Windows CI / windows ([bun run --cwd packages/app-core test bun run --cwd packages/elizaos test bun run --cwd packages/cloud/shared test], app-and-cli) (push) Has been cancelled
Windows CI / windows ([bun run --cwd packages/scenario-runner test bun run --cwd packages/vault test bun run --cwd packages/security test bun run --cwd plugins/plugin-coding-tools test], framework-packages) (push) Has been cancelled
Windows CI / windows ([bun run --cwd plugins/plugin-elizacloud test bun run --cwd plugins/plugin-discord test bun run --cwd plugins/plugin-anthropic test bun run --cwd plugins/plugin-openai test bun run --cwd plugins/plugin-app-control test bun run --cwd plugins/pl… (push) Has been cancelled
Windows CI / windows ([node packages/scripts/run-turbo.mjs run build --filter=@elizaos/core --filter=@elizaos/shared --filter=@elizaos/agent --concurrency=4 node packages/scripts/run-bash-linux-only.mjs scripts/verify-riscv64-buildpaths.sh node packages/scripts/run… (push) Has been cancelled
Windows CI / windows ([node packages/scripts/run-turbo.mjs run typecheck --filter=@elizaos/core --filter=@elizaos/shared --filter=@elizaos/cloud-shared --concurrency=4 bun run --cwd packages/core test bun run --cwd packages/shared test], core-runtime, 75) (push) Has been cancelled
Standalone C library + GGUF conversion scripts that port BlazeFace
(face detection) and a 128-d face embedding network (FaceNet-style
or ArcFace-mini) onto the elizaOS/llama.cpp fork's ggml dispatcher.
The native runtime replaces the ONNX-backed face-detector-mediapipe.ts
path and provides the face-embedding surface needed to retire the
face-recognition.ts face-api.js / TensorFlow.js stack after production
parity rollout.
Today the model entry points in include/face/face.h are implemented by
src/face_model.c with pure-C scalar BlazeFace and embedder forwards.
The model-independent helpers are real and tested:
face_blazeface_make_anchors — 896-anchor table for the
128x128 BlazeFace front model.
face_blazeface_decode — decode raw model outputs to
source-pixel detections.
face_align_5pt — 5-point similarity warp + bilinear sampler
producing a 112x112 RGB face crop.
face_embed_distance / face_embed_distance_l2 — cosine and L2
distance over unit-norm 128-d embeddings.
The full port plan — upstream pins, GGUF conversion approach, fork
integration steps, replacement path for the TS bindings — lives in
AGENTS.md. Read that before changing anything in this
directory.
include/face/face.h Public C ABI (frozen — see AGENTS.md).
src/face_model.c Native CPU model runtime.
src/face_blazeface.c BlazeFace forward path.
src/face_embed.c 128-d embedder forward path.
src/face_anchor_decode.c BlazeFace anchor table + decoder (real).
src/face_align.c 5-point affine warp + bilinear sampler (real).
src/face_distance.c Cosine + L2 distance helpers (real).
scripts/blazeface_to_gguf.py BlazeFace converter.
scripts/face_embed_to_gguf.py Embedder converter.
test/face_abi_smoke.c Build-only smoke test for the public ABI.
test/face_anchor_test.c Behavioural test for the anchor pipeline.
test/face_align_test.c Behavioural test for the 5-point aligner.
test/face_distance_test.c Behavioural test for the distance helpers.
CMakeLists.txt Builds libface + native test binaries.
License
Apache 2.0 — matches both google/mediapipe (BlazeFace) and
deepinsight/insightface (ArcFace-mini buffalo_s pack). The pinned
upstream commits recorded in scripts/*.py are the source of the
weights this library ships against.