Richard Palethorpe
2facfc0d88
* fix(vulkan): preserve host ICD discovery for packaged backends Add bundled Mesa manifests through VK_ADD_DRIVER_FILES instead of replacing the system driver list. Merge inherited and model-specific additive paths while preserving explicit operator overrides, with regression coverage. Assisted-by: Codex:gpt-5 golangci-lint Assisted-by: Codex:gpt-5.6-sol Signed-off-by: Richard Palethorpe <io@richiejp.com> * feat(3d): add Kimodo CPU and Vulkan animation backend Introduce a distinct animation capability and model-described 3D operations, with a typed /3d/animate API, RPC transport, distributed media staging, permissions, and tracing. Add a persistent kimodo.cpp adapter, skeleton GLB export, CPU/Vulkan packages, model and backend galleries, importer support, CI builds, and documentation. Adapt Studio inputs to each model and provide real-time skeleton playback, seeking, and history. Cover backend validation, packaging, API behavior, importer inventories, distributed staging, and Studio workflows. Validate real-model CPU/Vulkan generation and deploy the integration to the local QA instance. Assisted-by: Codex:gpt-5 golangci-lint Assisted-by: Codex:gpt-5.6-sol Signed-off-by: Richard Palethorpe <io@richiejp.com> * feat(kimodocpp): adopt monolithic encoders and resident inference Update upstream for resident weights, packed execution paths, and cached motion graphs. Default to all 32 text layers while retaining configurable streaming and legacy bundle support. Use monolithic Q8_0 encoders by default and offer all six published quantizations through the gallery and importer. Refresh pinned hashes, tests, and documentation; remove the obsolete thread patch and ensure cached source checkouts follow the upstream pin. Validated CPU and Vulkan generation, lower-bit streaming, gallery/importer suites, packaging, lint, and cold/warm Studio generation on localai-dev. Assisted-by: Codex:gpt-5 golangci-lint Assisted-by: Codex:gpt-5.6-sol Signed-off-by: Richard Palethorpe <io@richiejp.com> --------- Signed-off-by: Richard Palethorpe <io@richiejp.com>
kimodocpp
Persistent gRPC adapter for the pinned kimodo.cpp C API. The small C++ bridge exposes upstream skeleton metadata and runtime configuration; Go exports the returned motion as a node-only animated GLB. See 3D animation for the API and runtime options.
make protogen-go
make -C backend/go/kimodocpp BUILD_TYPE=cpu
KIMODO_TEST_PACKAGE="$PWD/backend/go/kimodocpp/package" make -C backend/go/kimodocpp test
make -C backend/go/kimodocpp BUILD_TYPE=vulkan
Requires CMake 3.25+, a C++23 compiler, and Go. Vulkan additionally requires the Vulkan SDK with shader tooling. Linux and macOS/Apple Silicon CPU are supported; Metal is disabled because upstream implements CPU/Vulkan only.
Normal CI tests need no model downloads. To exercise real weights, set
KIMODO_TEST_LIBRARY to the built libkimodo.so (or dylib),
KIMODO_TEST_MOTION to a motion GGUF, KIMODO_TEST_TEXT to the complete text
GGUF beside tokenizer.gguf (or legacy bundle directory), and
KIMODO_TEST_DEVICE=cpu or vulkan, then run the tests.
The real-model test generates two clips with one session to cover reuse.
Set KIMODO_TEST_TEXT_LAYER_CHUNK=8 to exercise bounded streaming instead of
the default all-layer residency.
Upstream now respects the configured thread count in both encoders and keeps
motion weights and execution graphs resident. The adapter defaults to 32 text
layers so the text weights also stay resident between requests. When bumping
the pinned upstream commit, verify the C ABI layout/version and all three
skeleton families. Pin changes update a clean cached checkout automatically;
local source modifications stop the update rather than being discarded. Preserve
any such changes before using make clean to replace that generated checkout.