Cargo.toml 3.79 KB
Newer Older
1
2
3
4
5
# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0

[workspace]
members = [
6
    "launch/dynamo-run",
7
8
    "lib/llm",
    "lib/runtime",
9
    "lib/tokens",
10
    "lib/async-openai",
11
    "lib/parsers",
12
    "lib/bindings/c",
13
    "lib/engines/*",
14
]
15
16
17
18
19
20
21
22
23
24
25
26
# Exclude certain packages that are slow to build and we don't ship as flagship
# features from default build, but keep them in workspace for convenience.
# - launch/dynamo-run
# - lib/engines/*
default-members = [
    "lib/llm",
    "lib/runtime",
    "lib/tokens",
    "lib/async-openai",
    "lib/parsers",
    "lib/bindings/c",
]
27
resolver = "3"
28
29

[workspace.package]
30
version = "0.5.1"
31
edition = "2024"
32
33
description = "Dynamo Inference Framework"
authors = ["NVIDIA Inc. <sw-dl-dynamo@nvidia.com>"]
34
license = "Apache-2.0"
Neelay Shah's avatar
Neelay Shah committed
35
36
homepage = "https://github.com/ai-dynamo/dynamo"
repository = "https://github.com/ai-dynamo/dynamo.git"
37
keywords = ["llm", "genai", "inference", "nvidia", "distributed"]
38
39

[workspace.dependencies]
40
# Local crates
41
42
43
44
dynamo-runtime = { path = "lib/runtime", version = "0.5.1" }
dynamo-llm = { path = "lib/llm", version = "0.5.1" }
dynamo-tokens = { path = "lib/tokens", version = "0.5.1" }
dynamo-async-openai = { path = "lib/async-openai", version = "0.5.1", features = [
45
46
47
    "byot",
    "rustls",
] }
48
49
dynamo-parsers = { path = "lib/parsers", version = "0.5.1" }

50

51
# External dependencies
52
anyhow = { version = "1" }
53
async-nats = { version = "0.40", features = ["service"] }
54
55
async-stream = { version = "0.3" }
async-trait = { version = "0.1" }
56
57
58
async_zmq = { version = "0.4.0" }
blake3 = { version = "1" }
bytes = { version = "1" }
59
60
61
62
63
64
65
chrono = { version = "0.4", default-features = false, features = [
    "alloc",
    "std",
    "clock",
    "now",
    "serde",
] }
66
67
68
derive_builder = { version = "0.20" }
derive-getters = { version = "0.5" }
either = { version = "1.13", features = ["serde"] }
69
etcd-client = { version = "0.16", features = ["tls"] }
70
futures = { version = "0.3" }
71
72
73
74
75
76
77
78
79
80
hf-hub = { version = "0.4.2", default-features = false, features = [
    "tokio",
    "rustls-tls",
    "ureq",
] }

# ModelExpress for model downloading
modelexpress-client = { git = "https://github.com/ai-dynamo/modelexpress.git", rev = "a232220bf268a475d293914d407f4ae186f443e3", package = "modelexpress-client" }
modelexpress-common = { git = "https://github.com/ai-dynamo/modelexpress.git", rev = "a232220bf268a475d293914d407f4ae186f443e3", package = "modelexpress-common" }

81
humantime = { version = "2.2.0" }
82
libc = { version = "0.2" }
Ryan Olson's avatar
Ryan Olson committed
83
oneshot = { version = "0.1.11", features = ["std", "async"] }
84
prometheus = { version = "0.14" }
85
rand = { version = "0.9.0" }
86
87
88
89
90
reqwest = { version = "0.12.22", default-features = false, features = [
    "json",
    "stream",
    "rustls-tls",
] }
91
92
serde = { version = "1", features = ["derive"] }
serde_json = { version = "1" }
93
strum = { version = "0.27", features = ["derive"] }
94
tempfile = "3"
95
96
97
thiserror = { version = "2.0.11" }
tokio = { version = "1", features = ["full"] }
tokio-stream = { version = "0.1" }
Ryan Olson's avatar
Ryan Olson committed
98
tokio-util = { version = "0.7", features = ["codec", "net", "rt"] }
99
100
tower-http = { version = "0.6", features = ["trace"] }
axum = { version = "0.8", features = ["macros"] }
101
tracing = { version = "0.1" }
102
103
104
105
106
tracing-subscriber = { version = "0.3", features = [
    "env-filter",
    "local-time",
    "json",
] }
107
validator = { version = "0.20.0", features = ["derive"] }
108
uuid = { version = "1.17", features = ["v4", "serde"] }
109
url = { version = "2.5", features = ["serde"] }
110
xxhash-rust = { version = "0.8", features = ["xxh3", "const_xxh3"] }
Graham King's avatar
Graham King committed
111
112
113

[profile.dev.package]
insta.opt-level = 3
114

115
116
117
118
[profile.dev]
# release level optimizations otherwise everything feels slow
opt-level = 3

119
[profile.release]
120
# These make the build much slower but shrink the binary, and could help performance
121
codegen-units = 1
122
lto = "thin"