physicsscript.md): PhysicsScript runtime lives
at FUN_0051bed0 → FUN_0051bfb0, runs per PhysicsObj; sky calls it
from NOWHERE.
- Agent B (2026-04-23-sky-pes-wiring.md): FUN_00508010 (sky render
loop) never reads SkyObject.DefaultPesObjectId — the field is dead
at render time. Rain/snow particles in retail come from a separate
camera-attached weather subsystem that has NOT yet been located.
So the correct behavior is: DayGroup name should only drive
fog/ambient tone (via keyframes, already in the Snapshot path),
never spawn particle emitters. Any retail-faithful particle rain
belongs to a future phase once we find the camera-attached weather
subsystem driver.
Change: MapDayGroupNameToKind now maps all weathery substrings
(storm/snow/rain/cloud/overcast/dark/fog) → Overcast — fog-only
visuals, no particle spawn. Clear names stay Clear. The Rain, Snow,
Storm enum values remain and are still accessible via ForceWeather()
for debug overrides.
Tests updated (WeatherSystemTests): the name→kind theory now expects
Overcast for Rainy/Snowy/Stormy variants.
Also commits the four research docs from this session's parallel
hunt: PhysicsScript dat+runtime, sky↔PES wiring (negative finding),
lightning timer (negative finding — agent #3), fog on sky
(positive: retail applies fog to sky geometry).
NOTE on lightning: agent #3's research only ruled out the CLIENT-SIDE
RANDOM TIMER hypothesis for lightning. User confirms retail does have
visible lightning + thunder. A follow-up agent (#5, in flight as of
this commit) is hunting the real mechanism — PlayScript opcode,
SetLight PhysicsScript hooks, AdminEnvirons side effects, or the
weather-volume draw. This commit does NOT attempt to port lightning.
Build + 733 tests green.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Captures where we stand after Phase 4b and lays out the remaining
retail-faithful port work across four phases (5-8):
- Phase 5: PhysicsScript loader + runtime + sky lifecycle. Replaces
WeatherSystem's crude "DayGroup name contains Rainy → spawn rain"
shortcut with retail's actual PES-driven particle emission.
- Phase 6: Fog on sky meshes. The sky frag currently ignores fog
uniforms; retail's D3D fog applies to sky.
- Phase 7: Lightning flash trigger + thunder audio for storm keyframes.
- Phase 8: Weather / DayGroup crossfade (DAT_008427a9 / _DAT_008427b8
lerp) + AdminEnvirons override → fog crossfade.
User observation 2026-04-23 during Phase 4b verification: "Now it is
raining when it should not be." Root cause traced to the
SetKindFromDayGroupName string match firing rain particles on a "Rainy"
DayGroup regardless of whether that DayGroup actually has a visible
rain-emitting SkyObject. Proper fix requires porting PhysicsScript.
Also commits the earlier research from agent Q1-Q6:
`docs/research/2026-04-23-sky-material-state.md`.
Four parallel decompile agents are in flight as of this commit:
- PhysicsScript dat + runtime
- Sky↔PES wiring + emitter lifecycle
- Lightning + weather crossfade
- Fog on sky + vertex distance
Phase 5 implementation starts once those land.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Ran a live memory probe against retail acclient.exe (new tool:
tools/RetailTimeProbe/) to read the TimeOfDay struct at
DAT_008ee9c8 and compare against our computed values. The decompile
agent's identification of TimeOfDay+0x10 as "SecondsPerDay (int
copy)" turned out to be WRONG — the live value is **360**, which is
GameTime.DaysPerYear.
The retail FUN_00501990 LCG seed is:
seed = Year × (*+0x10) + DayOfYear
= Year × DaysPerYear + DayOfYear
= flat "total days since epoch" day-index
Our previous Phase 3c port passed 7620 (DayLength in ticks) as the
multiplier, producing seed=883,967 against retail's seed=41,807 —
completely different LCG outputs, completely different DayGroup
picks. That's why the user's retail kept showing stormy/rainy while
acdream showed sunny/clear (or vice versa) even after Phases 3c.1
and 3f aligned Year and DayOfYear.
Also confirmed by the probe:
- EpochBase / ZeroTimeOfYear = 3600 ✓ Phase 3f already correct
- BaseYear / ZeroYear = 10 ✓ DerethDateTime.ZeroYear
- Year=116, DayOfYear=47 ✓ our AbsoluteYear / DayOfYear
- SecondsPerDay float (+0x0C) = 7620 ✓ DayTicks
- SecondsPerYear = 2,743,200 ✓ YearTicks
One "finding that's not a fix": retail's +0x48 DayFraction is a
sub-period fraction (fraction through current day/night window)
NOT a full-day fraction. CurDayEnd - CurDayStart = 2857.5 = 0.375
of a day = 6 Dereth hours = night duration. Not relevant for our
keyframe bracket interpolation, which correctly uses a full-day
0..1 scale matching the SkyTime.Begin values. Documented in the
probe research doc so future work doesn't trip on it.
Changes:
- tools/RetailTimeProbe/ — new P/Invoke tool. Forced x86 target to
match retail's bitness so hardcoded DAT_xxxxxxxx addresses are
pointer-width-correct. Handles ASLR relocation via
Process.MainModule.BaseAddress.
- src/AcDream.App/Rendering/GameWindow.cs: RefreshSkyForCurrentDay
passes 360 (DaysInAMonth × MonthsInAYear) not 7620.
- src/AcDream.Core/World/SkyDescLoader.cs: ActiveDayGroup(ticks)
and DefaultDayGroup same.
- docs/research/2026-04-23-retail-memory-probe.md — full probe
results + decompile-agent correction.
- AcDream.slnx — add tools/ folder.
Build + 733 tests green.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Decompile agent located the retail DayGroup selection function at
FUN_00501990 (chunk_00500000.c:1276). It is a straight-line 32-bit
signed LCG — NOT a ChanceOfOccur-weighted CDF. Replaces the SplitMix64
approximation from Phase 3a.
Algorithm (verbatim from the decompile):
seed = year * secondsPerDay + dayOfYear // TimeOfDay+0x64/+0x10/+0x68
hash = seed * 0x6A42FDB2 + 0x8ABE1652 // signed 32-bit LCG
index = floor(dayGroupCount * (uint)hash / 2^32)
if (index >= dayGroupCount) index = 0 // float-rounding safety
Uniform over all DayGroups. Dereth's 20 groups all carry ChanceOfOccur=5.0
so uniform matches the statistical intent; the weighted walk Phase 3a
attempted is NOT what retail does. The SecondsPerDay multiplier is
load-bearing — without it, adjacent years would share adjacent LCG
seeds and divergence from retail would recur annually.
Result (this session's local ACE):
server: PY106 ColdMeet 17 MorntideAndHalf, ticks=291130073
→ year=106, dayOfYear=(106×0 + 17 across ColdMeet) via DerethDateTime
→ retail picker returns a deterministic uniform index from LCG.
Acdream and retail now agree on the pick for any (Year, DayOfYear)
since both drive from the same server PortalYearTicks.
Changes:
- src/AcDream.Core/World/DerethDateTime.cs: add Year(ticks) and
DayOfYear(ticks) helpers (match retail TimeOfDay+0x64 / +0x68).
- src/AcDream.Core/World/SkyDescLoader.cs:
- SelectDayGroupIndex signature: (year, secondsPerDay, dayOfYear)
instead of the flat dayIndex used by the SplitMix64 approximation.
- Body: retail LCG line-by-line port with decompile citations.
- ACDREAM_DAY_GROUP env var still overrides (for A/B verification).
- src/AcDream.App/Rendering/GameWindow.cs: RefreshSkyForCurrentDay now
feeds Year / DayOfYear / SecondsPerDay=7620 to the picker instead
of a flat dayIndex. Composite `year*360+dayOfYear` still tracked
internally as the day-change key for provider-rebuild idempotence.
- docs/research/2026-04-23-daygroup-selection.md committed with the
full decompile trail (new agent-produced research).
Build + 717 tests green. User visual verification (retail side-by-side)
next.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The uncommitted uTint=AmbientColor-for-alpha-submeshes experiment (from
the 2026-04-22 inference) dimmed the sky dome's baked gradient — a
user-verified visual regression. Reverting to the eeae83a baseline
(uTint=Vector4.One for every submesh) while we execute the proper
retail-verbatim port.
Research: three parallel decompile-hunt agents landed verifying
retail's ground-truth sky pipeline for the first time (prior audits
searched for stripped symbol names; the trail opened via the Region
dat-type-index 0x1c registration at chunk_00410000.c:12952). Key
retail functions now mapped in chunk_00500000.c:1097-7535:
- FUN_00501530: keyframe bracket-picker (with 1.0f wrap denominator)
- FUN_00501600: sun+ambient interpolator (sunVec = DirBright ×
(sin yaw·cos pit, cos yaw·cos pit, sin pit))
- FUN_00501860: fog interpolator
- FUN_00502820: SkyDesc::Unpack (2 doubles + DayGroup list)
- FUN_00502a10: build per-frame sky-object table
- FUN_00505f30: apply light state + per-cell AdjustPlanes relight
- FUN_005062e0: per-frame sky tick (throttled by LightTickSize)
- FUN_00508010: sky-object render loop (enqueues through the NORMAL
mesh pipeline via FUN_00514b90 — not a bespoke path)
Surprise findings:
- D3DRS_AMBIENT is set to 0 once at init and NEVER changes per-frame
(chunk_005A0000.c). The r12-inferred "clouds = texture × D3DRS_
AMBIENT" formula is falsified. Retail instead routes keyframe
AmbColor through per-vertex lighting on non-Luminous sky meshes
via _DAT_008682bc/c0/c4.
- Retail does NOT anchor the sky to the camera or use a separate
sky projection. Sky meshes live in world space and follow the
camera via scene-graph parent.
- FUN_00532440 (AdjustPlanes) re-lights every terrain cell on every
keyframe tick — the "terrain follows the sky" effect we don't yet
reproduce.
Phase 1 code change (this commit):
- src/AcDream.App/Rendering/Sky/SkyRenderer.cs: revert uTint to white
for all submeshes (the per-submesh blend split stays — sun gets
additive, clouds get alpha). Keep the `keyframe` parameter in the
signature for Phase 2 readiness. Comments now cite the retail
functions and reference docs instead of the (disproven) r12 formula.
- src/AcDream.Core/World/SkyDescLoader.cs: ACDREAM_DUMP_SKY=1 logs
the entire Region SkyDesc on load — DayGroups, SkyObjects, every
SkyTimeOfDay keyframe, and every SkyObjectReplace with RAW pre-/100
Transparent/Luminosity/MaxBright values so we can settle the unit
question empirically.
- src/AcDream.App/Rendering/Sky/SkyRenderer.cs: ACDREAM_DUMP_SKY=1
additionally logs each sky GfxObj's Surfaces and their SurfaceType
flags on first load, so we can identify which meshes carry the
Luminous bit (dome? sun? moon? stars?) vs which are lit.
- src/AcDream.App/Rendering/GameWindow.cs: passes the interpolated
keyframe to the sky renderer (kept — needed for Phase 2).
Research docs (pushed as part of this commit):
- docs/research/2026-04-23-sky-retail-verbatim.md: full synthesis
with retail function map, struct layouts, globals, pseudocode, and
a 4-phase port plan.
- docs/research/2026-04-23-sky-decompile-hunt-{A,B,C}.md: raw hunt
outputs.
- docs/research/2026-04-23-sky-references-crossref.md: WorldBuilder/
ACE/ACViewer/holtburger/Chorizite coverage.
- docs/research/2026-04-23-sky-dat-schema.md: full dat schema + unit
analysis.
- docs/research/2026-04-22-sky-lighting-decompile.md: prior agent's
(superseded) inference — kept for provenance.
Phase 2 will port Surface.Luminous-flag-aware per-vertex lighting for
sky submeshes once the dump resolves the open questions (Luminous-flag
distribution per Dereth sky mesh; _DAT_007a1870 scale constant value).
Build + 717 tests green.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Retail's Region dat stores SkyObjectReplace.Luminosity / Transparent /
MaxBright as percentages in the 0..100 range. Our shader expects
fractions in 0..1. We were passing raw values (luminosity up to 100)
straight into the sky fragment shader's rgb-multiplier:
rgb = sampled.rgb * uTint.rgb * uLuminosity;
At the "Sunny" DayGroup's noon keyframes (verified via live diag),
Luminosity = 100 → shader multiplied the cloud texture RGB by 100 →
min(rgb, vec3(1.2)) clamped all channels to 1.2 → pure white sky.
Also gave the dawn/dusk purple sky effect on top of the pale texture.
Fix: SkyDescLoader.ConvertTimeOfDay divides Luminosity, Transparent
and MaxBright by 100 when loading each SkyObjectReplace. The Rotate
field stays as degrees (values like 270° are genuine headings, not
percentages).
Transparent was accidentally surviving via a 0..1 clamp downstream,
but we fix it for consistency and so brightness-attenuating values
in the 0..99 range (partial fade during dawn/dusk) work correctly
instead of rounding to full-transparent.
WorldBuilder's SkyboxRenderManager does NOT apply these fields at
all — that's why they never hit this bug. Our port applies them for
per-keyframe day-night fades, so we needed the unit conversion.
Also picked up in this commit (incidental, already running):
- Sky render: per-submesh blend mode from TranslucencyKind.Additive
for sun/moon-style self-bright objects (Additive bit 0x10000).
Luminous flag 0x40 intentionally NOT mapped to additive — that
flag is on the sky dome + cloud sheets and making them additive
produced the previous "fully white" iteration of this bug.
- ToD default seed: DayTicks/16 (Midsong = hour 9 = true noon)
instead of DayTicks*0.5 which landed on Gloaming-and-Half (sunset)
due to DerethDateTime's +7/16 day-fraction offset. Pre-TimeSync
view now correctly starts at noon.
- Lightning flash: brighter white-blue (vec3(1.5,1.5,1.8)) instead
of dim grey; ceiling relaxed during flash so the strobe actually
blows out. Cadence (strike intervals, decay) unchanged.
- Saved docs/research/2026-04-21-sky-deep-audit.md with the
decompile+ACE+ACME+WorldBuilder research done to corner this bug.
Open follow-up (not fixed here): sky clouds are white at noon /
don't get the dusk/night purple tint. Our sky shader is fully unlit
— doesn't apply sun/ambient directional light like the terrain
shader does. AmbientColor in the keyframe data carries the right
tint (purple at midnight, magenta at dusk) but we pass
uTint = Vector4.One instead of the keyframe value. Next commit will
wire directional-sun + ambient into sky.frag so cloud meshes pick
up the time-of-day color.
All 717 tests green. User-confirmed: sky colors are now "much
better" after this change (previously fully white).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Ground-truth audit of acdream's animation pipeline against retail
decompile (chunk_*.c), cross-referenced line-by-line with our code.
Previous audit relied on ACE and got wiring claims wrong (said our
PlayAction path was orphaned when it's wired via OnLiveMotionUpdated).
Findings:
- PerformMovement dispatcher (FUN_00529a90) matches our MotionInterpreter.
- apply_current_movement cycle priority (FUN_00529210) matches our
OnLiveMotionUpdated sequencer path.
- Commands list → PlayAction wiring matches retail.
- Falling / Jump / Dead substate routing matches.
- Frame-timing epsilon + negative-speed playback matches.
The agent's "hit-react missing" claim turned out to be wrong: the
referenced FUN_0048d760 call passes 32-bit IDs shaped like MotionCommand
values but user-confirmed retail shows NO body animation on damage, so
vtable +0x9c is almost certainly emit-effect / play-sound / spawn-
particle — not a motion play. Not an animation gap.
Open follow-up: CreateObject initial Commands list is parsed but not
replayed when the entity hydrates (minor; rare case).
Not a follow-up: on-hit combat feedback (particles, damage numbers).
That's a separate feature, not an animation pipeline concern.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Update the roadmap's 'shipped' table with the G.1+ entry covering the
end-to-end visual integration (sky renderer, weather system, particle
renderer, UBO-backed shader lighting, server time sync) — not just the
data-plumbing layer that went out yesterday.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closed: jump server packet (002), facing direction (003), run speed (004).
New: collision penetration (005), corner stuck (006), missing trees (007).
All collision bugs stem from static-overlap detection instead of
swept-sphere — needs Transition restructure to use FindTimeOfCollision.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Two fixes for jump physics:
- Skip ground-snap when velocity Z > 0 (prevents immediate re-landing
at high framerates where per-frame Z delta < 0.05 snap threshold)
- Guard apply_current_movement velocity write behind OnWalkable check
(prevents MotionInterpreter.DoMotion from zeroing jump velocity on
every frame while airborne)
- Guard PlayerMovementController velocity replacement behind OnWalkable
(preserves momentum during airborne flight)
Jump works locally but server packet not yet sent (BUG-002).
Facing direction mismatch logged as BUG-003.
RunRate not verified as BUG-004.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Four-layer design for retail-faithful movement: speed from RunRate,
charged jump, BSP collision from decompiled CTransition, cell-based
object collision. Decompile-first methodology per CLAUDE.md.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Start with BUG-001: wrong cloth textures on characters (observed
during rendering rebuild verification).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The single most important document in the project. Defines:
Architecture: 6-layer stack (Platform → Renderer → Network → World →
Game Objects → Plugin API). The code is modern C#; the behavior
matches the retail client exactly.
GameEntity: the unified entity class that replaces the current
scattered state (WorldEntity + AnimatedEntity + guid dicts + player
controller). Every world object is a GameEntity with PhysicsBody +
AnimationSequencer + CellTracker + MotionInterpreter + AppearanceState.
Per-frame update order: Network → Streaming → Input → Entity tick
(motion → physics → collision → cell → animation) → Render → Plugin.
Execution plan (R1-R8):
R1: GameEntity refactor (unify scattered state)
R2: Thin GameWindow (extract to proper systems)
R3: CellBSP + wall collision (indoor transitions)
R4: Complete animation state machine
R5: Lighting from decompiled AdjustPlanes
R6: Server compliance (authoritative Z, keepalive)
R7: Interaction (doors, NPCs, chat, inventory)
R8: Plugin API completion (Lua macros)
Also updates CLAUDE.md to establish the architect role and reference
the architecture doc as the single source of truth.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Sprint 2 research for indoor transitions. Documents:
- ACE EnvCell.find_transit_cells: sphere-plane + BSP containment
- ACE SortCell/BuildingObj.check_building_transit: outdoor→indoor
- PortalSide semantics and portal polygon plane testing
- Gap analysis: acdream needs CellBSP, BldPortal list, VisibleCells
Key finding: full accuracy requires CellBSP (physics BSP from dat)
for sphere_intersects_cell. Current PortalPlane.IsCrossing is a valid
approximation. ACME's AABB PointInCell is an intermediate option.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Complete pseudocode translation of the retail AC client's animation
system, extracted from chunk_00520000.c. Covers:
- Sequence::update_internal (1021 bytes, the core frame advance loop)
- Sequence::advance_to_next_animation (node transitions)
- Sequence::append_animation (queue management)
- MotionTableManager::PerformMovement (1878 bytes, full state machine)
- AddAnimationsToSequence (transition link → sequence nodes)
- GetStartFramePosition / GetEndFramePosition (reverse playback support)
- AdjustNodeSpeed (negative speed = swapped start/end frames)
Key findings:
- framePosition is a 64-bit DOUBLE, not float
- Negative speedScale swaps startFrame↔endFrame at the node level
- update_internal handles both forward and reverse in one loop
- Frame triggers fire at every integer boundary crossing
- The keyframe slerp lives in the renderer, not the sequencer
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
ROOT CAUSE FIX for missing left-side animations.
The AC client's MotionTable has NO cycles for TurnLeft (0x000E),
SideStepLeft (0x0010), or WalkBackward (0x0006). The real client
calls adjust_motion() which remaps these to their right-side
equivalents with NEGATIVE speed before looking up the cycle. Then
multiply_framerate() swaps LowFrame↔HighFrame so the animation
plays backward.
Source: ACE MotionInterp.cs:394-428, decompiled FUN_005267E0.
Changes:
- AnimationSequencer.SetCycle: adds adjust_motion block that remaps
left→right with speed *= -1 (TurnLeft, SideStepLeft) or
speed *= -0.65 (WalkBackward = BackwardsFactor)
- LoadAnimNode: when framerate < 0, swaps Low↔High (matching the
decompiled multiply_framerate)
- GameWindow.UpdatePlayerAnimation: passes original animCommand to
SetCycle (sequencer handles remapping internally), keeps legacy
fallback for non-sequencer entities
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Complete decompilation of the retail Asheron's Call client using
Ghidra 12.0.4 + pyghidra headless. 22,225 of 22,226 functions
successfully decompiled in 75 seconds.
Output: docs/research/decompiled/ (54 files, 688,567 lines of C)
Key findings already identified:
- CLandBlockStruct::ConstructPolygons at chunk_00530000.c:2270
(split direction formula with 0x0CCAC033 constants)
- Motion command handlers at chunk_00510000.c (0x45000005 etc)
- Motion interpreter at chunk_00520000.c
- Portal space UI at chunk_004D0000.c and chunk_00560000.c
Next: identify CPhysicsObj, CMotionInterp, collision, and movement
functions by cross-referencing against ACE's C# port.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Exhaustive analysis of two working AC clients revealing three critical
findings that reshape acdream's movement system:
1. Server-authoritative Z: neither AC2D nor holtburger computes local
terrain Z for the player. AC2D sends keys, receives position. Holtburger
dead-reckons for smoothing but the server overrides.
2. Terrain split formula mismatch: AC2D and ACViewer's render path use
0x0CCAC033-based FSplitNESW; WorldBuilder (our source) uses a different
214614067-based physics formula. Our terrain mesh triangulation doesn't
match the real AC client's, causing Z mismatches on slopes.
3. Movement deduplication: MoveToState sent once per state change, not per
frame. AutonomousPosition heartbeat every 1 second.
Also adds AC2D to CLAUDE.md reference repos section.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Comprehensive spec for completing the physics/movement system:
CellPortal-based indoor/outdoor/room-to-room transitions via
sphere-plane intersection, multi-landblock boundary crossing,
momentum-preserving jump with gravity arc, portal-space state
machine for teleports, Setup.StepUpHeight hookup, and
scenery-on-road exclusion fix.
Replaces the current "outdoor heightmap sampler with disabled
indoor transitions" with the full world-navigation system AC's
client uses. 12 acceptance criteria, ~20 new unit tests planned.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Tab-toggled player mode with WASD ground walking, A/D + mouse
turning, Z/X strafing, Shift for run, third-person chase camera,
local walk/run/turn/idle animations, and outbound MoveToState +
AutonomousPosition server messages. Uses PhysicsEngine from B.3
for collision-resolved positions.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
5-task TDD plan: TerrainSurface (outdoor heightmap Z + cell ID),
CellSurface (indoor floor polygon projection via barycentric interp),
PhysicsEngine (top-level resolver with step-height + cell transitions),
GameWindow integration (populate from streaming), and roadmap update.
~16 new unit tests with fake data. No dat files or rendering needed.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Pure-computation collision engine for ground-based entity movement.
Three components: TerrainSurface (outdoor heightmap Z interpolation),
CellSurface (indoor EnvCell floor polygon projection), PhysicsEngine
(top-level resolver with step-height enforcement, outdoor/indoor
cell transitions via CellPortals, and gravity reporting).
Uses PhysicsPolygons from CellStruct for walkable surfaces with
brute-force polygon iteration (< 20 polys per cell). BSP tree
acceleration deferred — same collision fidelity, simpler code.
Standalone module with no rendering or networking dependencies.
~15-20 unit tests with fake data covering flat terrain, slopes,
stairs, wall rejection, and cell transitions. Integration with
the streaming system via ApplyLoadedTerrainLocked. Consumed by
Phase B.2 (player movement mode, separate spec).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
All Foundation sub-pieces now shipped:
A.1 ✓ Streaming landblock loader
A.2 ✓ Frustum culling (~160fps)
A.3 ✓ Background net receive thread
A.4 Folded into A.1 (deferred — DatCollection not thread-safe)
Phase A (Foundation) is complete. Next: Phase B (Gameplay).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
~160fps uncapped at 5×5 radius with per-landblock AABB culling.
Perf overlay in window title shows visible/total landblock ratio
so the culling impact is visible at a glance.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move A.1 from "ahead" to "shipped" per the roadmap discipline rule.
The shipped row notes that the loader currently runs synchronously
(the original async-worker design hit DatCollection's lack of thread
safety) and that the Channel-based outbox API is preserved so async
loading can return cleanly when Phase A.3 lands a thread-safe dat
wrapper. Pending-spawn list in GpuWorldState handles live spawn /
streaming races without dropping data.
Quick-lookup table updated:
- "Can't walk past the loaded 3×3 window" → A.1 FIXED ✓
- "Frame hitch crossing landblock boundary" → Phase A.3
(synchronous loader for now; async returns when DatCollection is
thread-safe)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Review follow-up from commit 11df793. Three fixes:
1. Visible semantics: StreamingRegion.Visible now strictly describes the
current (2r+1)×(2r+1) window, not window + hysteresis retainees.
Added a parallel Resident property exposing the actual loaded set
(window + hysteresis buffer). This matters because StreamingController
(next task) reads these to decide what to render vs what to unload;
conflating them in one set would have forced awkward post-processing
downstream.
2. Doc/code disagreement: updated the RecenterTo and RegionDiff doc
comments from "radius + 1" to "radius + 2" to match the actual
implementation (which is what the tests require). Also updated the
plan doc so future readers don't hit the same contradiction.
3. Edge-clamping test coverage: added a single-axis edge test
(cx=0, cy=50 → 15 entries) and an ID-encoding test (radius=0 at
0x12,0x34 → 0x1234FFFE) so a swapped-shift bug in EncodeLandblockId
or an asymmetric off-by-one would fail a test instead of passing
silently.
9 tests green, full suite regressions-free.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
