v1.6.0 — Genre De-leakage

Genre De-leakage: euca-moba crate split

Extracted 11 MOBA-specific modules from euca-gameplay into a new euca-moba crate. Agents building non-MOBA games no longer inherit MOBA vocabulary at compile time.

What changed

• New crate: euca-moba — roshan, creep_wave, tower_aggro, neutral_camp, hero, shop, item_active, building, building_systems, attributes, fog_of_war
• euca-gameplay now has 29 genre-neutral modules only — zero MOBA vocabulary
• Two reverse dependencies severed:
  • TowerAggroOverride → generic TargetOverride in combat.rs
  • StatGrowth moved from hero.rs → leveling.rs
• Code structure matches --scope metadata from v1.5.0

Migration

// MOBA types moved — add euca-moba dependency and update imports:
use euca_moba::{HeroRegistry, spawn_roshan, buy_item};

// Genre-neutral types unchanged:
use euca_gameplay::{Health, Team, AutoCombat, GameState};

1,729 tests passing across the full workspace.

Full Changelog: v1.5.0...v1.6.0

v1.5.0 — Progressive discovery: euca explain + discover --scope

Progressive disclosure for agents

Reduces the token tax on every agent session by 533 lines. SKILL.md shrinks from 638 → 105 lines; the rest of the engine's surface area is now discoverable on demand.

New: euca explain <topic>

Focused worked examples compiled into the CLI binary. Agents pull specifics only when they need them:

euca explain              # Lists all topics
euca explain fork         # Full counterfactual workflow
euca explain scenario     # Declarative game setup format
euca explain combat       # Auto-targeting fighters + towers

Topics: quickstart, entity, combat, rule, assert, fork, scenario. Each is ~60–100 lines of focused documentation with executable examples.

New: euca discover --scope

Agents building a puzzle game shouldn't see Roshan. The new --scope flag filters the CLI surface by genre:

euca discover --scope core       # Generic primitives only
euca discover --scope gameplay   # Genre-neutral gameplay
euca discover --scope media      # Audio, animation, visuals
euca discover --scope moba       # MOBA-specific vocabulary
euca discover --scope tools      # Infrastructure, offline tools

Scopes: `core`, `gameplay`, `media`, `moba`, `tools`, `all`. Each group in the JSON manifest carries a `scope` field so programmatic callers can filter without re-parsing.

SKILL.md: 638 → 105 lines

The old SKILL.md was a full agent reference. Most of its content is now discoverable on demand, so the new SKILL.md is:

• One-line description
• Bootstrap sequence (5 commands)
• Core primitives table with explain pointers
• The counterfactual loop (5-command agent workflow)
• Pointers to `discover` / `explain` / `GUIDE.md`

The old content is preserved as `SKILL_FULL.md` for humans who want the full reference.

Why this matters

Per the "Seeing Like an Agent" design analysis, every line an agent preloads is a token tax paid on every session. 1,284 lines of SKILL + GUIDE was euca's preload tax. Blog's insight: agents discover better than we credit them for when given the right tools. `euca discover` + `euca explain` provide those tools, so SKILL.md can get out of the way.

What's next

The remaining items from the first-principles plan:

• Empirical validation — actually run an agent against the fork + scenario stack
• Priority 2: Genre de-leakage — split `euca-gameplay` into core/gameplay-kit/moba-kit crates (complements today's `--scope` filter)
• Secondary A: SpawnRequest refactor — reflection-driven entity construction

Full Changelog: v1.4.0...v1.5.0

v1.4.0

Scenario primitive (new)

A single declarative JSON document that describes an entire game state — templates, entities, rules, assertions, camera, game mode — replacing the imperative 28-command MOBA setup pipeline.

Composes with fork (v1.3.0): save a baseline scenario, fork the world, apply a modified scenario to the fork, run a counterfactual. Main world untouched.

Format (v2)

{
  "version": 2,
  "name": "single-lane-moba",
  "templates": {
    "hero":   { "mesh": "cube", "health": 500, "team": 1, "combat": true },
    "minion": { "mesh": "cube", "health": 80,  "team": 2 }
  },
  "entities": [
    { "template": "hero",   "position": [0, 1, 0] },
    { "template": "minion", "position": [5, 1, 0], "overrides": { "health": 200 } }
  ],
  "rules": [
    {
      "when":  { "kind": "death" },
      "filter": "team:2",
      "actions": [{ "action": "score", "target": "source", "points": 1 }]
    }
  ],
  "assertions": [
    {
      "name": "team1_alive",
      "condition": { "type": "entity_count", "filter": { "type": "team", "team": 1 }, "min": 1 },
      "severity": "error"
    }
  ],
  "game": { "mode": "deathmatch", "score_limit": 10 }
}

API

• POST /scenario — apply scenario to main world (wipes + loads atomically)
• GET /scenario — export current main world as scenario JSON
• POST /fork/{id}/scenario — apply scenario to a named fork

CLI

euca scenario load tests/fixtures/moba.json
euca scenario save --out /tmp/snapshot.json

# Counterfactual workflow:
euca fork create buff-test
euca scenario apply-to-fork buff-test tests/fixtures/moba_buffed.json
euca fork step buff-test --ticks 300
euca fork probe buff-test --assertions team1_wins

Why this matters

Before scenarios, agents setting up a MOBA had to issue 28 ordered euca commands. Any one could fail mid-setup, leaving the world half-built; agents had to figure out where the failure happened and resume from a partial state. Scenarios make setup atomic and declarative: the engine receives the entire description and applies it as one operation.

The scenario format uses already-typed enums everywhere (GameAction, RuleCondition, AssertCondition) — no string DSL parsing, no positional fields. Schema validation happens at the JSON deserialization layer, not at runtime.

Breaking changes (small)

• RuleCondition variants refactored from tuple to struct form for clean serde output:
  • Timer(f32) → Timer { interval: f32 }
  • HealthBelow(f32) → HealthBelow { threshold: f32 }
  • Score(i32) → Score { threshold: i32 }
  • Phase(String) → Phase { phase: String }
• SpawnRequest gained Debug + Default derives (additive)

Tests

19/19 integration tests pass (15 existing + 4 new scenario tests):

• scenario_apply_to_main
• scenario_round_trip_serializes_and_loads
• scenario_apply_to_fork_does_not_affect_main
• scenario_template_overrides_merge

v1.3.0 — Fork primitive for counterfactual simulation

Fork primitive (new)

Agents can now deep-clone the main world into independent forks to answer "what if?" questions without touching the main simulation.

• World::clone() — deep-clones every entity, component, resource, and event via per-type clone_fn infrastructure. Component trait now requires Clone; resources require Clone at insert time.
• SharedWorld::fork(id) — create named forks from the main world. Forks share the single Schedule but own independent World state.
• 6 HTTP endpoints: POST /fork, GET /fork/list, DELETE /fork/{id}, POST /fork/{id}/step, POST /fork/{id}/probe, GET /fork/{id}/observe
• euca fork CLI subcommands mirror the HTTP API
• 13 new tests (7 integration + 6 ECS-level fork tests)

Agent workflow

euca fork create scenario-a          # clone main world
euca fork step scenario-a --ticks 300  # advance physics/combat/AI
euca fork probe scenario-a --ticks 0 --assertions hero-alive  # evaluate
euca fork observe scenario-a          # read state
euca fork delete scenario-a           # clean up

Other changes since v1.2.0

• GPU timestamp queries for per-pass profiling
• Diagonal inertia tensor and angular collision impulses
• Reflection-driven scene serialization (v3 format)
• Joint limits and motors in physics
• GPU terrain generation wired into renderer
• ExternalForce component and force/torque pipeline
• CI benchmark regression detection for PRs
• MetalFX temporal upscaling integration
• Water shader, GPU particles, Metal mesh shader path
• Post-process passes, audio, animation, AI behavior trees
• SSGI, UI, Lua scripting, meshlets, deferred, HLOD
• Fix double-tick bug in probe endpoint
• Fix Metal backend compile errors

Breaking changes

• Component trait now requires Clone (blanket impl updated)
• World::insert_resource<T> now requires T: Clone
• World::send_event<T> now requires T: Clone
• AudioEngine and ScriptEngine must be wrapped in Arc<Mutex<>> at insertion site (use shared_engine() helpers)
• Box<dyn ChunkLoader> is now Arc<dyn ChunkLoader> in streaming

What's Changed

• Add CI benchmark regression detection for PRs by @stainlu in #147
• Add force/torque accumulation pipeline by @stainlu in #148
• Wire GPU terrain generation into renderer by @stainlu in #149
• Add joint limits and motors to physics system by @stainlu in #150
• Add reflection-driven scene serialization (v3 format) by @stainlu in #151
• Add inertia tensor and angular collision impulses by @stainlu in #152
• Add GPU timestamp queries for per-pass profiling by @stainlu in #153

Full Changelog: v1.2.0...v1.3.0

v1.2.0

What's Changed

• refactor: make TextureStore generic over RenderDevice by @stainlu in #106
• refactor: make DecalRenderer generic over RenderDevice by @stainlu in #107
• refactor: make UiOverlayRenderer generic over RenderDevice by @stainlu in #108
• refactor: make IblResources generic over RenderDevice by @stainlu in #109
• refactor: make BindlessMaterialSystem generic over RenderDevice by @stainlu in #110
• refactor: make MotionBlurPass and DofPass generic over RenderDevice by @stainlu in #111
• refactor: make SsgiPass and SsrPass generic over RenderDevice by @stainlu in #112
• refactor: make TaaPass generic over RenderDevice by @stainlu in #113
• refactor: make GpuParticleSystem generic over RenderDevice by @stainlu in #114
• refactor: make PrepassPipeline generic over RenderDevice by @stainlu in #115
• refactor: make GpuDrivenPipeline generic over RenderDevice by @stainlu in #116
• refactor: make VelocityTextures generic over RenderDevice by @stainlu in #117
• refactor: make VolumetricFogPass generic over RenderDevice by @stainlu in #118
• refactor: make GBuffer and DeferredPipeline generic over RenderDevice by @stainlu in #119
• refactor: make ComputePipeline and GpuBuffer generic over RenderDevice by @stainlu in #120
• refactor: make PostProcessStack generic over RenderDevice by @stainlu in #121
• docs: update tech report, roadmap, changelog with RHI and Metal backend by @stainlu in #122
• feat: add native MSL shaders for Metal backend (PBR, shadow, sky) by @stainlu in #123
• feat: add GET /engine/gpu endpoint for GPU capability reporting by @stainlu in #124
• feat: Metal indexed drawing, capability detection, memoryless render targets by @stainlu in #125
• refactor: convert TextureStore methods to generic RenderDevice by @stainlu in #126
• refactor: convert TaaPass methods to generic RenderDevice by @stainlu in #129
• refactor: convert VolumetricFog, IBL, GpuParticles to generic RenderDevice by @stainlu in #131
• refactor: convert SSGI, SSR, DOF, MotionBlur to generic RenderDevice by @stainlu in #133
• refactor: convert GBuffer and DeferredPipeline to generic RenderDevice by @stainlu in #127
• perf: retain BFS queue across frames in transform propagation by @stainlu in #134
• fix: recover from poisoned locks in asset loader by @stainlu in #135
• fix: recover from poisoned locks instead of crashing in ECS by @stainlu in #136
• perf: allocation-free animation blend workspace by @stainlu in #137
• perf: retain frame vectors across physics ticks by @stainlu in #138
• perf: auto-compact RenderExtractor on 25% fragmentation by @stainlu in #139
• Add CONTRIBUTING.md with build, test, and workflow docs by @stainlu in #140
• Expand GUIDE.md into a proper getting-started tutorial by @stainlu in #141
• Enable SSR, DoF, and volumetric fog in MOBA demo by @stainlu in #142
• Wire MetalFX temporal upscaling into Renderer by @stainlu in #143
• Wire terrain_panel into editor and fix clippy warnings by @stainlu in #144
• Add procedural grass foliage to Tiled level viewer by @stainlu in #145

Full Changelog: v1.1.0...v1.2.0

v1.1.0

Full Changelog: v1.0.0...v1.1.0

v1.0.0

What's Changed

• v6: Deep engine audit — fix dead code, unwrap, quality issues by @stainlu in #58
• v7-1B: Remove RefCell from Renderer by @stainlu in #59
• v7-2A: Use SpatialIndex for O(n) combat targeting by @stainlu in #60
• v7-2BC: Eliminate unnecessary clones in combat/rules by @stainlu in #61
• v7-1A: Remove inline postprocess, always use PostProcessStack by @stainlu in #62
• Merge v7-1C editor refactor by @stainlu in #64
• v7-1C: Extract editor resource init into helpers by @stainlu in #63
• v7-2E: Optimize MOBA minion wave spawning by @stainlu in #65
• v7-2D: Add minion corpse cleanup by @stainlu in #66
• MOBA: Skip AutoCombat for PlayerHero entities by @stainlu in #67
• MOBA: Wire input system into editor by @stainlu in #68
• MOBA: Add PlayerHero + command system by @stainlu in #69
• MOBA: Player-controlled hero — full integration by @stainlu in #72
• MOBA: Add camera follow + edge panning system by @stainlu in #70
• MOBA: Add click-to-move + attack-target input by @stainlu in #71
• Add moba.json level file by @stainlu in #73
• Play mode: level file loading, hide editor UI, camera auto-follow by @stainlu in #77
• v0.8.0: Doc comments for remaining public APIs + version bump by @stainlu in #79
• Split euca-audio/lib.rs into focused modules by @stainlu in #80
• Add doc comments to euca-render public API by @stainlu in #82
• Add doc comments to euca-gameplay and euca-physics public APIs by @stainlu in #84
• refactor(euca-agent): replace wildcard re-exports with explicit named exports by @stainlu in #87
• Add missing Cargo.toml metadata to all crates by @stainlu in #78
• docs(euca-ecs): add comprehensive doc comments to public API by @stainlu in #81
• Replace println!/eprintln! with log:: macros by @stainlu in #83
• Review clippy suppressions: fix 5, document 11 by @stainlu in #86
• refactor: split euca-cli main.rs into command modules by @stainlu in #85
• Add audio clip unload, input stack pop guard, and editor entity lookup docs by @stainlu in #88
• Fix hardcoded projectile radius and terrain physics panic by @stainlu in #89
• Bind per-decal uniforms before each draw call by @stainlu in #90
• Fix fragile color parsing and document hardcoded game init by @stainlu in #91
• Remove dead code in App framework and simplify reflect JSON dispatch by @stainlu in #92
• Add turn & phase management system (Unit 10) by @stainlu in #95
• Add spatial grid for O(k) interest queries by @stainlu in #96
• Wire 8 missing v0.9.3 systems into headless server schedule by @stainlu in #98
• Implement stat resolution with equipment and status effect modifiers by @stainlu in #99
• Add v0.9.3/v0.9.4 pipeline integration tests by @stainlu in #100
• Add shop system — buy/sell items with gold, recipe combining by @stainlu in #104

Full Changelog: v0.3.0...v1.0.0

EucaEngine v0.3.0

Changelog

v0.3.0 (2026-03-21)

Rendering

• Deferred rendering path (G-buffer + lighting pass)
• Screen-space reflections (SSR)
• Volumetric fog with god rays
• SSAO (GTAO) + bilateral blur
• FXAA anti-aliasing
• Post-process stack (bloom, color grading, ACES tonemapping)
• GPU-driven rendering (draw indirect)
• HZB occlusion culling
• LOD system (screen-space mesh selection)
• Depth+normal pre-pass
• Clustered light culling (256+ lights)
• Render quality presets (Low/Medium/High/Ultra)
• Material system: transparency, emissive, metallic/roughness/AO textures
• Shader extraction to .wgsl files
• Foliage system (Poisson disk instancing)
• HLOD (hierarchical LOD)
• Decal system
• Particle render data pipeline
• Compute shader infrastructure

Physics

• Collision layers and masks
• Mass and inertia properties
• Scene queries (overlap_sphere, sweep_sphere, raycast_world)
• Collision events
• Character controller (capsule, ground detection, slopes, jumping)
• Vehicle physics (suspension, tires, engine, transmission)

Animation

• Animation blending and state machines
• Blend spaces (1D parametric)
• Root motion extraction
• Animation events
• Montage player
• Inverse kinematics (two-bone IK, FABRIK, look-at)

AI

• Behavior trees with blackboard
• Decorators, composites, action/condition nodes

Gameplay

• Role-aware targeting (heroes/minions/towers have different priorities)
• Persistent target tracking (CurrentTarget component)
• March direction (units advance toward enemy base)

Scale

• World streaming / chunk loading
• Spatial index (uniform grid queries)
• Prefab system (PrefabRegistry, spawn by name)

Networking

• Property replication with delta compression
• RPCs (ServerRpc, ClientRpc)
• Replication priority

Performance

• SIMD math (SSE2/NEON) for Vec3/Vec4/Mat4/Quat
• ECS query caching with generation invalidation
• Parallel system execution (ParallelSchedule)
• Frame profiler (per-system timing via CLI)
• Criterion benchmarks (ECS, physics, math)

Apple Silicon

• Metal TBDR render hints
• Unified memory SmartBuffer
• Optimized 32-thread compute dispatch

Scripting

• Lua scripting via mlua (hot reload, sandboxing, ECS bridge)

UI

• Runtime UI framework (anchored layout, flex, widgets)

Terrain

• Heightmap terrain (chunk LOD, splatting, physics, brush editing)

Audio

• Bus mixing hierarchy (Master/Music/SFX/Voice/UI)
• Reverb zones
• Sound priority and occlusion

Reflection

• Runtime field access (field_ref, field_mut, set_field)
• TypeRegistry for dynamic type creation
• JSON serialization

CLI / Agent API

• 70+ HTTP endpoints
• CLI commands: terrain, prefab, material, postprocess, fog, foliage, profile
• Render quality presets via CLI

Infrastructure

• 715+ tests
• Criterion benchmark suites
• CI: build-essential for mlua

What's Changed

• WU-15: Add behavior trees and blackboard (euca-ai crate) by @stainlu in #7
• WU-02: Add collision layers, mass, and scene queries by @stainlu in #14
• WU-05: Add SIMD math intrinsics (SSE2/NEON) by @stainlu in #8
• WU-11: Audio mixing buses, reverb zones, priority, occlusion, and fading by @stainlu in #9
• WU-08: SSAO and post-processing pipeline by @stainlu in #10
• WU-09: Add LOD system — screen-space mesh selection by @stainlu in #4
• WU-14: Add compute shader infrastructure by @stainlu in #3
• WU-16: Add terrain system — heightmap, chunked LOD, splatting, physics by @stainlu in #5
• Update docs for v3: 24 crates, expanded feature table, CI for mlua by @stainlu in #15
• Integrate all v3 features into main by @stainlu in #16
• WU-19: Add vehicle physics by @stainlu in #17
• WU-13: Add prefabs and spatial index by @stainlu in #18
• WU-03: Add character controller by @stainlu in #19
• WU-12: Add particle rendering data pipeline by @stainlu in #20
• WU-18: Add inverse kinematics by @stainlu in #21
• WU-17: Add decal system by @stainlu in #22
• WU-10: Add network property replication and RPCs by @stainlu in #23
• WU-12: Add particle render data pipeline by @stainlu in #24
• E1: Add built-in frame profiler by @stainlu in #25
• B1: Integrate post-process stack into renderer by @stainlu in #26
• A1: Wire all new systems into editor tick loop by @stainlu in #27
• D1: Metal-optimized render path for Apple Silicon by @stainlu in #28
• C3: Add foliage system with instanced vegetation by @stainlu in #31
• B6: Extract shaders to .wgsl files by @stainlu in #32
• E2: Add ECS query caching by @stainlu in #33
• A2: Add CLI/HTTP endpoints for new systems by @stainlu in #34
• Merge remaining batch 1: unified memory + occlusion culling by @stainlu in #35
• D3: Unified memory optimization for Apple Silicon by @stainlu in #29
• C1: Add occlusion culling with hierarchical Z-buffer by @stainlu in #30
• A3: Update SKILL.md with all new commands by @stainlu in #36
• C4: Add hierarchical LOD system by @stainlu in #37
• B4: Add volumetric fog and god rays by @stainlu in #38
• B3: Add screen-space reflections by @stainlu in #39
• E3: Add parallel system execution by @stainlu in #40
• C2: Add world streaming / chunk loading by @stainlu in #41
• B5: Add GPU-driven rendering with draw indirect by @stainlu in #42
• D2: Apple GPU compute optimization by @stainlu in #43
• B2: Add deferred rendering path by @stainlu in #44
• v5-1A: Enable SSAO/FXAA by default, add render quality presets by @stainlu in #45
• v5-1G: Add clustered light culling for Forward+ by @stainlu in #48
• v5: Wire all rendering systems into pipeline by @stainlu in #52
• v5-1F: Add depth+normal pre-pass by @stainlu in #46
• v5-1E: Wire foliage instancing into render pipeline by @stainlu in #47
• v5-1D: Wire HZB occlusion culling into renderer by @stainlu in #49
• v5-1H: Add render quality presets by @stainlu in #50
• v5-1C: Wire volumetric fog into render pipeline by @stainlu in #51
• v6-1C: Add math SIMD benchmark suite by @stainlu in #53
• v6-1B: Add physics benchmark suite by @stainlu in #54
• v6-1D: Integrate profiler into editor + CLI by @stainlu in #55
• v6-1A: Add ECS benchmark suite by @stainlu in #56
• Add CHANGELOG.md for v0.3.0 release by @stainlu in #57

New Contributors

• @stainlu made their first contribution in #7

Full Changelog: v0.2.0...v0.3.0

v0.2.0 — MOBA Release

EucaEngine v0.2.0

Major release: full gameplay systems + working MOBA demo.

What's New Since v0.1.0

New Engine Systems

• MSAA 4x anti-aliasing
• Audio (spatial, kira)
• Animation (glTF skeletal)
• Particles (CPU emitters)
• Navigation (grid navmesh, A* pathfinding)
• Economy (gold, bounty, gold-on-kill)
• Leveling (XP, auto stat boost, max level 18)
• Abilities (Q/W/E/R, cooldowns, mana, effects)
• Entity Roles + targeting priority
• AttackStyle (Melee/Stationary)
• Patrol-combat hybrid AI
• SpawnPoints + Respawn
• Diagnostics (euca diagnose, euca events)

Architecture Improvements

• Routes split from 1738-line monolith into 10 modules
• Physics decoupled from Collider (entities move without colliders)
• Kinematic body type for gameplay entities
• Persistent entities survive world reset
• RuleSpawnEvent bridges gameplay→render for visible spawned entities

MOBA Demo

Complete 1-lane MOBA: heroes, towers, minions, economy, leveling.

./scripts/moba.sh

Quickstart

git clone https://github.com/stainlu/euca-engine.git && cd euca-engine
cargo build -p euca-editor --example editor -p euca-cli
cargo run -p euca-editor --example editor &
./scripts/moba.sh

Full Changelog: v0.1.0...v0.2.0

v0.1.0

Full Changelog: https://github.com/stainlu/euca-engine/commits/v0.1.0