Repository navigation
Expand file tree
/
Copy pathdata.ts
More file actions
234 lines (222 loc) · 12.4 KB
/
Copy pathdata.ts
File metadata and controls
234 lines (222 loc) · 12.4 KB
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
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
// The page copy is the maintainer's, supplied 2026-09-07 and used as written.
// Do not add sentences here: earlier versions of this file accumulated
// carried-over prose from previous drafts, and the page said more than the
// author had written.
//
// Every figure in MEASURED was checked against the frozen claims table in
// `reference/docs/CLAIMS.md` before it landed; the `claim` field records which
// row carries it, so the check is repeatable rather than remembered.
export const ORG = {
name: 'Agentic Datasets',
tagline:
'A machine-readable contract that lets datasets declare what AI agents can do — and makes those decisions observable, portable, and testable.',
subline:
'15 normative assertions · language-neutral conformance vectors · tested across multiple agent runtimes and dataset boundaries',
github: 'https://github.com/agentic-datasets',
docs: '/reference/',
showcase: '/showcase/',
maintainer: 'Alexander Chernov',
orcid: 'https://orcid.org/0009-0007-3198-2712',
orcidLabel: '0009-0007-3198-2712',
} as const
// Vision, Qubit and Spectra are named by their own page titles and linked at
// their live addresses: their sources are not in this organization, so there
// is no repository to point at.
export const NAV = [
{ label: 'Read the manifesto', href: '#manifesto' },
{ label: 'Explore the showcases', href: '/showcase/' },
{ label: 'Explore Agentic Vision', href: 'https://vision.agenticdatasets.org/' },
{ label: 'Explore Agentic Spectra', href: 'https://spectra.agenticdatasets.org/' },
{ label: 'Explore Agentic Qubit', href: 'https://qubit.agenticdatasets.org/' },
{ label: 'Read the specification', href: '/reference/' },
{ label: 'View the repositories', href: 'https://github.com/agentic-datasets' },
] as const
export const DEFINITION = {
heading: 'What is an Agentic Dataset?',
statement:
'An Agentic Dataset carries a machine-readable description of its capabilities, policies, and operating boundaries.',
body: [
'Before an agent acts on the dataset, a control plane evaluates the requested action against that contract and records the decision and supporting evidence.',
'This moves governance from documentation around the dataset into the execution path itself.',
],
verdictsLead: 'The contract produces three explicit admission outcomes:',
verdicts: [
{ term: 'Approved', gloss: 'the requested action is admissible under the descriptor' },
{ term: 'Refused', gloss: 'a declared policy prevents the requested action' },
{
term: 'Indeterminate',
gloss: 'the available descriptor does not establish whether the action is admissible',
},
],
close: [
'Each outcome is attributable to the descriptor evidence that produced it.',
'The result is a dataset interface that participates directly in governed agentic execution.',
],
} as const
// The manifesto is the maintainer's, approved as written 2026-09-29. Its source is
// `programme/MANIFESTO.md`; keep the two identical. Each point names the
// assertions in `reference/CONFORMANCE.md` that it restates, so every line
// traces to the contract rather than adding to it.
export const MANIFESTO = {
heading: 'Manifesto: what makes a dataset agentic',
lead: 'A data system is an agentic dataset when these can be found in its implementation, not just in its documentation.',
points: [
{
title: 'It describes itself.',
body: "A machine-readable descriptor states the dataset's identity, capabilities and prohibitions. A dataset without a well-formed descriptor takes no part in admission.",
asserts: 'AD-001',
},
{
title: 'Only what it advertises can run.',
body: 'Every executable action is a registered capability; an unregistered one is denied, never allowed by default.',
asserts: 'AD-002, AD-006',
},
{
title: 'It decides before anything runs.',
body: 'A control plane rules on each requested action before it executes: Approved, Refused or Indeterminate, each with a typed reason.',
asserts: '',
},
{
title: 'Only approval grants authority.',
body: 'Execution requires an approval token, and only Approved mints one. After a refusal, or an unknown, there is nothing to execute with.',
asserts: 'AD-003, AD-004, AD-005',
},
{
title: 'Scope never widens.',
body: 'What was admitted is what runs: no wider at execution, through a cache, through a remote tool server, or in a handoff to another agent.',
asserts: 'AD-007, AD-008, AD-013, AD-014',
},
{
title: 'Every decision leaves evidence.',
body: 'Every decision gets an append-only, hash-chained record, refusals included. It names the descriptor clause, the dataset revision and the policy version that applied, and every result traces to what produced it.',
asserts: 'AD-009 to AD-012',
},
{
title: 'Prohibited means never.',
body: 'A prohibited action executes zero times, not rarely.',
asserts: 'AD-015',
},
{
title: 'It is checked by behavior, not wording.',
body: 'Each point above is checked through the public interface, without an LLM, by the absence of authority rather than the wording of a refusal. That means any implementation, in any framework, is checked the same way.',
asserts: '',
},
],
close: [
'Refusal is not a message.',
'Meeting these points is a claim an implementation makes about itself, made checkable. It is not a security guarantee.',
],
trace: 'The assertions cited are those of the behavioral contract, and its language-neutral vectors check them.',
source: 'https://github.com/agentic-datasets/programme/blob/main/MANIFESTO.md',
} as const
export const CONTRACT = {
heading: 'A portable behavioral contract',
lead: 'Agentic Datasets defines behavior rather than prescribing a particular implementation.',
body: [
'The contract consists of 15 normative assertions represented through language-neutral executable vectors. Implementations can therefore be evaluated through their public interfaces without depending on their internal architecture or programming language.',
'The same vectors are demonstrated directly in the browser and exercised against the reference implementation, agent-framework integrations, and an implementation sharing no reference code.',
],
} as const
export const MEASURED = {
heading: 'Measured portability',
figures: [
{ figure: '15 / 15', label: 'Normative assertions satisfied through the public interface', claim: 4 },
{ figure: '85', label: 'Language-neutral executable vector steps', claim: 2 },
{ figure: '8', label: 'Measured combinations across four agent runtimes and two dataset boundaries', claim: 5 },
{ figure: '15 / 15', label: 'Assertions passed by a code-independent implementation', claim: 6 },
{ figure: '17 / 17', label: 'Targeted behavioral defects detected by their corresponding assertions', claim: 7 },
{ figure: '0', label: 'Policy-prohibited actions executed across the measured conformance matrix', claim: 11 },
],
close: [
'These measurements make portability testable rather than assumed.',
'The reference repository contains the vectors, conformance suite, measurements, and reproducible evidence behind each result.',
],
} as const
export const ARTIFACTS = {
heading: 'From contract to implementation',
lead: 'The programme is organized around several complementary artifacts.',
items: [
{
name: 'Reference',
href: 'https://github.com/agentic-datasets/reference',
summary:
'The normative behavioral contract, language-neutral vectors, reference control plane, conformance suite, Authorized Recall@K, and measured results.',
},
{
name: 'Programme',
href: 'https://github.com/agentic-datasets/programme',
summary:
'The conceptual model, terminology, architecture, and licensing boundary for Agentic Datasets.',
},
{
name: 'Showcase',
href: 'https://github.com/agentic-datasets/showcase',
summary:
'Interactive browser demonstrations of the contract using the normative vectors and a TypeScript implementation.',
},
],
close:
'Together they separate the idea, the executable contract, and its implementations while keeping them independently inspectable.',
} as const
export const VALIDATION = {
heading: 'Validation',
body: [
"The current conformance work tests observable behavior through an implementation's public interface.",
'Evidence includes multiple agent runtimes, local and MCP dataset boundaries, mutation testing, public-interface isolation, and a code-independent implementation.',
'The next stage is broader independent implementation and third-party conformance evidence.',
'The contract is designed so independent implementations can be built and evaluated without permission from the programme.',
],
} as const
export const PROGRAMME = {
heading: 'An open technical programme',
body: [
'Agentic Datasets has a dedicated organizational home so the contract can be implemented, tested, cited, and extended independently of any one repository.',
'The specification, executable vectors, and conformance tooling use open licenses chosen to support independent implementation and experimentation.',
'Authorship and provenance remain explicit through citation metadata, ORCID, repository history, and associated publications.',
],
} as const
export const MARK = {
heading: 'The mark',
body: [
'The Agentic Datasets mark expresses the architecture in two shapes: a governed boundary and an admitted path to the dataset.',
'The outer square represents the boundary. The opening represents the controlled path through it. The inner square represents the dataset.',
'The mark belongs to the Agentic Datasets programme and provides a consistent identity across its specifications, implementations, demonstrations, and documentation.',
],
href: 'https://github.com/agentic-datasets/reference/blob/main/brand/README.md',
} as const
export const STATUS = {
heading: 'Current status',
lead: 'Agentic Datasets is a release candidate under active validation.',
body: [
'The behavioral contract, executable vectors, reference control plane, conformance suite, framework integrations, and browser demonstrations are available now.',
'The programme welcomes independent implementations, conformance results, integrations, technical discussion, and research building on the contract.',
],
call: 'Implement the contract. Run the vectors. Compare the behavior.',
} as const
// The notice, in two forms. The footer carries the summary; /legal.html carries
// the terms, and the terms are what govern. Both live here so the summary
// cannot come to say something the full text does not.
export const LEGAL = {
summary:
'Agentic Datasets is an independent research and technical project, unaffiliated with the author’s employer. Nothing on this site represents the employer’s views. Materials are provided as is, without warranty.',
ownership:
'Site and original code © 2026 Alexander Chernov. Site text is licensed CC BY 4.0. The Agentic Datasets mark is all rights reserved; see the brand notes.',
terms: [
{
title: 'No warranty',
body: 'Materials on this site are provided as is, for demonstration and research only. No warranty of any kind, express or implied, including without limitation any warranty of merchantability, fitness for a particular purpose, accuracy, availability or non-infringement. No claim is made that they are complete, correct, reliable or continuously available, and they may change or be withdrawn without notice.',
},
{
title: 'Affiliation',
body: 'Agentic Datasets is an independent technical project, unaffiliated with the author’s employer. Nothing on this site represents the employer’s views.',
},
{
title: 'Liability',
body: 'Use is entirely at your own risk, and to the fullest extent permitted by law the author accepts no liability or responsibility for any loss or damage whatsoever arising from its use.',
},
{
title: 'Licensing',
body: 'Site and original code © 2026 Alexander Chernov. The text of this site is licensed CC BY 4.0. The specification is CC BY 4.0, the vectors are CC0, and the conformance software is Apache-2.0, so anyone can implement the contract without asking. The Agentic Datasets mark is the exception: all rights reserved, because an identifier anyone may modify identifies nothing.',
},
],
} as const