Repository navigation
Releases
Blender Agent Kit v0.2.3
Blender Agent Kit v0.2.3
v0.2.3 is an agent-efficiency and architecture-hardening release driven by two sources of evidence: the independent SupplyCrate/Helios Conduit production trial and a later blind two-subagent quadcopter workflow comparison.
The release focuses on making production rules explicit, reducing agent discovery overhead, improving review/delivery automation, and replacing documentation-only batch claims with executable caching/telemetry.
Agent-first discovery
Two host-only commands now expose executable truth without locating or launching Blender:
python skills/core/blender-core/scripts/run.py capabilities --json
python skills/core/blender-core/scripts/run.py contract --target unity --asset-class solid_propcapabilitiesreports executable commands, host-vs-Blender execution mode, installed targets/policies/project profiles, evidence surfaces, and guidance/configuration skills.contractresolves engine target -> default production policy -> optional project profile and returns naming/classification, QA, LOD/material/pivot/UV/rigging rules, texture packing, export/round-trip behavior, and review styles/presets in one JSON contract.- Public JSON schemas cover both surfaces.
- Invalid host arguments/errors remain machine-readable.
- The 10 hot-path
SKILL.mdfiles were reduced from roughly 1,800 words to 1,229 words (~31.7% less) while retaining critical Blender/PBR safety guardrails.
Engine target vs production policy
Unity engine semantics are now separated from generic real-time asset policy.
The raw Unity target owns:
- coordinate/export behavior;
- hierarchy-safe static Y-up FBX handling;
- standard URP/HDRP texture layouts;
- engine-relevant round-trip settings.
The reusable realtime-default policy owns toolkit production defaults such as:
- triangle/material budgets;
- pivot/UV/texel-density expectations;
- LOD/collider naming;
- directional-empty naming;
- LOD transition defaults.
This keeps statements such as a 15k triangle budget correctly attributable to Blender Agent Kit policy for the Unity target, not to Unity as a universal engine requirement.
LOD, collider, and directional-empty naming are configuration-driven. Blender tests prove alternate naming can be supplied without changing core code.
Configurable review styles
The default production policy now supplies reusable review styles:
technical— headless-safe low-sample Cycles, brighter neutral presentation, deterministic high-contrast lighting, and tighter inspection framing.showcase— higher-sample Cycles with presentation-oriented background, light rig, lens, and framing.
Render engine, samples, background, camera factors, and light rig are policy/configuration data rather than hidden constants. Project review presets can select and overlay styles.
Eevee remains configurable when the runtime has a suitable graphics/EGL stack, but is intentionally not the headless CI default.
QA-gated delivery
New host command:
python skills/core/blender-core/scripts/run.py deliver \
--blend asset.blend \
--target unity --asset-class solid_prop \
--review-style technical --out delivery/Default flow:
inspect -> QA gate -> optional pack -> review -> export/bundle -> fresh-import verification
- QA failures stop before expensive review/export by default.
- An explicit diagnostic override can continue after QA failure.
- Results record per-stage status and measured duration, QA-gate state, retry guidance, and generated evidence.
- Delivery reports validate against a public JSON schema.
Executable batch processing and safe caching
blender-batch is no longer guidance-only. The public batch command supports:
qapackexportbundledeliver
Blender-backed cache misses remain process-isolated. Cache identity includes:
- source bytes;
- operation-relevant resolved configuration;
- executable toolkit identity;
- relevant operation options;
- Blender executable/version identity.
Export/bundle cache hits additionally verify manifest build hashes and required generated files before a skip is accepted.
Reports include measured per-asset duration, cache hit/miss counts, processed/skipped/failed counts, and tool/Blender identity. The runtime explicitly does not fabricate model/API-token or agent-turn telemetry.
Tier-4 CI proves a real fresh SupplyCrate batch miss followed by an unchanged zero-processing cache hit.
Canonical reproducible demos
The independently authored source-bound demonstrations under demo/ are now the canonical v0.2.3+ showcase lineage.
python demo/SupplyCrate/run_demo.py --mode full --clean
python demo/SciFiKit/run_demo.py --mode full --cleanBoth also provide bounded --mode smoke entrypoints used directly by Tier-4 CI.
The older top-level demos/ tree is retained as historical v0.2.1 provenance/backwards links rather than rewritten.
The current Sci-Fi full showcase also verifies and displays the optional source-bound Unity _EmissionMap, which the original v0.2.1 evidence predated.
Independent evidence context
The earlier independent production trial reported retrospective estimates of:
- 74–80% lower pipeline-engineering effort;
- 55–63% lower combined end-to-end effort.
Those figures remain explicitly labeled estimates, not controlled benchmarks.
A later blind two-subagent quadcopter workflow comparison observed:
- 90 -> 0 Blender Agent Kit
solid_proppolicy errors for the Unity target; - 376.12 s -> 56.49 s recorded Blender compute across the two delivered workflows;
- 150 -> 0 evaluator-classified custom pipeline-boilerplate lines;
- 8 -> 13 agent turns;
- kit export fresh-import parity at 10,240 -> 10,240 triangles.
The compute/render workloads and final assets were not identical, so the 85% compute difference is an observed workflow outcome, not an isolated library-speed claim or guaranteed saving. Raw subagent orchestration transcripts were not included in the submitted evidence package, so complete prompt isolation remains experimenter-reported.
Verification
The release candidate is gated by:
- Ubuntu host/schema tests;
- Windows host/schema tests;
- Blender 5.2 Tier-2/3;
- Tier-4 canonical demo smoke using the documented
run_demo.pyentrypoints; - fresh Blender FBX round-trip verification;
- tag-specific CI before GitHub Release publication.
Known limits
- No Unity Editor import/prefab/LODGroup verification is claimed.
- Geometric UV1 island-overlap proof remains unimplemented; requested UV1 emits
uv1_overlap_unverifiedwhen only presence can be proven. - Collider convexity/decomposition quality is not geometrically certified by core QA.
- The original large independent-trial binary commit remains in Git history to preserve provenance.
Blender Agent Kit v0.2.2
Blender Agent Kit v0.2.2 — Candidate Notes
This is a candidate, not a published release. The final release tag is created only after an exact-head review and CI approval.
Independent agent showcase
- Promotes two newly authored external-agent demonstrations in the README: a procedural SupplyCrate and the Helios Conduit six-module Sci-Fi kit. The new master/optimized images, scenes, sources, original machine reports, manifests, and analysis remain in
demo/. - SupplyCrate: 4,204 / 1,084 / 136 visible tris; 12-triangle collision; four Blender FBX round trips passed; nine-map source PBR → URP sheet includes 16-bit normal and metallic smoothness alpha evidence.
- Helios Conduit: six piece types, 21 module instances, 109 objects in the assembled scene, 2,392 / 904 / 336 / 60 visible bulkhead LOD triangles, six socket empties, three collision hulls, six Blender FBX round trips passed.
- The independent agent's retrospectively estimated effort savings are presented with explicit scope/methodology: 74–80% pipeline-only and 55–63% combined end-to-end. The estimates are not controlled benchmark results or guarantees. It reported two genuine QA-caught authoring defects, zero round-trip failures in ten exported FBXs, and particularly useful hierarchy-safe export, QA, URP packing, and review automation.
- The trial did not execute Unity Editor; all engine-target checks are Blender round trips, profile assertions, and texture validation.
General-purpose toolkit improvements
- Treat recognized collision meshes as collision inventory, not render-LOD mesh/triangle/bounds contributions; carry configured LOD exclusions into review inventory.
- Validate
required_parent_pairsstructure before launching QA, including merged target overlays; enforce corresponding target schema constraints. - Provide fast
run.py <subcommand> --helpbased on the real argument declarations, without launching Blender. - Upgrade host review compositor with readable scalable fonts, title/subtitle, non-cropped layouts, explicit 16-bit RGB/RGBA PNG channel reading, and a visible smoothness alpha split.
- Add optional source-bound Unity
_EmissionMappacking for URP/HDRP without fabricating emission maps when no source exists. - Surface a bounded
stderr_tailin machine-readable errors when headless Blender fails, while still retaining the full local log. - Improve onboarding docs with a minimal external profile and Blender executable discovery.
Version-independent release safety
- A new Tier-4 independent-source CI smoke regenerates both procedural texture sets, rebuilds the SupplyCrate and representative Sci-Fi bulkhead with Blender 5.2.4, runs QA, packs URP textures (including optional emission), and verifies all 10 newly exported FBXs by Blender fresh import. This does not rerender all beauty images or exercise Unity Editor.
- Delete one-off
publish-v0.2.1.yml. - Add
workflow_callto existing host/Blender CI and a version-neutralrelease.ymltriggered byv*.*.*tags. - Require the exact tagged SHA's Ubuntu, Windows and Blender CI jobs to pass before
gh release create, and require matchingRELEASES/<tag>.mdnotes. The release job verifies the supplied tag and SHA.
Evidence and storage limits
- Machine results and effort figures are independently agent-reported; no fresh rebuild by this release author has yet been performed.
demo/contains substantial historical binary evidence. This release preserves the originally delivered files and attaches a deterministic independent-trial archive plus SHA256 to the GitHub Release. Future generated binary outputs are ignored bydemo/.gitignore, while source/configuration, JSON evidence and optimized WebP remain suitable for normal Git. Existing Git history is not rewritten, so the original 115 MB commit remains in repository history.
Blender Agent Kit v0.2.1
Blender Agent Kit v0.2.1
Patch release — CLI correctness, repeatable review sheets, and an independently evaluated modular-kit showcase.
Fixed
- Public
--result-jsonis now owned by the host entrypoint instead of leaking into Blender's internal bootstrap result transport.inspect,qa,pack,export,bundle, andrevieware covered by host regression tests. - SupplyCrate texture showcase generation now includes all five source PBR maps and all four Unity URP packed maps. The 16-bit PNG reader validates the PNG signature, supported channel/depth combinations, and unfiltered scanlines; PNG/WebP content is validated and missing inputs fail closed.
- The texture generator exposes the actual metallic/smoothness alpha channel rather than presenting RGB alone. Synthetic fixtures cover nine tiles and a 16-bit normal PNG.
- The README no longer displays the previously reported blank legacy
demo_textures.webp. Instead, it shows a compressed preview of the original six-tile PNG supplied in the SupplyCrate archive; this is not represented as regenerated nine-map evidence.
Added
- Generic, host-side
tools/review_compose.pyfor configurable multiview, LOD, range, and collision contact sheets using real source images and caller-supplied evidence labels. Includes host tests and CI Pillow installation. - A second, materially different independent-agent case study under
demos/SciFiKit/, including source-archive assessment, machine-evidence summary, provenance limitations, and a hash-checked importer. - Sci-Fi trial documents a 60-object / 30-mesh six-piece modular assembly, directional snap sockets, a two-material bulkhead, LOD0–3 (3,628 / 1,260 / 372 / 72 visible triangles), QA pass with explicitly unverified UV1 overlap, 5,380-triangle Blender FBX round-trip including collision, and four Unity URP packed texture slots. These figures describe supplied original machine evidence, not a fresh v0.2.1 rebuild.
Evidence and limitations
- Blender 5.2 tests validate the tooling; they are not a fresh rerun of the supplied Sci-Fi kit or SupplyCrate full asset build.
- SupplyCrate's corrected nine-map contact sheet requires raw and packed map regeneration. No manually reconstructed sheet or invented production evidence was committed in this patch.
- Sci-Fi's original proof images and editable builder exist in the supplied archive; the repository publishes an evidence assessment, importer and two lightweight WebP previews from the original agent images, not a newly regenerated canonical asset gallery. The asset conversion/rebuild workflow remains follow-up work.
- Unity Editor asset import was not executed or verified.
- Independent agent-reported productivity savings are subjective feedback, not a measured benchmark.
Verification
GitHub CI on the final release candidate must pass: Ubuntu Tier-1, Windows Tier-1, and Blender 5.2 Tier-2/3. Review and exact-head status should be recorded in PR #3.
See SupplyCrate, Sci-Fi trial, and PR #3.