artifactize 0.22.0

The AI-native collaboration layer for one-of-a-kind teammates and their agents: split a project into Artifacts with their own evals, and reuse every review while it still holds.
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

name = "engine-cli"

[views.agent_tools]
list = { builtin = "list" }
glob = { builtin = "glob" }
grep = { builtin = "grep" }
read = { builtin = "read" }

[evals.follows-style]
title = "The CLI code follows the code checklist"
profile = { kind = "agent", backend = "codex", model = "gpt-6.1-sol", reasoning = "xhigh", timeout_ms = 1800000 }

[evals.follows-style.profile_variants]
sol-medium = { kind = "agent", backend = "codex", model = "gpt-6.1-sol", reasoning = "medium", timeout_ms = 1800000 }
sol-max = { kind = "agent", backend = "codex", model = "gpt-6.1-sol", reasoning = "max", timeout_ms = 1800000 }
luna-max = { kind = "agent", backend = "codex", model = "gpt-6-luna", reasoning = "max", timeout_ms = 1800000 }

[evals.follows-style.payload]
instruction = """
Read the checklist {rules-code-style}. Then review the Rust code of {engine-cli} (src/ \
and Cargo.toml; tests/ for M-TAUTOLOGICAL-TESTS and every TEST- rule) against every item. Tool paths are relative to the Artifact, for example src/main.rs. Submit \
GREEN when no item is clearly violated. Otherwise submit RED with up to 20 violations, the most \
important first, each with the item id, a path:line and one sentence on what to change."""

[evals.follows-style.pass_schema]
type = "object"
properties.summary = { type = "string", minLength = 1 }
required = ["summary"]
additionalProperties = false

[evals.follows-style.fail_schema]
type = "object"
required = ["violations"]
additionalProperties = false

[evals.follows-style.fail_schema.properties.violations]
type = "array"
minItems = 1
maxItems = 20

[evals.follows-style.fail_schema.properties.violations.items]
type = "object"
properties.item = { type = "string", minLength = 1 }
properties.where = { type = "string", minLength = 1 }
properties.change = { type = "string", minLength = 1 }
required = ["item", "where", "change"]
additionalProperties = false

[evals.follows-platform]
title = "The CLI code follows the platform checklist"
profile = { kind = "agent", backend = "codex", model = "gpt-6.1-sol", reasoning = "xhigh", timeout_ms = 1800000 }

[evals.follows-platform.profile_variants]
sol-medium = { kind = "agent", backend = "codex", model = "gpt-6.1-sol", reasoning = "medium", timeout_ms = 1800000 }
sol-max = { kind = "agent", backend = "codex", model = "gpt-6.1-sol", reasoning = "max", timeout_ms = 1800000 }
luna-max = { kind = "agent", backend = "codex", model = "gpt-6-luna", reasoning = "max", timeout_ms = 1800000 }

[evals.follows-platform.payload]
instruction = """
Read the checklist {rules-platform}. Then review the Rust code of {engine-cli} (src/, \
tests/ and Cargo.toml) against every P- item. Tool paths are relative to the Artifact, for example src/main.rs. Submit \
GREEN when no item is clearly violated. Otherwise submit RED with up to 20 violations, the most \
important first, each with the item id, a path:line and one sentence on what to change."""

[evals.follows-platform.pass_schema]
type = "object"
properties.summary = { type = "string", minLength = 1 }
required = ["summary"]
additionalProperties = false

[evals.follows-platform.fail_schema]
type = "object"
required = ["violations"]
additionalProperties = false

[evals.follows-platform.fail_schema.properties.violations]
type = "array"
minItems = 1
maxItems = 20

[evals.follows-platform.fail_schema.properties.violations.items]
type = "object"
properties.item = { type = "string", minLength = 1 }
properties.where = { type = "string", minLength = 1 }
properties.change = { type = "string", minLength = 1 }
required = ["item", "where", "change"]
additionalProperties = false

[evals.fmt]
title = "The Rust formatting check passes"
profile = { kind = "runtime", command = "cargo", args = ["fmt", "--check", "-p", "artifactize"], timeout_ms = 60000 }

[evals.fmt.payload]
instruction = "Check the formatting of {engine-cli}."

[evals.clippy]
title = "Clippy passes for every Rust target"
profile = { kind = "runtime", command = "cargo", args = ["clippy", "--locked", "-p", "artifactize", "--all-targets", "--", "-D", "warnings"], timeout_ms = 1800000 }

[evals.clippy.payload]
instruction = "Check every Rust target of {engine-cli} with {engine-core} with {engine-tools} with {engine-tui} with {engine-mcp} with the workspace {engine-cargo-manifest} and {engine-cargo-lock}, treating every clippy warning as an error."

[evals.tests]
title = "The Rust crate tests pass"
profile = { kind = "runtime", command = "cargo", args = ["test", "--locked", "-p", "artifactize"], timeout_ms = 1800000 }

[evals.tests.payload]
instruction = "Run the tests of {engine-cli} with {engine-core} with {engine-tools} with {engine-tui} with {engine-mcp} with the workspace {engine-cargo-manifest} and {engine-cargo-lock}. Fixtures build or declare every executable they use and own their resources."

[evals.publish]
title = "Every published Rust crate packages and builds"
profile = { kind = "runtime", command = "node", args = ["publish.mjs"], timeout_ms = 1800000 }

[evals.publish.payload]
instruction = "Package and verify {engine-cli}, {engine-core}, {engine-tools}, {engine-tui} and {engine-mcp} with the workspace {engine-cargo-manifest} and {engine-cargo-lock} exactly as publication would, without uploading. Allow the builder's uncommitted candidate; the package verification still builds every crate."