prx 0.5.9

Praxis — agent-native Unix tools. Single binary replacing grep, cat, find, sed, diff for AI coding agents.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140

[package]
name = "prx"
version = "0.5.9"
edition = "2024"
rust-version = "1.85"
description = "Praxis — agent-native Unix tools. Single binary replacing grep, cat, find, sed, diff for AI coding agents."
license = "Apache-2.0"
repository = "https://github.com/civitas-io/prx"
keywords = ["cli", "ai", "agent", "search", "tools"]
categories = ["command-line-utilities", "development-tools"]
readme = "README.md"
exclude = [
    "models/",
    "benchmarks/",
    "book/",
    "docs/",
    ".prx/",
    ".github/",
]

[dependencies]
# CLI framework — derive API with multicall for busybox-style dispatch
clap = { version = "4.6", features = ["derive"] }

# AST parsing — 0.26 confirmed compatible with all grammar crates via tree-sitter-language bridge
tree-sitter = "0.26"
# Structural search patterns (fn $NAME($$$))
ast-grep-core = "0.42"

# Embedding model weights — zero-copy deserialization from include_bytes!
safetensors = "0.7"
# Dense matrix ops — embedding lookup, mean pool, L2 normalize, cosine similarity
ndarray = "0.17"
# Sparse matrices — pre-computed BM25 scores in CSC format
sprs = "0.11"

# Token counting — cl100k_base tokenizer for --budget enforcement
tokenizers = "0.23"

# Diff computation — TextDiff for ag diff, ChangeTag iteration
similar = "3.1"

# Bloom filter — O(1) existence checks for ag exists
bloomfilter = "3.0"

# JSON serialization — output envelope, structured errors
serde = { version = "1", features = ["derive"] }
serde_json = "1"

# Content hashing — xxh3 128-bit for change detection
xxhash-rust = { version = "0.8", features = ["xxh3"] }

# .gitignore-aware file walking — from ripgrep
ignore = "0.4"

# Error handling
thiserror = "2"                    # Typed errors for library code (AgError enum)
anyhow = "1"                      # Ergonomic errors for CLI main()

# Regex for literal search and identifier extraction
regex = "1"

# Home directory resolution for stats file
dirs-next = "2"

# Parallel iteration — embarrassingly parallel indexing (chunking, embedding, import extraction)
rayon = "1"

# Memory-mapped files — zero-copy embedding index access, OS page cache keeps index warm across queries
memmap2 = "0.9"
# Safe byte-to-float casting — reinterpret mmap'd &[u8] as &[f32] with alignment validation
bytemuck = { version = "1", features = ["extern_crate_alloc"] }

# Binary serialization for persistent index
# postcard: compact serde serialization for index files (replaces unmaintained bincode, RUSTSEC-2025-0141)
postcard = { version = "1", features = ["alloc"] }

# f16 support for loading float16 model weights
half = "2"

# Continuation token encoding for paginated search
base64 = "0.22"

# MCP server (optional, enabled by default)
rmcp = { version = "1", features = ["server", "transport-io", "schemars"], optional = true }
tokio = { version = "1", features = ["full"], optional = true }
schemars = { version = "1", optional = true }

# File watching (optional, for ag index --watch)
notify = { version = "8", optional = true }

# Tree-sitter language grammars — all confirmed working with tree-sitter 0.26
tree-sitter-rust = "0.24"
tree-sitter-python = "0.25"
tree-sitter-javascript = "0.25"
tree-sitter-typescript = "0.23"
tree-sitter-go = "0.25"
tree-sitter-java = "0.23"
tree-sitter-c = "0.24"
tree-sitter-cpp = "0.23"
tree-sitter-ruby = "0.23"
tree-sitter-bash = "0.25"
tree-sitter-json = "0.24"
tree-sitter-html = "0.23"
tree-sitter-css = "0.25"

[features]
default = ["mcp"]
mcp = ["rmcp", "tokio", "schemars"]
watch = ["notify", "tokio"]

[profile.release]
lto = true
strip = true
codegen-units = 1
opt-level = "z"                    # Optimize for size (model weights dominate binary)

[dev-dependencies]
assert_cmd = "2"                   # CLI integration testing
predicates = "3"                   # Assertion helpers for CLI output
tempfile = "3"                     # Temp dirs for test fixtures
criterion = "0.8"                  # Benchmarking

[[bench]]
name = "search"
harness = false

[[bench]]
name = "chunking"
harness = false

[build-dependencies]
# HTTP downloads for model weights — minimal blocking client, no async runtime
ureq = "3"
# SHA-256 verification of downloaded artifacts (pinned hashes)
sha2 = "0.10"
# F32 -> F16 conversion for safetensors embeddings (reuses regular dependency)
half = "2"
# Parse/rewrite the safetensors JSON header during F32 -> F16 conversion
serde_json = "1"