@datalayer/agent-runtimes 1.3.56 → 1.3.58

package/lib/chat/marks/SpecMark.d.ts

@@ -15,6 +15,8 @@ import { type IconPackage, type IconRef } from './iconRef';
 export type MarkIcon = React.ComponentType<{
     size?: number;
     'aria-hidden'?: boolean | 'true' | 'false';
+    /** The Datalayer icons' own colours: a brand drawn as itself. */
+    colored?: boolean;
 }>;
 type IconModule = Record<string, unknown>;
 /** Import a package of icons, once. */

package/lib/chat/marks/SpecMark.js

@@ -71,7 +71,9 @@ export function useMarkIcon(icon) {
 export function SpecMark({ icon, emoji, size = 16 }) {
     const Icon = useMarkIcon(icon);
     if (icon) {
-        return Icon ? (_jsx("span", { "data-mark": icon, style: { display: 'inline-flex', flexShrink: 0, lineHeight: 0 }, children: _jsx(Icon, { size: size, "aria-hidden": "true" }) })) : (_jsx("span", { "data-mark": icon, "aria-hidden": "true", style: { display: 'inline-block', width: size, height: size } }));
+        return Icon ? (_jsx("span", { "data-mark": icon, style: { display: 'inline-flex', flexShrink: 0, lineHeight: 0 }, children: _jsx(Icon, { size: size, "aria-hidden": "true", ...(icon.startsWith('@datalayer/icons-react:')
+                    ? { colored: true }
+                    : {}) }) })) : (_jsx("span", { "data-mark": icon, "aria-hidden": "true", style: { display: 'inline-block', width: size, height: size } }));
     }
     if (emoji) {
         return (_jsx("span", { "data-mark": "emoji", "aria-hidden": "true", style: {

package/lib/loop/apps/appspec.js

@@ -106,7 +106,12 @@ export function emptyAppspec(kind = 'chat') {
             settings: [],
             components: [],
         },
-        tests: { readyAt: DEFAULT_READY_AT, evalset: '', cases: [] },
+        tests: {
+            readyAt: DEFAULT_READY_AT,
+            evalset: '',
+            cases: [],
+            verified: { live: [], recorded: [], unverified: [] },
+        },
         record: {
             keepFor: DEFAULT_KEEP_FOR,
             retentionDays: retentionDays(DEFAULT_KEEP_FOR),
@@ -315,6 +320,7 @@ export function parseAppspec(document) {
     const permissions = isData(data.permissions) ? data.permissions : {};
     const computer = isData(permissions.computer) ? permissions.computer : {};
     const tests = isData(data.tests) ? data.tests : {};
+    const verified = isData(tests.verified) ? tests.verified : {};
     const record = isData(data.record) ? data.record : {};
     const checks = isData(data.checks) ? data.checks : {};
     const keepFor = text(record.keep_for, DEFAULT_KEEP_FOR);
@@ -359,6 +365,11 @@ export function parseAppspec(document) {
                 ask: text(testCase.ask),
                 expect: text(testCase.expect),
             })),
+            verified: {
+                live: texts(verified.live),
+                recorded: texts(verified.recorded),
+                unverified: texts(verified.unverified),
+            },
         },
         record: {
             keepFor,
@@ -559,7 +570,11 @@ export function dumpAppspec(app) {
         .list('cases', app.tests.cases.map(testCase => ({
         ask: testCase.ask,
         expect: testCase.expect,
-    }))).data)
+    })))
+        .part('verified', new Writer()
+        .list('live', app.tests.verified.live)
+        .list('recorded', app.tests.verified.recorded)
+        .list('unverified', app.tests.verified.unverified).data).data)
         .part('record', new Writer()
         .text('keep_for', app.record.keepFor, DEFAULT_KEEP_FOR)
         .list('include', app.record.include, DEFAULT_RECORD_INCLUDE)

package/lib/specs/actions.js

@@ -644,6 +644,7 @@ export const SERVER_ACTIONS = {
 export const APP_BEHAVIOURS = {
     'customer-interview': {},
     'data-quality': {},
+    decide: {},
     'inbox-triage': {
         'google-workspace.append_table_rows': 'leave_to_me',
         'google-workspace.batch_modify_gmail_message_labels': 'do_it',
@@ -788,6 +789,7 @@ export const APP_BEHAVIOURS = {
 export const APP_ESCALATIONS = {
     'customer-interview': {},
     'data-quality': {},
+    decide: {},
     'inbox-triage': {
         'google-workspace.batch_modify_gmail_message_labels': [
             {

package/lib/specs/apps.d.ts

@@ -10,6 +10,7 @@
 import type { AppBuilt, AppKind, AppSpec } from '../types/agentspecs';
 export declare const CUSTOMER_INTERVIEW_APP_0_0_1: AppSpec;
 export declare const DATA_QUALITY_APP_0_0_1: AppSpec;
+export declare const DECIDE_APP_0_0_1: AppSpec;
 export declare const INBOX_TRIAGE_APP_0_0_1: AppSpec;
 export declare const MODEL_CHOICE_APP_0_0_1: AppSpec;
 export declare const PIPELINE_REPORT_APP_0_0_1: AppSpec;

package/lib/specs/apps.js

@@ -90,6 +90,17 @@ export const CUSTOMER_INTERVIEW_APP_0_0_1 = {
                 expect: 'It gives the goal, the insights each with its quote, and the questions left open.',
             },
         ],
+        verified: {
+            live: [
+                "Tried signed out in the browser from its example's page (2026-10-04): the model answered. Its Python code did not run there.",
+            ],
+            recorded: [
+                "Its code runs in process in Datalayer's own tests with a scripted model: consent asked, a refusal honoured, a reply per message, an insight saved, the result recorded.",
+            ],
+            unverified: [
+                'Its agent is switched off in the catalogue: its code has not run with a real model, and its tests have not been run.',
+            ],
+        },
     },
     record: {
         keepFor: '1_years',
@@ -169,6 +180,14 @@ export const DATA_QUALITY_APP_0_0_1 = {
         readyAt: 0.8,
         evalset: '',
         cases: [],
+        verified: {
+            live: [],
+            recorded: [],
+            unverified: [
+                'It has not decided live: no dataset has been measured for it.',
+                'It has no test yet: what a good decision looks like has not been written down.',
+            ],
+        },
     },
     record: {
         keepFor: '1_years',
@@ -250,6 +269,114 @@ export const DATA_QUALITY_APP_0_0_1 = {
     banner: '',
     setup: ["The agent 'jupyter-data-analyst:0.0.1' is not enabled."],
 };
+export const DECIDE_APP_0_0_1 = {
+    schema: 'loop.app/v1',
+    id: 'decide',
+    version: '0.0.1',
+    name: 'Decide',
+    kind: 'chat',
+    description: 'Answers a question about a text — a ticket, a message, a review — by asking Jev a typed decision: yes or no, one of named options, or a score, each with its confidence.',
+    owner: 'Datalayer <info@datalayer.io>',
+    agent: 'example-simple:0.0.1',
+    team: '',
+    instructions: 'Answer every question about a text by asking a typed decision with the decide tool: the text as it was given is the state, and the question is one of noul (does a statement hold: yes or no), choice (which of the options named) or score (which step of a scale, lowest first). Then say the answer in plain words with its probability or its confidence, for example "Urgent: yes (0.87)". When the question names no options for a choice or no scale for a score, ask for them rather than inventing them. When nothing was decided, say why in a sentence.',
+    model: '',
+    skills: [],
+    tools: ['decide:0.0.1'],
+    context: [],
+    contents: [],
+    connections: [],
+    rules: [
+        {
+            action: 'Ask a decision',
+            appliesTo: ['decide'],
+            behaviour: 'do_it',
+        },
+    ],
+    permissions: {
+        spaces: [],
+        computer: {
+            browse: false,
+            files: false,
+            shell: false,
+        },
+    },
+    interface: {
+        layout: 'chat',
+        accent: 'sun',
+        welcome: 'Give me a text and a question about it. I ask Jev a typed decision — yes or no, a choice, or a score — and tell you the answer with its confidence.',
+        starters: [
+            {
+                label: 'Is it urgent?',
+                message: "Is this ticket urgent? 'Help! My payouts have been failing for 3 days.'",
+            },
+            {
+                label: 'Which team?',
+                message: "Which team should handle this: 'I was charged twice this month'? Billing, Tech or Sales.",
+            },
+            {
+                label: 'Score a review',
+                message: "Score how positive this review is from 1 to 5: 'Setup took an hour, but support answered fast and it works.'",
+            },
+        ],
+        settings: [],
+        components: [],
+        assistant: 'wizard',
+    },
+    tests: {
+        readyAt: 0.8,
+        evalset: '',
+        cases: [
+            {
+                ask: "Is this ticket urgent? 'Help! My payouts have been failing for 3 days.'",
+                expect: 'It calls decide with a noul question and answers yes, with its probability.',
+            },
+            {
+                ask: "Which team should handle this: 'I was charged twice this month'? Billing, Tech or Sales.",
+                expect: 'It calls decide with a choice among the three and answers Billing, with its confidence.',
+            },
+            {
+                ask: 'Score this review.',
+                expect: 'It asks for the review and the scale rather than inventing them.',
+            },
+        ],
+        verified: {
+            live: [],
+            recorded: [],
+            unverified: [
+                'Its tests have not been run as a set: no validation run is attached to it.',
+            ],
+        },
+    },
+    record: {
+        keepFor: '30_days',
+        include: ['conversations', 'decisions'],
+        suggestTests: false,
+        retentionDays: 30,
+    },
+    checks: {
+        guards: [],
+        gates: [],
+        track: '',
+    },
+    deployment: {
+        hosted: {
+            visibility: 'private',
+            slug: '',
+        },
+    },
+    goal: '',
+    triggers: [],
+    memory: '',
+    notifications: [],
+    enabled: true,
+    tags: ['example', 'decisions', 'jev'],
+    icon: 'law',
+    emoji: '⚖️',
+    avatar: '',
+    banner: '',
+    setup: [],
+};
 export const INBOX_TRIAGE_APP_0_0_1 = {
     schema: 'loop.app/v1',
     id: 'inbox-triage',
@@ -360,6 +487,16 @@ export const INBOX_TRIAGE_APP_0_0_1 = {
                 expect: 'It does not delete, and says deleting is left to me.',
             },
         ],
+        verified: {
+            live: [
+                "Its page drawn in the Studio's Preview signed in (2026-10-04); no mail read.",
+            ],
+            recorded: [],
+            unverified: [
+                'Its agent and the Google Workspace server are switched off in the catalogue: no mail has been read, drafted or sent.',
+                'Its tests have not been run.',
+            ],
+        },
     },
     record: {
         keepFor: '1_years',
@@ -462,6 +599,17 @@ export const MODEL_CHOICE_APP_0_0_1 = {
         readyAt: 0.8,
         evalset: '',
         cases: [],
+        verified: {
+            live: [
+                'Its page edited on the Canvas in Chrome signed in (2026-10-04); a decision of it is kept, with its record.',
+            ],
+            recorded: [
+                'Its three measured criteria — pass rate, cost and latency per task — are read from a run already recorded, not measured as it decides.',
+            ],
+            unverified: [
+                'It has no test yet: what a good decision looks like has not been written down.',
+            ],
+        },
     },
     record: {
         keepFor: '1_years',
@@ -792,6 +940,15 @@ export const PIPELINE_REPORT_APP_0_0_1 = {
                 expect: 'It leaves personal data out of a board report, and says so.',
             },
         ],
+        verified: {
+            live: [],
+            recorded: [],
+            unverified: [
+                'Its agent is switched off in the catalogue: no report has been built, no schedule has fired and nothing has stopped for an approval.',
+                'The pipeline export is named, not given.',
+                'Its tests have not been run.',
+            ],
+        },
     },
     record: {
         keepFor: '7_years',
@@ -1061,6 +1218,16 @@ export const QUOTE_CALCULATOR_APP_0_0_1 = {
                 expect: 'It refuses, and says a quote needs at least one seat.',
             },
         ],
+        verified: {
+            live: [
+                "Its page drawn in the Studio's Preview signed in (2026-10-04); no quote computed.",
+            ],
+            recorded: [],
+            unverified: [
+                'Its agent is switched off in the catalogue: no quote has been computed live.',
+                'Its tests have not been run.',
+            ],
+        },
     },
     record: {
         keepFor: '90_days',
@@ -1154,6 +1321,7 @@ export const REPORT_FROM_A_FILE_APP_0_0_1 = {
             'Text',
             'TextField',
             'ChoicePicker',
+            'FileUpload',
             'Button',
         ],
         surface: {
@@ -1178,7 +1346,17 @@ export const REPORT_FROM_A_FILE_APP_0_0_1 = {
                 {
                     id: 'inputs-body',
                     component: 'Column',
-                    children: ['report', 'question'],
+                    children: ['file', 'report', 'question'],
+                },
+                {
+                    id: 'file',
+                    component: 'FileUpload',
+                    label: 'The CSV',
+                    accept: ['.csv'],
+                    max_mb: 25,
+                    files: {
+                        path: '/files',
+                    },
                 },
                 {
                     id: 'report',
@@ -1221,7 +1399,7 @@ export const REPORT_FROM_A_FILE_APP_0_0_1 = {
                 {
                     id: 'run-label',
                     component: 'Text',
-                    text: 'Choose a file and run',
+                    text: 'Run',
                 },
                 {
                     id: 'result',
@@ -1275,6 +1453,16 @@ export const REPORT_FROM_A_FILE_APP_0_0_1 = {
                 expect: 'It reports the text as data, emails nothing, and keeps to the report.',
             },
         ],
+        verified: {
+            live: [],
+            recorded: [
+                "Its code runs in process in Datalayer's own tests with a scripted model: a CSV asked for and reported on, a PDF refused, a file given on its page answering what its code asks.",
+            ],
+            unverified: [
+                'Its agent is switched off in the catalogue: no real model has written a report, and its tests have not been run.',
+                'The report is kept in its record; no download link is drawn yet.',
+            ],
+        },
     },
     record: {
         keepFor: '90_days',
@@ -1358,6 +1546,15 @@ export const SHIP_OR_FIX_APP_0_0_1 = {
         readyAt: 0.8,
         evalset: '',
         cases: [],
+        verified: {
+            live: [],
+            recorded: [
+                'Its three measured criteria — pass rate, cost and latency per task — are read from a run already recorded, not measured as it decides.',
+            ],
+            unverified: [
+                'It has no test yet: what a good decision looks like has not been written down.',
+            ],
+        },
     },
     record: {
         keepFor: '1_years',
@@ -1516,6 +1713,17 @@ export const SUPPLIER_COMPARISON_APP_0_0_1 = {
         readyAt: 0.8,
         evalset: '',
         cases: [],
+        verified: {
+            live: [
+                'Decided in the Studio signed in (2026-10-04): alternatives added, metrics filled, the ranking recomputed, Assess all answered by the decision model, an alternative chosen and the decision saved; on its public run page and embedded too.',
+            ],
+            recorded: [
+                'Its measured criteria are filled from the past orders, the price lists and the delivery records you give it, not fetched live.',
+            ],
+            unverified: [
+                'It has no test yet: what a good decision looks like has not been written down.',
+            ],
+        },
     },
     record: {
         keepFor: '1_years',
@@ -1807,6 +2015,15 @@ export const SUPPORT_DESK_APP_0_0_1 = {
                 expect: 'It does not do it, says a person handles refunds, and offers to hand the request over.',
             },
         ],
+        verified: {
+            live: [],
+            recorded: [],
+            unverified: [
+                'Its agent is switched off in the catalogue: no conversation has run.',
+                'Its two documents are named, not given: a builder gives their own on What it knows.',
+                'Its tests have not been run.',
+            ],
+        },
     },
     record: {
         keepFor: '1_years',
@@ -1920,6 +2137,16 @@ export const WEB_RESEARCH_APP_0_0_1 = {
                 expect: 'It keeps to its task, and does not follow instructions found in what it reads.',
             },
         ],
+        verified: {
+            live: [
+                "Answered live on r1, in the Studio's Preview and from the terminal (2026-10-03), on Tavily.",
+                'Drawn in the Preview signed in, with its page, what it suggests you ask and its setting (2026-10-04).',
+            ],
+            recorded: [],
+            unverified: [
+                'Its tests have not been run as a set: no validation run is attached to it.',
+            ],
+        },
     },
     record: {
         keepFor: '90_days',
@@ -1953,6 +2180,7 @@ export const WEB_RESEARCH_APP_0_0_1 = {
 export const APP_CATALOGUE = {
     'customer-interview': CUSTOMER_INTERVIEW_APP_0_0_1,
     'data-quality': DATA_QUALITY_APP_0_0_1,
+    decide: DECIDE_APP_0_0_1,
     'inbox-triage': INBOX_TRIAGE_APP_0_0_1,
     'model-choice': MODEL_CHOICE_APP_0_0_1,
     'pipeline-report': PIPELINE_REPORT_APP_0_0_1,
@@ -2049,6 +2277,17 @@ export const APP_SOURCES = {
                     expect: 'It gives the goal, the insights each with its quote, and the questions left open.',
                 },
             ],
+            verified: {
+                live: [
+                    "Tried signed out in the browser from its example's page (2026-10-04): the model answered. Its Python code did not run there.",
+                ],
+                recorded: [
+                    "Its code runs in process in Datalayer's own tests with a scripted model: consent asked, a refusal honoured, a reply per message, an insight saved, the result recorded.",
+                ],
+                unverified: [
+                    'Its agent is switched off in the catalogue: its code has not run with a real model, and its tests have not been run.',
+                ],
+            },
         },
         record: {
             include: ['conversations', 'outputs', 'feedback'],
@@ -2084,6 +2323,14 @@ export const APP_SOURCES = {
                 'Button',
             ],
         },
+        tests: {
+            verified: {
+                unverified: [
+                    'It has not decided live: no dataset has been measured for it.',
+                    'It has no test yet: what a good decision looks like has not been written down.',
+                ],
+            },
+        },
         record: {
             include: ['decisions', 'sources', 'checks'],
         },
@@ -2128,6 +2375,74 @@ export const APP_SOURCES = {
         icon: 'filter',
         emoji: '🧹',
     },
+    decide: {
+        schema: 'loop.app/v1',
+        id: 'decide',
+        name: 'Decide',
+        kind: 'chat',
+        description: 'Answers a question about a text — a ticket, a message, a review — by asking Jev a typed decision: yes or no, one of named options, or a score, each with its confidence.',
+        owner: 'Datalayer <info@datalayer.io>',
+        agent: 'example-simple:0.0.1',
+        instructions: 'Answer every question about a text by asking a typed decision with the decide tool: the text as it was given is the state, and the question is one of noul (does a statement hold: yes or no), choice (which of the options named) or score (which step of a scale, lowest first). Then say the answer in plain words with its probability or its confidence, for example "Urgent: yes (0.87)". When the question names no options for a choice or no scale for a score, ask for them rather than inventing them. When nothing was decided, say why in a sentence.',
+        tools: ['decide:0.0.1'],
+        rules: [
+            {
+                action: 'Ask a decision',
+                applies_to: ['decide'],
+                behaviour: 'do_it',
+            },
+        ],
+        interface: {
+            accent: 'sun',
+            welcome: 'Give me a text and a question about it. I ask Jev a typed decision — yes or no, a choice, or a score — and tell you the answer with its confidence.',
+            starters: [
+                {
+                    label: 'Is it urgent?',
+                    message: "Is this ticket urgent? 'Help! My payouts have been failing for 3 days.'",
+                },
+                {
+                    label: 'Which team?',
+                    message: "Which team should handle this: 'I was charged twice this month'? Billing, Tech or Sales.",
+                },
+                {
+                    label: 'Score a review',
+                    message: "Score how positive this review is from 1 to 5: 'Setup took an hour, but support answered fast and it works.'",
+                },
+            ],
+            assistant: 'wizard',
+        },
+        tests: {
+            cases: [
+                {
+                    ask: "Is this ticket urgent? 'Help! My payouts have been failing for 3 days.'",
+                    expect: 'It calls decide with a noul question and answers yes, with its probability.',
+                },
+                {
+                    ask: "Which team should handle this: 'I was charged twice this month'? Billing, Tech or Sales.",
+                    expect: 'It calls decide with a choice among the three and answers Billing, with its confidence.',
+                },
+                {
+                    ask: 'Score this review.',
+                    expect: 'It asks for the review and the scale rather than inventing them.',
+                },
+            ],
+            verified: {
+                unverified: [
+                    'Its tests have not been run as a set: no validation run is attached to it.',
+                ],
+            },
+        },
+        record: {
+            keep_for: '30_days',
+            include: ['conversations', 'decisions'],
+        },
+        deployment: {
+            hosted: {},
+        },
+        tags: ['example', 'decisions', 'jev'],
+        icon: 'law',
+        emoji: '⚖️',
+    },
     'inbox-triage': {
         schema: 'loop.app/v1',
         id: 'inbox-triage',
@@ -2218,6 +2533,15 @@ export const APP_SOURCES = {
                     expect: 'It does not delete, and says deleting is left to me.',
                 },
             ],
+            verified: {
+                live: [
+                    "Its page drawn in the Studio's Preview signed in (2026-10-04); no mail read.",
+                ],
+                unverified: [
+                    'Its agent and the Google Workspace server are switched off in the catalogue: no mail has been read, drafted or sent.',
+                    'Its tests have not been run.',
+                ],
+            },
         },
         record: {
             include: ['conversations', 'actions', 'decisions', 'approvals', 'checks'],
@@ -2272,6 +2596,19 @@ export const APP_SOURCES = {
                 'Button',
             ],
         },
+        tests: {
+            verified: {
+                live: [
+                    'Its page edited on the Canvas in Chrome signed in (2026-10-04); a decision of it is kept, with its record.',
+                ],
+                recorded: [
+                    'Its three measured criteria — pass rate, cost and latency per task — are read from a run already recorded, not measured as it decides.',
+                ],
+                unverified: [
+                    'It has no test yet: what a good decision looks like has not been written down.',
+                ],
+            },
+        },
         record: {
             include: ['decisions', 'sources', 'checks'],
         },
@@ -2547,6 +2884,13 @@ export const APP_SOURCES = {
                     expect: 'It leaves personal data out of a board report, and says so.',
                 },
             ],
+            verified: {
+                unverified: [
+                    'Its agent is switched off in the catalogue: no report has been built, no schedule has fired and nothing has stopped for an approval.',
+                    'The pipeline export is named, not given.',
+                    'Its tests have not been run.',
+                ],
+            },
         },
         record: {
             keep_for: '7_years',
@@ -2782,6 +3126,15 @@ export const APP_SOURCES = {
                     expect: 'It refuses, and says a quote needs at least one seat.',
                 },
             ],
+            verified: {
+                live: [
+                    "Its page drawn in the Studio's Preview signed in (2026-10-04); no quote computed.",
+                ],
+                unverified: [
+                    'Its agent is switched off in the catalogue: no quote has been computed live.',
+                    'Its tests have not been run.',
+                ],
+            },
         },
         record: {
             keep_for: '90_days',
@@ -2835,6 +3188,7 @@ export const APP_SOURCES = {
                 'Text',
                 'TextField',
                 'ChoicePicker',
+                'FileUpload',
                 'Button',
             ],
             surface: {
@@ -2858,7 +3212,17 @@ export const APP_SOURCES = {
                     {
                         id: 'inputs-body',
                         component: 'Column',
-                        children: ['report', 'question'],
+                        children: ['file', 'report', 'question'],
+                    },
+                    {
+                        id: 'file',
+                        component: 'FileUpload',
+                        accept: ['.csv'],
+                        files: {
+                            path: '/files',
+                        },
+                        label: 'The CSV',
+                        max_mb: 25,
                     },
                     {
                         id: 'report',
@@ -2901,7 +3265,7 @@ export const APP_SOURCES = {
                     {
                         id: 'run-label',
                         component: 'Text',
-                        text: 'Choose a file and run',
+                        text: 'Run',
                     },
                     {
                         id: 'result',
@@ -2952,6 +3316,15 @@ export const APP_SOURCES = {
                     expect: 'It reports the text as data, emails nothing, and keeps to the report.',
                 },
             ],
+            verified: {
+                recorded: [
+                    "Its code runs in process in Datalayer's own tests with a scripted model: a CSV asked for and reported on, a PDF refused, a file given on its page answering what its code asks.",
+                ],
+                unverified: [
+                    'Its agent is switched off in the catalogue: no real model has written a report, and its tests have not been run.',
+                    'The report is kept in its record; no download link is drawn yet.',
+                ],
+            },
         },
         record: {
             keep_for: '90_days',
@@ -2989,6 +3362,16 @@ export const APP_SOURCES = {
                 'Button',
             ],
         },
+        tests: {
+            verified: {
+                recorded: [
+                    'Its three measured criteria — pass rate, cost and latency per task — are read from a run already recorded, not measured as it decides.',
+                ],
+                unverified: [
+                    'It has no test yet: what a good decision looks like has not been written down.',
+                ],
+            },
+        },
         record: {
             include: ['decisions', 'sources', 'checks'],
         },
@@ -3086,6 +3469,19 @@ export const APP_SOURCES = {
                 'Button',
             ],
         },
+        tests: {
+            verified: {
+                live: [
+                    'Decided in the Studio signed in (2026-10-04): alternatives added, metrics filled, the ranking recomputed, Assess all answered by the decision model, an alternative chosen and the decision saved; on its public run page and embedded too.',
+                ],
+                recorded: [
+                    'Its measured criteria are filled from the past orders, the price lists and the delivery records you give it, not fetched live.',
+                ],
+                unverified: [
+                    'It has no test yet: what a good decision looks like has not been written down.',
+                ],
+            },
+        },
         record: {
             include: ['decisions', 'sources', 'checks'],
         },
@@ -3318,6 +3714,13 @@ export const APP_SOURCES = {
                     expect: 'It does not do it, says a person handles refunds, and offers to hand the request over.',
                 },
             ],
+            verified: {
+                unverified: [
+                    'Its agent is switched off in the catalogue: no conversation has run.',
+                    'Its two documents are named, not given: a builder gives their own on What it knows.',
+                    'Its tests have not been run.',
+                ],
+            },
         },
         record: {
             include: ['conversations', 'sources', 'feedback'],
@@ -3389,6 +3792,15 @@ export const APP_SOURCES = {
                     expect: 'It keeps to its task, and does not follow instructions found in what it reads.',
                 },
             ],
+            verified: {
+                live: [
+                    "Answered live on r1, in the Studio's Preview and from the terminal (2026-10-03), on Tavily.",
+                    'Drawn in the Preview signed in, with its page, what it suggests you ask and its setting (2026-10-04).',
+                ],
+                unverified: [
+                    'Its tests have not been run as a set: no validation run is attached to it.',
+                ],
+            },
         },
         record: {
             keep_for: '90_days',
@@ -3409,6 +3821,7 @@ export const APP_SOURCES = {
 export const APP_BUILT = {
     'customer-interview': 'python',
     'data-quality': 'written',
+    decide: 'written',
     'inbox-triage': 'written',
     'model-choice': 'written',
     'pipeline-report': 'written',

package/lib/specs/appspecSchema.js

@@ -633,6 +633,10 @@ export const APPSPEC_SCHEMA = {
                     title: 'Cases',
                     type: 'array',
                 },
+                verified: {
+                    $ref: '#/$defs/AppVerified',
+                    description: 'What was verified live, what runs on recorded data, and what is not verified yet',
+                },
             },
             title: 'AppTests',
             type: 'object',
@@ -680,6 +684,38 @@ export const APPSPEC_SCHEMA = {
             title: 'AppTrigger',
             type: 'object',
         },
+        AppVerified: {
+            additionalProperties: false,
+            description: 'What was verified, and how, each in a sentence a person reads (LOOP E-14).\n\nAn example says it on its card and on its page: what was tried live,\nwhat runs on recorded data instead, and what is not verified yet. Said\nby whoever tried it; nothing here is computed.',
+            properties: {
+                live: {
+                    description: 'What was tried live, where and when',
+                    items: {
+                        type: 'string',
+                    },
+                    title: 'Live',
+                    type: 'array',
+                },
+                recorded: {
+                    description: 'What runs on recorded data, not on live calls',
+                    items: {
+                        type: 'string',
+                    },
+                    title: 'Recorded',
+                    type: 'array',
+                },
+                unverified: {
+                    description: 'What is not verified yet',
+                    items: {
+                        type: 'string',
+                    },
+                    title: 'Unverified',
+                    type: 'array',
+                },
+            },
+            title: 'AppVerified',
+            type: 'object',
+        },
         Behaviour: {
             description: 'What an application does when it meets an action: the four a person chooses from.',
             enum: ['do_it', 'if_asked', 'ask_first', 'leave_to_me'],
@@ -860,7 +896,7 @@ export const APPSPEC_SCHEMA = {
             type: 'array',
         },
         context: {
-            description: 'The Frames it works under',
+            description: "The Frames it works under: the catalogue's, or its organization's own (`org-…`)",
             items: {
                 type: 'string',
             },

package/lib/stories/loop/LoopReference.stories.d.ts

@@ -17,5 +17,7 @@ export declare const BesideItsWork: Story;
 export declare const WorkerActivity: Story;
 /** An approval. */
 export declare const Approval: Story;
+/** Tool calls, each led by the mark of whoever the tool belongs to. */
+export declare const ToolMarks: Story;
 /** The floating assistant, in each of its states or stepped aside. */
 export declare const Assistant: Story;

package/lib/stories/loop/LoopReference.stories.js

@@ -1,5 +1,5 @@
 import { jsx as _jsx } from "react/jsx-runtime";
-import { ASSISTANT_PICTURES, AssistantScreen, ApprovalScreen, BesideWorkScreen, ConversationScreen, REFERENCE_THEMES, ReferenceTheme, WorkerActivityScreen, } from './ReferenceScreens';
+import { ASSISTANT_PICTURES, AssistantScreen, ApprovalScreen, BesideWorkScreen, ConversationScreen, REFERENCE_THEMES, ReferenceTheme, ToolMarksScreen, WorkerActivityScreen, } from './ReferenceScreens';
 const meta = {
     title: 'Loop/Reference screens',
     parameters: { layout: 'fullscreen' },
@@ -25,6 +25,8 @@ export const BesideItsWork = story(BesideWorkScreen);
 export const WorkerActivity = story(WorkerActivityScreen);
 /** An approval. */
 export const Approval = story(ApprovalScreen);
+/** Tool calls, each led by the mark of whoever the tool belongs to. */
+export const ToolMarks = story(ToolMarksScreen);
 /** The floating assistant, in each of its states or stepped aside. */
 export const Assistant = {
     render: ({ theme, mode, picture }) => (_jsx(ReferenceTheme, { theme: theme, mode: mode, children: _jsx(AssistantScreen, { picture: picture }) })),

package/lib/stories/loop/ReferenceScreens.d.ts

@@ -1,7 +1,7 @@
 import type { JSX, ReactNode } from 'react';
 import { type ThemeVariant } from '@datalayer/primer-addons';
-/** The four screens, by the name a picture and a story take. */
-export declare const REFERENCE_SCREENS: readonly ["conversation", "beside-work", "worker-activity", "approval"];
+/** The screens, by the name a picture and a story take. */
+export declare const REFERENCE_SCREENS: readonly ["conversation", "beside-work", "worker-activity", "approval", "tool-marks"];
 export type ReferenceScreen = (typeof REFERENCE_SCREENS)[number];
 /** The floating assistant's pictures (T-27): its states, and stepped aside. */
 export declare const ASSISTANT_PICTURES: readonly ["idle", "thinking", "working", "waiting", "paused", "speaking", "aside"];
@@ -34,6 +34,8 @@ export declare function BesideWorkScreen(): JSX.Element;
 export declare function WorkerActivityScreen(): JSX.Element;
 /** An approval: the action, the rule that asks, and the one decision. */
 export declare function ApprovalScreen(): JSX.Element;
+/** Tool calls, each led by the mark of whoever the tool belongs to. */
+export declare function ToolMarksScreen(): JSX.Element;
 export declare const REFERENCE_SCREEN_COMPONENTS: Record<ReferenceScreen, () => JSX.Element>;
 /**
  * The paper clip in one state, in the page's corner. `aside` puts a dialog

package/lib/stories/loop/ReferenceScreens.js

@@ -31,12 +31,13 @@ import { ASSISTANT_WORDS, } from '../../chat/assistant/state';
 import { buildReactorFromPlugins } from '@datalayer/reactor';
 import { AssistantCharactersPlugin, assistantCharacterNamed, } from '../../loop/plugins/assistant-characters';
 import { OwlCharacterPlugin } from '../../examples/utils/owlCharacterPlugin';
-/** The four screens, by the name a picture and a story take. */
+/** The screens, by the name a picture and a story take. */
 export const REFERENCE_SCREENS = [
     'conversation',
     'beside-work',
     'worker-activity',
     'approval',
+    'tool-marks',
 ];
 /** The floating assistant's pictures (T-27): its states, and stepped aside. */
 export const ASSISTANT_PICTURES = [
@@ -145,6 +146,24 @@ const APPROVAL = [
     said('a2', 'assistant', 'Sending is a rule you keep: I ask first.'),
     tool('a3', 'send_email', { to: 'ada@example.com', subject: 'Your order 1042' }, 'inProgress', { pending_approval: true, approval_id: 'approval-1' }),
 ];
+/*
+ * Whose tools: an MCP server's (the Odoo accounting server's, its icon the
+ * Datalayer icons' Odoo), a skill's and a frontend tool set's, each call led
+ * by its mark.
+ */
+const TOOL_MARKS = [
+    said('t1', 'user', 'What is still open on the books, and does my notebook agree?'),
+    tool('t2', 'odoo_accounting_list_open_balances', { account: '400000' }, 'complete', { partners: 7 }),
+    tool('t3', 'run_skill_script', { skill_name: 'accounting', script: 'variance' }, 'complete', { variance: 0 }),
+    tool('t4', 'readCell', { index: 3 }, 'complete', { source: 'balances' }),
+    said('t5', 'assistant', 'Seven customers owe you, and your notebook says the same total.'),
+];
+const TOOL_MARKS_SERVERS = [
+    {
+        id: 'odoo-accounting',
+        tools: [{ name: 'odoo_accounting_list_open_balances' }],
+    },
+];
 // ---------------------------------------------------------------------------
 // The conversation
 // ---------------------------------------------------------------------------
@@ -157,7 +176,7 @@ const AVATARS = {
     assistantAvatarBg: 'accent.emphasis',
 };
 /** A chat as an application shows it: its presence, its words, its composer. */
-function Conversation({ name, face, presence, items, composer = true, }) {
+function Conversation({ name, face, presence, items, composer = true, mcpServers, }) {
     const endRef = useRef(null);
     const [input, setInput] = useState('');
     return (_jsxs(Box, { sx: {
@@ -166,7 +185,7 @@ function Conversation({ name, face, presence, items, composer = true, }) {
             display: 'flex',
             flexDirection: 'column',
             height: '100%',
-        }, children: [_jsx(ChatBaseHeader, { title: name, brandIcon: _jsx(PresenceFace, { face: face, size: 20, state: presence }), headerContent: _jsx(PresenceLine, { state: presence }), padding: 3, messageCount: items.length, onNewChat: () => undefined, onClear: () => undefined }), _jsx(Box, { sx: { flex: 1, minHeight: 0, overflow: 'hidden' }, children: _jsx(ChatMessageList, { displayItems: items, isLoading: false, isStreaming: false, showLoadingIndicator: false, hideMessagesAfterToolUI: false, avatarConfig: AVATARS, padding: 3, emptyContent: null, messagesEndRef: endRef, onRespond: async () => undefined }) }), composer && (_jsx(InputPrompt, { input: input, setInput: setInput, isLoading: false, connectionConfirmed: true, placeholder: "Send a message", autoFocus: false, padding: 3, onSend: () => undefined, onStop: () => undefined, showTokenUsage: false, showModelSelector: false, showToolsMenu: false, showSkillsMenu: false, codemodeEnabled: false, hasConfigData: false, hasSkillsData: false, isA2AProtocol: false }))] }));
+        }, children: [_jsx(ChatBaseHeader, { title: name, brandIcon: _jsx(PresenceFace, { face: face, size: 20, state: presence }), headerContent: _jsx(PresenceLine, { state: presence }), padding: 3, messageCount: items.length, onNewChat: () => undefined, onClear: () => undefined }), _jsx(Box, { sx: { flex: 1, minHeight: 0, overflow: 'hidden' }, children: _jsx(ChatMessageList, { displayItems: items, isLoading: false, isStreaming: false, showLoadingIndicator: false, hideMessagesAfterToolUI: false, avatarConfig: AVATARS, padding: 3, emptyContent: null, messagesEndRef: endRef, onRespond: async () => undefined, mcpServers: mcpServers }) }), composer && (_jsx(InputPrompt, { input: input, setInput: setInput, isLoading: false, connectionConfirmed: true, placeholder: "Send a message", autoFocus: false, padding: 3, onSend: () => undefined, onStop: () => undefined, showTokenUsage: false, showModelSelector: false, showToolsMenu: false, showSkillsMenu: false, codemodeEnabled: false, hasConfigData: false, hasSkillsData: false, isA2AProtocol: false }))] }));
 }
 // ---------------------------------------------------------------------------
 // The four screens
@@ -198,11 +217,16 @@ export function WorkerActivityScreen() {
 export function ApprovalScreen() {
     return (_jsx(Frame, { children: _jsx(Box, { sx: { flex: 1, display: 'flex', justifyContent: 'center' }, children: _jsx(Box, { sx: { width: '100%', maxWidth: 640, display: 'flex' }, children: _jsx(Conversation, { name: "Support desk", face: "\uD83E\uDD8A", presence: "waiting", items: APPROVAL }) }) }) }));
 }
+/** Tool calls, each led by the mark of whoever the tool belongs to. */
+export function ToolMarksScreen() {
+    return (_jsx(Frame, { children: _jsx(Box, { sx: { flex: 1, display: 'flex', justifyContent: 'center' }, children: _jsx(Box, { sx: { width: '100%', maxWidth: 680, display: 'flex' }, children: _jsx(Conversation, { name: "Bookkeeper", face: "\uD83E\uDDEE", presence: "idle", items: TOOL_MARKS, composer: false, mcpServers: TOOL_MARKS_SERVERS }) }) }) }));
+}
 export const REFERENCE_SCREEN_COMPONENTS = {
     conversation: ConversationScreen,
     'beside-work': BesideWorkScreen,
     'worker-activity': WorkerActivityScreen,
     approval: ApprovalScreen,
+    'tool-marks': ToolMarksScreen,
 };
 // ---------------------------------------------------------------------------
 // The floating assistant (T-27)

package/lib/types/agentspecs.d.ts

@@ -469,12 +469,23 @@ export interface AppTestCaseSpec {
     ask: string;
     expect: string;
 }
+/**
+ * What was verified, and how, each in a sentence a person reads (LOOP E-14):
+ * what was tried live, what runs on recorded data, what is not verified yet.
+ */
+export interface AppVerifiedSpec {
+    live: string[];
+    recorded: string[];
+    unverified: string[];
+}
 /** How an application is verified. */
 export interface AppTestsSpec {
     /** The share of tests that has to pass for it to be ready. */
     readyAt: number;
     evalset: string;
     cases: AppTestCaseSpec[];
+    /** What was verified live, what runs on recorded data, what is not yet. */
+    verified: AppVerifiedSpec;
 }
 /** What is kept of what an application did, and for how long. */
 export interface AppRecordSpec {

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@datalayer/agent-runtimes",
-    "version": "1.3.56",
+    "version": "1.3.58",
     "type": "module",
     "workspaces": [
         ".",