Docs.rs

Build and parse RFC 5322 / MIME email messages, without an SMTP dependency

Safe Rust API for NVIDIA TensorRT runtime inference.

Build + raw FFI bindings to baracuda's port of NVIDIA TransformerEngine's FP8 cast/transpose + delayed-scaling recipe primitives. Cast/recipe subset only — `normalization` / `fused_rope` / `fused_attn` / `fused_softmax` / `activation` / `gemm` deliberately skipped (overlap existing baracuda Phase 3/5/14/17/30/31/36/41/42). NO cuDNN dep (recipe + cast paths don't need it; `fused_attn` would, and we skip it); NO pybind11 (the safe wrapper lives in `baracuda-transformer-engine` and exposes a raw C ABI defined in `csrc/baracuda_te_shim.cu`). Apache-2.0 per upstream — see `ATTRIBUTION.md`.

Safe Rust wrappers for the CUDA Runtime API (devices, streams, events, managed memory, kernel launch via the library API).

Build + raw FFI bindings to baracuda's clean-fork of Hiroyuki Ootomo's ozIMMU — the Ozaki-scheme FP64 GEMM library that synthesizes a DGEMM from S² int8 tensor-core matmuls. Phase 44b internalized the upstream sources under `cuda/` (no more `vendor/` subdir; cutf submodule eliminated). Linked statically into the baracuda CUDA stack; consumed by the safe wrapper crate `baracuda-ozimmu`. MIT-licensed (original ozIMMU MIT — see `ATTRIBUTION.md`).

Optimizer kernels (Adam / LAMB / SGD) for the baracuda CUDA stack, built on the multi_tensor_apply idiom vendored from NVIDIA Apex (BSD-3-Clause). One launch over thousands of parameter tensors — critical for the optimizer step on large-model training stacks. NEW in Phase 49; deliberate scope expansion (training-framework-adjacent). Off-by-default in baracuda-kernels via the `optim` cargo feature so inference-only consumers don't pay the FFI surface cost.

Raw FFI bindings + dynamic loader for NVIDIA TensorRT, plus an optional C++ shim exposing a flat C ABI over TensorRT's C++-only runtime.

Credential-acquisition mechanism for the Ppoppo SDK family: the single-flight token cache, the OAuth2 client_credentials and RFC 8693 token-exchange sources, the one token-endpoint POST and its RFC 6749 §5.2 refusal reading, RFC 8414 discovery, and the reqwest arm of the JWKS fetcher port. Errors are ppoppo-call vocabulary. Not a stable public API.

Safe Rust wrappers for NVIDIA NVSHMEM host API (symmetric-heap one-sided GPU communication).

JWT (RFC 9068 access token + OIDC id_token, EdDSA) issuance engine for the Ppoppo SDK family: SigningKey and the one encoder per profile. PAS mints with it; the SDK test kits mint through it so a fake issuer signs exactly what PAS signs. Verification is ppoppo-engine-jwt-verify; the token payload is ppoppo-claims. Not a stable public API.

JWT (RFC 9068 access token + OIDC id_token, EdDSA) verification engine for the Ppoppo SDK family: the one decode function behind RFC 8725 mitigations, KeySet / JWKS parsing, and the JWKS cache's fresh → stale → expired policy over an injected fetcher port. No signing, no HTTP. Not a stable public API — consume a ppoppo SDK face.

Display object tree for composited 2D rendering: the Node2D family, sprites, nine-slice sprites, html views and the Scene2D presentation context

Python API for ripgrep-style file walking and searching

What a ppoppo face's test kit scripts and records — route markers, a scripted stream dial, the calls a served face received

Universal Clock + Timer port for ppoppo workspace — chat-as-infrastructure primitive