Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
226 commits
Select commit Hold shift + click to select a range
4c8036b
Prepare public inference and evaluation release
iChubai Sep 8, 2026
832ce01
Refresh bilingual model docs and fix docs startup and mobile navigation
iChubai Sep 8, 2026
5f2c606
Polish the model command builder into one run card and drop unused De…
iChubai Sep 11, 2026
eae83d9
Refresh supported-model homepages so generated pages can show paper f…
iChubai Sep 11, 2026
90979c7
Make catalog YAML the only official-runner table and keep a single ro…
iChubai Sep 13, 2026
3c124bc
Inject docs theme bootstrap through useServerInsertedHTML and tighten…
iChubai Sep 13, 2026
7df1f8c
Resolve Vid2World eval lists and YAML through the packaged config root.
iChubai Sep 13, 2026
c182a85
Point Evoke and Solaris helper scripts at packaged runtime configs.
iChubai Sep 13, 2026
7f8749f
Probe attention backends by walking package paths instead of importin…
iChubai Sep 17, 2026
12cbc20
Keep image metrics on native CPU backends and pin torch-fidelity impo…
iChubai Sep 17, 2026
f69f4b3
Share sample scoring across online and offline runs and resume from v…
iChubai Sep 17, 2026
32e6fb6
Integrate inference-only interactive world model routes
iChubai Sep 18, 2026
dd705fe
Complete interactive model contracts across Workspace CLI and TUI
iChubai Sep 18, 2026
8647440
Fix inference failures found in multi-model GPU validation
iChubai Sep 18, 2026
f6d7581
Clarify SolarWM source-length requirements for long inference
iChubai Sep 18, 2026
42573b8
Fix Matrix and LongCat inference failures found in GPU validation
iChubai Sep 18, 2026
4d59302
Fix inference failures found in older model GPU validation
iChubai Sep 18, 2026
9eaa653
Fix WorldCam inference and flag missing public runtimes
iChubai Sep 18, 2026
342bfbc
Restore inference-only GameCraft and minWM HY runtimes
iChubai Sep 18, 2026
4c2b229
Fix FlashWorld camera controls and restore InSpatio inference inputs
iChubai Sep 19, 2026
ac6844c
Fix conditioned inference and packaged camera runtimes
iChubai Sep 19, 2026
1a53ff8
Integrate HyperFlow inference and fix GPU demo contracts
iChubai Sep 19, 2026
4a53509
Fix Cosmos sampling windows and MAGI-2 inference parity
iChubai Sep 19, 2026
84c0ef2
Align Cosmos Transfer edge preprocessing with upstream
iChubai Sep 19, 2026
303e7d5
Fix inference offload, video frame counts and point cloud I/O
iChubai Sep 19, 2026
3cab072
Restore inference runtimes and fix release model packaging
iChubai Sep 19, 2026
9f65224
Fix inference parameter handling and native video chunk boundaries
iChubai Sep 19, 2026
ab936f3
Fix VGGT frame results and Omega inference configuration
iChubai Sep 19, 2026
9f509f6
Verify public inference routes and record GPU coverage
iChubai Sep 23, 2026
27f4da2
Ignore private test sources
iChubai Sep 23, 2026
4d69bba
Keep local validation artifacts private
iChubai Sep 23, 2026
b45d965
Limit public scripts to supported workflows
iChubai Sep 23, 2026
2024fc9
Fully ignore private tests
iChubai Sep 23, 2026
2ba593d
Polish public README and fix broken links
iChubai Sep 23, 2026
0ddbcc2
Remove ViewCrafter integration and trim public demos
iChubai Sep 23, 2026
e023ec0
Fix LiveWorld asset resolution and DINO-WM preflight
iChubai Sep 23, 2026
ac888a5
Make MME-VLA JAX inference imports compatible
iChubai Sep 23, 2026
1506a46
Mark LIBERO-Para public inference as blocked
iChubai Sep 23, 2026
3b9403a
Correct blocked Emu3.5 and WonderWorld availability
iChubai Sep 23, 2026
f8843e0
Fix WonderJourney depth preprocessing import
iChubai Sep 23, 2026
a2a3ab3
Report runtime model readiness from actual plans
iChubai Sep 23, 2026
447bb0b
Clarify missing official policy checkpoint diagnostics
iChubai Sep 23, 2026
73f114c
Load Evoke checkpoint components with compatible imports
iChubai Sep 23, 2026
b31f95a
Load MiniMax H3 across GPUs and restore faithful video output
iChubai Sep 24, 2026
892c486
Run Open-Sora from pinned in-tree source with safe T5 cache conversion
iChubai Sep 24, 2026
35f36ae
Run Emu3.5 interleaved generation through the official runtime
iChubai Sep 24, 2026
4665d3f
Remove unsupported model entries and refresh public catalog docs
iChubai Sep 24, 2026
52dfb43
Report HY-World renderer source and scene prerequisites accurately
iChubai Sep 24, 2026
ee694b5
Fix Pandora plan runtime path and WorldGrow checkpoint discovery
iChubai Sep 24, 2026
0776c17
Honor explicit Being-H05 legacy metadata variant selection
iChubai Sep 24, 2026
2fdcf51
Add verified LTX-2 IC-LoRA video control bindings
iChubai Sep 24, 2026
9de296c
Validate XGEN-JING H3 codecs and GPU output
iChubai Sep 24, 2026
c071bbe
Require typed control inputs for LTX V2V CLI
iChubai Sep 24, 2026
03b6738
Record XGEN-JING and pi05 checkpoint-backed validation
iChubai Sep 24, 2026
89cde59
Stabilize HunyuanVideo I2V latent initialization
iChubai Sep 24, 2026
ef8da7c
Validate Voyager and LingBot public video runners
iChubai Sep 24, 2026
a996ebd
Reject invalid video defaults and unavailable LaST-R1 route
iChubai Sep 24, 2026
2a0305a
Preserve requested frame count in MAGI low-FPS decode
iChubai Sep 24, 2026
38a845a
Align LAPA and Track-Anything runtime assets
iChubai Sep 24, 2026
1131726
Load RoboFlamingo policy weights and dual-camera inputs correctly
iChubai Sep 24, 2026
4fccde4
Remove canceled model media and refresh public catalog data
iChubai Sep 24, 2026
7325463
Preserve exact frame count across WorldCrafter resume chunks
iChubai Sep 24, 2026
47068ee
Route Track-Anything weights and text prompts to its runtime
iChubai Sep 24, 2026
d832c98
Remove unsupported LeRobot policy integrations
iChubai Sep 24, 2026
a8c9a47
Support verified WorldCrafter Fast resume
iChubai Sep 24, 2026
20e8c1e
Remove DROID-W catalog integration
iChubai Sep 24, 2026
34b57ab
Enable OmniVinci video question answering
iChubai Sep 24, 2026
b74c1cd
Preserve UniAnimate demo rewrite compatibility
iChubai Sep 24, 2026
cf2ca9d
Reject remote DVLT checkpoints in offline mode
iChubai Sep 24, 2026
5280caa
Enable Qwen2.5 Omni video question answering
iChubai Sep 24, 2026
676e414
Enable Qwen2.5 VL video question answering
iChubai Sep 24, 2026
1b8b56d
Keep official video plans aligned with invocation overrides
iChubai Sep 24, 2026
d0b1def
Honor official video runtime default overrides
iChubai Sep 24, 2026
8cab94d
Expose TeaCache toggle for Wan Fun camera validation
iChubai Sep 24, 2026
d08e3c7
Support forward FramePack F1 inference
iChubai Sep 24, 2026
0b4aa36
Honor Echo Infinity inference seed and frame count
iChubai Sep 24, 2026
e7df340
Forward HyDRA guidance scale to official inference
iChubai Sep 24, 2026
6bb5110
Validate HyDRA on packaged official example
iChubai Sep 24, 2026
abdcc6b
Integrate LaST-R1 offline action inference
iChubai Sep 24, 2026
05b71e0
fix(vchitect): align sampling and T5 conditioning with release
iChubai Sep 24, 2026
cf2153f
Use released Ctrl-World sampling default
iChubai Sep 24, 2026
8abb9cf
Correct RDT-1B local checkpoint availability
iChubai Sep 24, 2026
69d7008
Align Echo-Memory camera action defaults with upstream
iChubai Sep 24, 2026
dc64a5f
Repair pi0 WorldFoundry preflight config and asset status
iChubai Sep 24, 2026
752dc65
Align EgoWM 3-DoF actions and guidance with upstream
iChubai Sep 24, 2026
abb2159
Respect CUDA visibility in SCOPE subprocess routing
iChubai Sep 24, 2026
e012dbd
Honor explicit negative prompt in SAMA CLI
iChubai Sep 24, 2026
2cc1cae
Remove RTC passthrough masquerading as model inference
iChubai Sep 24, 2026
d714cec
Reflect staged Bernini GPU inference in runtime profiles
iChubai Sep 24, 2026
0f7c5ee
Route original GR00T-N1 through isolated official worker
iChubai Sep 24, 2026
4c714f0
Align Bernini renderer defaults with official CLI
iChubai Sep 24, 2026
c50f7a7
Match Vchitect inference precision to official pipeline
iChubai Sep 24, 2026
ee3a70e
Align Cosmos Predict2.5 sampling with official inference
iChubai Sep 25, 2026
129d90a
Align Cosmos Predict2.5 2B with official Karras inference
iChubai Sep 25, 2026
26b70fd
Reject unavailable residual AdaLN fusion option
iChubai Sep 25, 2026
879d892
Match GigaWorld text context to upstream limit
iChubai Sep 25, 2026
d40b31a
Use native Gemini video API for Veo 3.1
iChubai Sep 25, 2026
e272a62
Show asset hints in missing-path diagnostics
iChubai Sep 25, 2026
0328a3e
Refresh video asset readiness and MAGI-1 lookup
iChubai Sep 25, 2026
8a5e3b1
Preserve LTX-2.5 virtualenv interpreter in runtime plans
iChubai Sep 25, 2026
f347449
Align FastVideo CausalWan sampling and image canvas with upstream
iChubai Sep 25, 2026
407cdd0
Report actual MultiWorld environment observation source
iChubai Sep 25, 2026
0a6a8a0
Document verified 204-frame Step-Video inference
iChubai Sep 25, 2026
33c19f2
Correct Step-Video text-only input documentation
iChubai Sep 25, 2026
edfb262
Match official TesserAct RGB preprocessing
iChubai Sep 25, 2026
5e291f4
Reject unsupported Gamma-World bidirectional frame counts
iChubai Sep 25, 2026
76584b1
Document verified Voyager 50-step output length
iChubai Sep 25, 2026
1d4af9a
Clarify verified Oasis camera pitch control
iChubai Sep 25, 2026
0eccc8d
Normalize ATI track visibility across input formats
iChubai Sep 25, 2026
9b1a2cd
Route original GR00T-N1 cameras by official data config
iChubai Sep 26, 2026
a545b6b
Use official physical pose conditioning for Ctrl-World
iChubai Sep 26, 2026
cea04e2
Require paired physical conditions for EgoWM 25-DoF
iChubai Sep 26, 2026
c253b81
Support official Genie Envisioner camera histories
iChubai Sep 26, 2026
64c2537
Support paired EgoWM 3-DoF navigation actions
iChubai Sep 26, 2026
905d9d5
Support paired LangTable histories in HMA
iChubai Sep 26, 2026
92ff8ac
Route HMA CLI variants to their matching checkpoints
iChubai Sep 26, 2026
6d7226e
Honor released Gamma-World rollout length and cache capacity
iChubai Sep 26, 2026
48c1d6a
Forward Spatia foreground masks with source-size checks
iChubai Sep 26, 2026
e46a0a9
Integrate pinned UniWorld-View image and video inference
iChubai Sep 26, 2026
c41dd53
Guard UniWorld-View downloads and isolate its Python runtime
iChubai Sep 26, 2026
92706f7
Honor UniWorld-View seed and preserve upstream imports
iChubai Sep 26, 2026
6818f83
Validate UniWorld-View GPU routes and expose fixed sizing
iChubai Sep 26, 2026
909a864
Validate model integrations and align official inference paths
iChubai Sep 27, 2026
16cc493
Polish site navigation and model documentation
iChubai Sep 27, 2026
ae6bfd4
Align docs quality checks with curated pages
iChubai Sep 27, 2026
19eba75
Validate new checkpoints and repair inference routes
iChubai Sep 27, 2026
198abae
Validate DINO-WM, WildDet3D, and MolmoBot Pi0 runtimes
iChubai Sep 28, 2026
89c2411
Validate benchmark integrations against pinned official runtimes
iChubai Sep 28, 2026
ba03658
Remove WorldAtlas Arena benchmark integration
iChubai Sep 28, 2026
61cd523
Correct benchmark import evidence and preflight metadata
iChubai Sep 28, 2026
e51eb50
Exclude benchmark probe scripts from public release
iChubai Sep 28, 2026
ab241ca
Remove redundant Studio asset wrapper
iChubai Sep 28, 2026
aabe783
Trim redundant public helper scripts
iChubai Sep 29, 2026
9c584ec
Fix public release checks and consolidate runtime docs
iChubai Sep 30, 2026
ef38dd0
Keep release packaging audit separate from lint
iChubai Sep 30, 2026
f6a83d7
Harden inference lifecycle and publish unified tests
iChubai Oct 2, 2026
10456e9
Consolidate geometry runtimes into 3D base models
iChubai Oct 2, 2026
ce6301b
Guard 3D inference with real checkpoint replays
iChubai Oct 2, 2026
b0a26e0
Organize core infrastructure by responsibility
iChubai Oct 2, 2026
1b48037
Add layered inference regression gates and fix failure recovery
iChubai Oct 2, 2026
073c50c
Remove unified Gradio Studio frontend
iChubai Oct 2, 2026
dfdca11
Add affected-model planning and scheduled GPU regression
iChubai Oct 2, 2026
3a0ed7a
Preserve prepared tool paths in GPU cron jobs
iChubai Oct 2, 2026
8a82ca4
Preflight accepted runtimes and preserve per-model environments
iChubai Oct 2, 2026
2ccde75
Extend video and world-model regression gates and seed LTX decoding
iChubai Oct 2, 2026
e74dc27
Match exact plain and setuptools-activated GPU environment inventories
iChubai Oct 2, 2026
69e330b
Respect deterministic convolution policy in DreamX realtime sessions
iChubai Oct 2, 2026
36ab841
Restore terminable GPU holders after interrupted regression cleanup
iChubai Oct 2, 2026
e4c1a48
Fix causal attention padding and CUDA graph output ownership
iChubai Oct 2, 2026
3132fe5
Consolidate license texts and notices in THIRD-PARTY-NOTICES
iChubai Oct 2, 2026
d5d5547
Gate inference benchmarks on numerical parity and real CUDA execution
iChubai Oct 2, 2026
30fb47e
Fuse FP8 FFN activation quantization and invalidate stale CUDA captures
iChubai Oct 2, 2026
a32fc59
Tune mixed convolution layouts before VAE compilation
iChubai Oct 2, 2026
1c80a06
Optimize MG2 causal projections and cache bookkeeping with parity tests
iChubai Oct 2, 2026
0f316ac
Coordinate streaming step admission and verify fullgraph FP8 FFNs
iChubai Oct 2, 2026
472fdb7
Add explicit scaled cuDNN FP8 attention with numerical and capture co…
iChubai Oct 2, 2026
b23eeeb
Flatten visual world-model packages into visual_generation
iChubai Oct 2, 2026
bd85147
Overlap MG2 streaming decode with clean context refresh safely
iChubai Oct 2, 2026
70d1015
Add opt-in dense Triton TMA attention with replay and numerical contr…
iChubai Oct 2, 2026
6dc4ff8
Retain world-model inference resources in public distributions
iChubai Oct 2, 2026
4d7283e
Preserve owned CUDA frames through NVENC and survive bounded streamin…
iChubai Oct 2, 2026
6ebd07c
Require Triton 3.5 context-local allocation for explicit TMA attention
iChubai Oct 2, 2026
4870881
Gate optimized inference on full-checkpoint parity and reproducible C…
iChubai Oct 2, 2026
aef9ea1
Prove model relocations strictly and preserve lazy no-grad discovery
iChubai Oct 2, 2026
0768679
Preserve MG2 image conditioning through causal rollouts
iChubai Oct 3, 2026
51cbe27
Expose completed diffusion latents before decoding across runner stra…
iChubai Oct 3, 2026
cf0611d
Bound video conditioning reads while preserving full-input normalization
iChubai Oct 3, 2026
b587425
Validate bounded Reason1 caching and explicit Cosmos inference parall…
iChubai Oct 3, 2026
f5cf70d
Isolate multistep sampling histories and reuse unchanged DreamX promp…
iChubai Oct 3, 2026
7742010
Own resident preload and cleanup after request failure or bounded shu…
iChubai Oct 3, 2026
61230b0
Keep QKV receipt identities constant during fullgraph recompilation
iChubai Oct 3, 2026
c484ac9
test: extend native replay recipes and numerical contract dependencies
iChubai Oct 3, 2026
4c85013
test: require complete CPU contracts and trusted historical inference…
iChubai Oct 3, 2026
679c677
fix: safely snapshot committed sources and prove verification-only ch…
iChubai Oct 3, 2026
216e7d2
test: wait for detached supervisor exit after completion receipt
iChubai Oct 3, 2026
5d36693
ci: keep evidence dispatch skips separate from required checks
iChubai Oct 3, 2026
20069cb
feat: integrate InSpatio World 1.5 with shared Wan base models
iChubai Oct 3, 2026
7081f6c
test: bind recipe replay selection to immutable case diffs
iChubai Oct 3, 2026
a76dafc
fix: classify runtime dependencies and require focused checkpoint rep…
iChubai Oct 3, 2026
4591cbb
fix: improve TUI swan logo rendering and sizing
iChubai Oct 3, 2026
4821814
test: require real runtime coverage and wire Oasis checkpoint replay
iChubai Oct 3, 2026
53cbcd8
Merge commit '4591cbb0831c32253d9e9397d6477b52d01c418f' into HEAD
iChubai Oct 3, 2026
f4569b5
fix(kernels): preserve eager rounding in functional diffusion fusion
iChubai Oct 3, 2026
df211b7
Merge TUI swan logo fix into release
iChubai Oct 3, 2026
5c9b58b
Merge remote TUI release while preserving strict replay coverage fixes
iChubai Oct 3, 2026
9bfe870
feat(diffusion): add reversible acceleration plugins and scoped atten…
iChubai Oct 3, 2026
8ed2a48
test(diffusion): record quality-gated H100 acceleration diagnostics
iChubai Oct 3, 2026
c8428c0
fix(kernels): preserve eager rounding in functional diffusion fusion
iChubai Oct 3, 2026
93a8e42
fix(kernels): preserve vendor LayerNorm bits in accelerated AdaLN
iChubai Oct 3, 2026
8b44f92
fix: preserve the official OpenEnvision logo in TUI
iChubai Oct 3, 2026
85f38c0
Merge latest release into official logo fix
iChubai Oct 3, 2026
6e01b99
fix: sharpen the official logo in text terminals
iChubai Oct 3, 2026
92727ca
feat(diffusion): add guarded projection precision, KV fusion and matc…
iChubai Oct 3, 2026
036f954
merge: integrate strict diffusion kernel rounding validation
iChubai Oct 3, 2026
0c56052
test(diffusion): gate native timing and matched LightVAE encoder diag…
iChubai Oct 3, 2026
3dbb1eb
feat: enlarge the official TUI logo from source pixels
iChubai Oct 3, 2026
5328565
fix(benchmarks): scope codec provenance to execution dependencies
iChubai Oct 3, 2026
1808aa2
docs(diffusion): publish source-qualified native acceleration evidence
iChubai Oct 3, 2026
a2a26b9
merge: integrate audited pluggable diffusion acceleration
iChubai Oct 3, 2026
fb64d5d
docs(diffusion): record final merged-release regression gates
iChubai Oct 3, 2026
5762200
merge: include final acceleration release validation
iChubai Oct 3, 2026
92f2a13
docs(diffusion): record final SANA execution and quality gates
iChubai Oct 3, 2026
53ce8df
merge: include source-qualified final SANA acceleration evidence
iChubai Oct 3, 2026
8f92b0e
feat(diffusion): add calibrated packed precision and independent MHA …
iChubai Oct 3, 2026
c1a368c
fix(benchmarks): qualify native acceleration receipts from effective …
iChubai Oct 3, 2026
183147e
fix(attention): avoid unsafe Hopper autotuning tiles on older Triton
iChubai Oct 3, 2026
aeb1ade
docs(diffusion): publish source-bound acceleration quality and timing…
iChubai Oct 3, 2026
31fb97a
merge: integrate calibrated native acceleration with verified quality…
iChubai Oct 3, 2026
78e132a
docs(diffusion): record merged-release acceleration regression gates
iChubai Oct 3, 2026
85ba6ea
merge: include final native acceleration release validation
iChubai Oct 3, 2026
74bd33b
chore: install HERO development scope limits
iChubai Oct 4, 2026
191ff77
fix(acceleration): preserve parent hooks and simplify diagnostics
iChubai Oct 4, 2026
4b75958
merge: simplify acceleration diagnostics with HERO scope limits
iChubai Oct 4, 2026
c2af343
fix(runtime): avoid retrying failed stateful operations
iChubai Oct 4, 2026
0a64ae9
fix(cli): surface invalid runtime profiles
iChubai Oct 4, 2026
ce876c4
fix(checkpoints): limit cache recovery and validation to relevant fai…
iChubai Oct 4, 2026
6fea4ce
merge: simplify runtime, CLI, and checkpoint failure handling
iChubai Oct 4, 2026
f1caafc
fix(inspatio): propagate seed and deterministic settings to workers
iChubai Oct 3, 2026
f1a2273
test(inspatio): isolate leaf imports from the test process
iChubai Oct 3, 2026
3476bbd
test(diffusion): require acceleration contracts in CPU and CUDA gates
iChubai Oct 4, 2026
dcac3a5
fix(attention): route CPU camera tensors and require calibrated contr…
iChubai Oct 4, 2026
a941b87
fix(ci): resolve lazy replay ownership and validate exported outputs
iChubai Oct 5, 2026
872802f
feat(docs): refresh model commands, validation status and terminal br…
iChubai Oct 5, 2026
5b8fd68
merge: integrate public documentation, TUI and strict replay coverage
iChubai Oct 5, 2026
62b624f
fix(ci): keep generated model defaults portable and gate documentatio…
iChubai Oct 5, 2026
4e4c015
merge: align upstream main with the public inference release
iChubai Oct 5, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
The diff you're trying to view is too large. We only load the first 3000 changed files.
64 changes: 64 additions & 0 deletions .dockerignore
Original file line number Diff line number Diff line change
@@ -1,8 +1,26 @@
.git
.github
.codex
.codex/
.claude/
.agents/
.cursor/
.vscode/
.idea/
.env*
.envrc
**/.env*
**/.envrc
**/.direnv/
credentials/*.secret
**/credentials/*.secret
token
*.token
**/*.token
.mypy_cache
.pytest_cache
.ruff_cache
.runtime/
__pycache__/
*.py[cod]
*.egg-info/
Expand All @@ -17,6 +35,9 @@ runs/
tmp/
inspatio_world_output/
worldfoundry_outputs/
*.log
*.out
*.err

node_modules/
docs/fumadocs/.next/
Expand All @@ -32,3 +53,46 @@ docs/fumadocs/node_modules/
*.tar.gz
*.tgz
*.zip

worldfoundry/training
worldfoundry/cli/training.py
worldfoundry/cli/training_commands
scripts/training
configs/training
configs/post_training
test
tests
test_stream

testcase
worldfoundry/data/test_cases
**/tmp
**/test_*.py
**/*_test.py

# Include the public test sources for development/test containers.
!tests/
!tests/**/
!tests/**/*.py
!tests/**/*.md
!tests/**/*.sh
!tests/**/*.yaml
!tests/**/*.yml
!tests/**/*.json
!tests/**/*.jsonl
tests/**/__pycache__/
tests/**/.pytest_cache/
tests/**/.mypy_cache/
tests/**/.ruff_cache/
tests/**/tmp/
tests/**/temp/
tests/**/logs/
tests/**/outputs/
tests/**/.env*
tests/**/*.py[cod]
tests/**/*.log
tests/**/*.ckpt
tests/**/*.pt
tests/**/*.pth
tests/**/*.safetensors
tests/**/*.mp4
5 changes: 3 additions & 2 deletions .gitattributes
Original file line number Diff line number Diff line change
Expand Up @@ -6,8 +6,9 @@
*.bin filter=lfs diff=lfs merge=lfs -text
*.mp4 filter=lfs diff=lfs merge=lfs -text
docs/fumadocs/public/home-demos/*.mp4 -filter -diff -merge -text
docs/fumadocs/public/readme-demos/*.mp4 -filter -diff -merge -text
docs/fumadocs/public/cover_4x4_hero.mp4 -filter -diff -merge -text
worldfoundry/data/test_cases/**/*.mp4 -filter -diff -merge -text
*.mov filter=lfs diff=lfs merge=lfs -text
*.pt filter=lfs diff=lfs merge=lfs -text
*.pth filter=lfs diff=lfs merge=lfs -text
# Retained license texts preserve upstream whitespace byte for byte.
THIRD-PARTY-NOTICES -text whitespace=-blank-at-eol
13 changes: 13 additions & 0 deletions .github/CODEOWNERS
Original file line number Diff line number Diff line change
@@ -0,0 +1,13 @@
# Review coverage for high-impact infrastructure and generated-data surfaces.
# GitHub ignores owners that do not have write access to this repository.

.github/workflows/ @dhsaikhgdius
.github/dependabot.yml @dhsaikhgdius
.github/CODEOWNERS @dhsaikhgdius

docs/ @dhsaikhgdius
scripts/docs/ @dhsaikhgdius
worldfoundry/data/ @dhsaikhgdius
pyproject.toml @dhsaikhgdius
requirements/ @dhsaikhgdius
docker/ @dhsaikhgdius
254 changes: 251 additions & 3 deletions .github/workflows/ci.yml
Original file line number Diff line number Diff line change
Expand Up @@ -5,13 +5,216 @@ on:
push:
branches:
- main
- 'release/**'
workflow_dispatch:
inputs:
source_revision:
description: Exact candidate commit for trusted inference evidence
required: false
type: string
base_revision:
description: Exact comparison commit
required: false
type: string
evidence_json:
description: Compact numerical evidence from a trusted model host
required: false
type: string

permissions:
contents: read
actions: read

jobs:
geometry-impact:
name: ${{ github.event_name == 'workflow_dispatch' && inputs.evidence_json != '' && 'evidence-dispatch-idle-geometry-impact' || 'geometry-impact' }}
if: ${{ github.event_name != 'workflow_dispatch' || inputs.evidence_json == '' }}
runs-on: ubuntu-latest
steps:
- name: Checkout with diff history
uses: actions/checkout@v5
with:
fetch-depth: 0
- name: Set up Python
uses: actions/setup-python@v6
with:
python-version: "3.11"
- name: Plan affected 3D, video and world-model replays without weights or GPU
env:
IMPACT_BASE_SHA: ${{ inputs.base_revision || github.event.pull_request.base.sha || github.event.before }}
run: |
if [ -n "$IMPACT_BASE_SHA" ] && [ "$IMPACT_BASE_SHA" != "0000000000000000000000000000000000000000" ]; then
python tests/manual/geometry_regression_impact.py --base "$IMPACT_BASE_SHA" --output tmp/geometry-impact.json
else
python tests/manual/geometry_regression_impact.py --changed-file worldfoundry/core/__init__.py --output tmp/geometry-impact.json
fi
- name: Publish required replay plan, including uncovered changes
if: always()
uses: actions/upload-artifact@v4
with:
name: geometry-impact
path: tmp/geometry-impact.json
if-no-files-found: error
- name: Install declaration reader
if: always()
run: python -m pip install "PyYAML>=6"
- name: Reject new, regressed or affected uncovered native variants
if: always()
env:
IMPACT_BASE_SHA: ${{ inputs.base_revision || github.event.pull_request.base.sha || github.event.before }}
run: |
if [ -n "$IMPACT_BASE_SHA" ] && [ "$IMPACT_BASE_SHA" != "0000000000000000000000000000000000000000" ]; then
python tests/manual/inference_regression_coverage.py --base "$IMPACT_BASE_SHA" --output tmp/inference-coverage.json
else
python tests/manual/inference_regression_coverage.py --output tmp/inference-coverage.json
fi
- name: Publish coverage definitions and gaps
if: always()
uses: actions/upload-artifact@v4
with:
name: inference-coverage
path: tmp/inference-coverage.json
if-no-files-found: error

inference-replay:
name: ${{ github.event_name == 'workflow_dispatch' && inputs.evidence_json != '' && 'evidence-dispatch-idle-inference-replay' || 'inference-replay' }}
if: ${{ github.event_name != 'workflow_dispatch' || inputs.evidence_json == '' }}
runs-on: ubuntu-latest
env:
GITHUB_TOKEN: ${{ github.token }}
INFERENCE_EVIDENCE_CONTROL_REF: ${{ vars.INFERENCE_EVIDENCE_CONTROL_REF }}
INFERENCE_EVIDENCE_WORKFLOW_PATH: ${{ vars.INFERENCE_EVIDENCE_WORKFLOW_PATH }}
INFERENCE_REFERENCE_INDEX_SHA256: ${{ vars.INFERENCE_REFERENCE_INDEX_SHA256 }}
IMPACT_BASE_SHA: ${{ inputs.base_revision || github.event.pull_request.base.sha || github.event.before }}
steps:
- name: Checkout exact candidate with comparison history
uses: actions/checkout@v5
with:
fetch-depth: 0
- name: Set up Python
uses: actions/setup-python@v6
with:
python-version: "3.11"
- name: Compute committed inference impact
run: |
test -n "$IMPACT_BASE_SHA"
test "$IMPACT_BASE_SHA" != "0000000000000000000000000000000000000000"
python tests/manual/geometry_regression_impact.py --base "$IMPACT_BASE_SHA" --output tmp/geometry-impact.json
- name: Require successful official evidence for this exact commit
run: python tests/manual/inference_actions_evidence.py verify-actions --plan tmp/geometry-impact.json --output tmp/inference-evidence.json
- name: Publish replay gate result
if: always()
uses: actions/upload-artifact@v4
with:
name: inference-evidence-verification
path: tmp/inference-evidence.json
if-no-files-found: error

attest-inference-evidence:
if: ${{ github.event_name == 'workflow_dispatch' && inputs.evidence_json != '' }}
runs-on: ubuntu-latest
env:
INFERENCE_EVIDENCE_CONTROL_REF: ${{ vars.INFERENCE_EVIDENCE_CONTROL_REF }}
INFERENCE_EVIDENCE_WORKFLOW_PATH: ${{ vars.INFERENCE_EVIDENCE_WORKFLOW_PATH }}
INFERENCE_REFERENCE_INDEX_SHA256: ${{ vars.INFERENCE_REFERENCE_INDEX_SHA256 }}
INFERENCE_REPORT_JSON: ${{ inputs.evidence_json }}
CANDIDATE_REVISION: ${{ inputs.source_revision }}
BASE_REVISION: ${{ inputs.base_revision }}
steps:
- name: Checkout trusted control workflow source
uses: actions/checkout@v5
with:
path: control
persist-credentials: false
- name: Checkout candidate for immutable source verification
uses: actions/checkout@v5
with:
ref: ${{ inputs.source_revision }}
path: candidate
fetch-depth: 0
persist-credentials: false
- name: Set up Python
uses: actions/setup-python@v6
with:
python-version: "3.11"
- name: Verify numerical bundle and control provenance
run: |
python control/tests/manual/inference_actions_evidence.py attest \
--base "$BASE_REVISION" --head "$CANDIDATE_REVISION" \
--control-revision "$GITHUB_SHA" --actual-control-ref "$GITHUB_REF_NAME" \
--run-id "$GITHUB_RUN_ID" --source-root candidate --output tmp/receipt.json
- name: Publish trusted inference receipt
uses: actions/upload-artifact@v4
with:
name: trusted-inference-evidence
path: tmp/receipt.json
if-no-files-found: error
retention-days: 90

cpu-tests:
name: ${{ github.event_name == 'workflow_dispatch' && inputs.evidence_json != '' && 'evidence-dispatch-idle-cpu-tests' || 'cpu-tests' }}
if: ${{ github.event_name != 'workflow_dispatch' || inputs.evidence_json == '' }}
runs-on: ubuntu-latest
steps:
- name: Checkout
uses: actions/checkout@v5
- name: Set up Python
uses: actions/setup-python@v6
with:
python-version: "3.11"
- name: Install CPU test dependencies
run: python -m pip install -e ".[test]"
- name: Run public CPU inference and packaging tests
run: make test
- name: Validate documented commands and catalog claims without weights
run: make test-docs-contracts
- name: Publish complete CPU selection and results
if: always()
uses: actions/upload-artifact@v4
with:
name: public-cpu-contracts
path: |
tmp/public-cpu-selection.json
tmp/public-cpu-contracts.xml
tmp/docs-contract-selection.json
tmp/docs-contracts.xml
if-no-files-found: error

inference-tensors:
name: ${{ github.event_name == 'workflow_dispatch' && inputs.evidence_json != '' && 'evidence-dispatch-idle-inference-tensors' || 'inference-tensors' }}
if: ${{ github.event_name != 'workflow_dispatch' || inputs.evidence_json == '' }}
runs-on: ubuntu-latest
env:
CUDA_VISIBLE_DEVICES: ""
PYTEST_DISABLE_PLUGIN_AUTOLOAD: "1"
steps:
- name: Checkout
uses: actions/checkout@v5
- name: Set up Python
uses: actions/setup-python@v6
with:
python-version: "3.11"
- name: Install CPU Torch and tensor test dependencies
run: |
python -m pip install "torch==2.9.0+cpu" --index-url https://download.pytorch.org/whl/cpu
python -m pip install -e ".[test]" \
"safetensors>=0.4" "einops>=0.7,<0.9" "omegaconf>=2.3,<3" "triton==3.5.0" \
"transformers>=4.57,<4.58" ftfy regex opencv-python-headless "aiohttp>=3.10,<4" scikit-image
- name: Run inference tensors and streaming contracts without model weights
run: make test-infer-tensors
- name: Publish inference contract results
if: always()
uses: actions/upload-artifact@v4
with:
name: inference-cpu-contracts
path: |
tmp/inference-cpu-selection.json
tmp/inference-cpu-contracts.xml
if-no-files-found: error

public-surface:
name: ${{ github.event_name == 'workflow_dispatch' && inputs.evidence_json != '' && 'evidence-dispatch-idle-public-surface' || 'public-surface' }}
if: ${{ github.event_name != 'workflow_dispatch' || inputs.evidence_json == '' }}
runs-on: ubuntu-latest
defaults:
run:
Expand Down Expand Up @@ -40,7 +243,10 @@ jobs:
- name: Install quality gate dependencies
run: |
python -m pip install --upgrade pip
python -m pip install PyYAML
python -m pip install PyYAML "ruff==0.12.7"

- name: Check generated docs
run: make docs-check

- name: Run lightweight quality gates
run: make lint
Expand All @@ -51,5 +257,47 @@ jobs:
- name: Run CLI check
run: make cli-check

- name: Run documented CLI checks
run: make docs-check
- name: Run documented CLI entrypoint checks
run: make cli-entrypoint-check


packaging-license-gate:
name: ${{ github.event_name == 'workflow_dispatch' && inputs.evidence_json != '' && 'evidence-dispatch-idle-packaging-license-gate' || 'packaging-license-gate' }}
if: ${{ github.event_name != 'workflow_dispatch' || inputs.evidence_json == '' }}
runs-on: ubuntu-latest
defaults:
run:
shell: sh

steps:
- name: Checkout clean source tree
uses: actions/checkout@v5

- name: Set up Python
uses: actions/setup-python@v6
with:
python-version: "3.11"

- name: Install wheel build and audit tooling
run: python -m pip install --upgrade build "setuptools>=77" "setuptools-scm[toml]" wheel

- name: Build and audit clean distributions
run: |
test -z "$(git status --porcelain)"
test ! -e build
audit_dir="$(mktemp -d)"
python -m build --outdir "$audit_dir"
set -- "$audit_dir"/*.whl
test "$#" -eq 1
test -f "$1"
wheel_path="$1"
set -- "$audit_dir"/*.tar.gz
test "$#" -eq 1
test -f "$1"
make packaging-check WHEEL="$wheel_path" SDIST="$1"
python -m pip install "${wheel_path}[test]"
source_dir="$audit_dir/source"
mkdir "$source_dir"
tar -xzf "$1" -C "$source_dir"
cd "$source_dir"/*
make test
Loading
Loading