[package]
edition = "2021"
rust-version = "1.85"
name = "spec-stream"
version = "0.1.1"
authors = [
"GLASTOR® Core <ventas@glastor.es>",
"Andrés Antonio Cardoso <glastor.info@gmail.com>",
]
build = false
autolib = false
autobins = false
autoexamples = false
autotests = false
autobenches = false
description = "High-performance, model-agnostic speculative decoding and paged KV-cache engine for accelerated LLM inference"
homepage = "https://glastor-core.github.io/spec-stream/"
readme = "README.md"
keywords = [
"llm",
"speculative-decoding",
"kv-cache",
"inference",
"ai",
]
categories = [
"algorithms",
"asynchronous",
"concurrency",
]
license = "MIT OR Apache-2.0"
repository = "https://github.com/glastor-core/spec-stream"
[features]
cache = ["dep:spec-stream-cache"]
default = ["full"]
engine = [
"dep:spec-stream-engine",
"cache",
]
full = [
"cache",
"engine",
]
[lib]
name = "spec_stream"
path = "src/lib.rs"
[[example]]
name = "speculative_pipeline"
path = "examples/speculative_pipeline.rs"
[[bench]]
name = "verification_bench"
path = "benches/verification_bench.rs"
harness = false
[dependencies.spec-stream-cache]
version = "0.1.1"
optional = true
[dependencies.spec-stream-core]
version = "0.1.1"
[dependencies.spec-stream-engine]
version = "0.1.1"
optional = true
[dev-dependencies.tokio]
version = "1.40"
features = ["full"]
[dev-dependencies.tokio-stream]
version = "0.1"
[dev-dependencies.tracing-subscriber]
version = "0.3"