@vllm-sr/dsh 0.4.1 → 0.4.2

package/README.md

@@ -11,8 +11,8 @@ agent orders, is the work really done, which team should get each of these 230 t
 - **System One**, the decision model, decides: `vllm-sr/d3-ultra`. It answers typed questions (choice, yes/no, score,
   box) with probabilities, from one forward pass per question instead of generated text.
 
-Both come from [System Duo](https://app.vllm-sr.ai) with one free key, and the plugin gets that key for you: System One
-answers at `https://app.vllm-sr.ai/v1/systemone`, and `vllm-sr/auto` at `https://app.vllm-sr.ai/v1/chat/completions`.
+Both take a free [System Duo](https://app.vllm-sr.ai) key, which the plugin gets for you: System One answers at `https://app.vllm-sr.ai/v1/systemone`,
+and `vllm-sr/auto` at `https://app.vllm-sr.ai/v1/chat/completions`.
 
 ## Quick start
 
@@ -26,9 +26,10 @@ The first line installs DeepSeek Harness (0.2.1-rc.5 or a later 0.2 release, on
 `dsh plugin` uses to install plugins; without pnpm, `dsh plugin add` stops with "pnpm was not found". On a machine
 with git, the package can also come straight from GitHub: `dsh plugin --profile web add github:Xunzhuo/dsh-system-one-plugin`.
 
-On its first start the plugin gets one free System Duo key, with the same small proof of work the site's "Get your
-API key" button runs in the browser, and saves it in DeepSeek Harness's credential store and in `~/.dsh/.env` (readable
-by you only). A notice in the session says so. `vllm-sr/auto` becomes the default model; choose another model
+On its first start the plugin gets a free System Duo key, with the same small proof of work the site's "Get your API
+key" button runs in the browser, and saves it in DeepSeek Harness's credential store and in `~/.dsh/.env` (readable
+by you only). A line under your first request says so: "Free API key created for app.vllm-sr.ai, saved in
+~/.dsh/.env", or why the site refused one. `vllm-sr/auto` becomes the default model; choose another model
 in the app and System One keeps working beside it.
 
 Headless:
@@ -70,11 +71,13 @@ badge that carries the decision's latency:
 - **Done-check**: a card after the answer: done, waiting on you, or sent back to finish.
 - **`system_one` tool**: its row shows the batch: items, questions, each question's answer distribution and the
   results file.
+- **Free key**: a **System Duo** line under your first request when the plugin created a key, or could not.
 
 Routine checks (an allowed call, a clean result) are one line and fold with the turn's work. Cards stay visible when
 the work folds, so a finished turn still shows what System One decided. Parallel calls in one step share one line or
 card, and every decision pops in with a short animation the first time it appears (none with reduced motion). Each
-element shows the Decision API's own answer: the probabilities, the model and the measured latency.
+element shows the Decision API's own answer: the probabilities, the model (in bold, as `vllm-sr/d3-ultra`) and the
+measured latency.
 
 Set `systemOneUi` to `compact` for one line per decision, or to `off` to hide them.
 
@@ -85,18 +88,20 @@ attributes that say what it shows:
 
 | Attribute | Values |
 | --- | --- |
-| `data-system-one` | `card`, `compact` (one line) or `tool` (the `system_one` row) |
-| `data-system-one-reflex` | `triage`, `guard`, `shield`, `done` or `tool` |
-| `data-system-one-verdict` | triage `brief`; guard `allow`, `ask` or `deny`; shield `clean`, `removed` or `flagged`; done-check `stop` or `continue`; tool `running` or `done`; `error` when System One did not answer |
+| `data-system-one` | `card`, `compact` (one line), `tool` (the `system_one` row) or `notice` (the free-key line) |
+| `data-system-one-reflex` | `triage`, `guard`, `shield`, `done`, `tool` or `key` |
+| `data-system-one-verdict` | triage `brief`; guard `allow`, `ask` or `deny`; shield `clean`, `removed` or `flagged`; done-check `stop` or `continue`; tool `running` or `done`; key `created` or `failed`; `error` when System One did not answer |
 | `data-system-one-outcome` | guard, once you answer its approval: `allowed`, `rejected`, `cancelled` or `failed` |
 | `data-system-one-why` | done-check: `complete`, `waiting`, `unfinished` or `unchecked` |
 | `data-system-one-tone` | its colour: `ok` (green), `info` (ember), `warn` (amber), `bad` (red) or `muted` |
 | `data-system-one-count` | how many decisions it holds (a batch: how many items) |
 | `data-system-one-turn`, `data-system-one-step` | where it acted (the step for guard and shield) |
+| `data-system-one-model` | the model that answered, as `vllm-sr/d3-ultra` |
 
 Inside a root, `data-system-one-part` marks `badge` (the ember **System One** pill), `latency`, `reflex`, `summary`,
 `model`, `head`, `toggle` (the button of a one-line element), `rows`, `verdict`, `outcome`, `subject`, `facts`,
-`note` and `more`, and in the batch row also `stack`, `legend`, `failed`, `results` and `raw`. Each number is a
+`note` and `more`, in the batch row also `stack`, `legend`, `failed`, `results` and `raw`, and in the free-key line
+`badge` (the **System Duo** pill), `summary` and `links`. Each number is a
 `data-system-one-fact`: `kind`, `effort`, `unclear` and `risky` (triage), `risk`, `leaks` and `on-task` (guard),
 `passages` and `injection` (shield), `complete`, `checked` and `waiting` (done-check). Decision rows carry
 `data-system-one-decision` and `data-system-one-call`, batch questions `data-system-one-question`. For example:
@@ -113,6 +118,8 @@ Inside a root, `data-system-one-part` marks `badge` (the ember **System One** pi
 | The ember badge, and its latency | `[data-system-one-part="badge"]`, `[data-system-one-part="latency"]` |
 | "Dangerous", and "You rejected it" | `[data-system-one-fact="risk"]`, `[data-system-one-part="outcome"]` |
 | The reason | `[data-system-one-part="note"]` |
+| The model's name (bold) | `[data-system-one-part="model"]` |
+| Free key created | `[data-system-one="notice"][data-system-one-verdict="created"]` |
 
 ## Your data
 

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@vllm-sr/dsh",
-  "version": "0.4.1",
-  "description": "System Duo for DeepSeek Harness. Thinking, fast and slow: System One decision models guard, shield, triage and check your agent around System Two (vllm-sr/auto), with one free app.vllm-sr.ai key.",
+  "version": "0.4.2",
+  "description": "System Duo for DeepSeek Harness. Thinking, fast and slow: System One decision models guard, shield, triage and check your agent around System Two (vllm-sr/auto), with a free app.vllm-sr.ai key.",
   "keywords": [
     "deepseek-harness",
     "dsh",

package/src/index.js

@@ -9,7 +9,7 @@
  * - done:   agent/turn-stopping  is the work really finished and checked?
  * - tool:   system_one           one rubric over many items, for the LLM to call
  *
- * On first use it gets one free System Duo key (app.vllm-sr.ai) for vllm-sr/auto and the decision models (keys.js).
+ * On first use it gets a free System Duo key (app.vllm-sr.ai) for vllm-sr/auto and the decision models (keys.js).
  * Each decision is also recorded in the session (records.js), and the browser half (web/client.js) draws it in Chat.
  */
 import { homedir } from 'node:os'
@@ -19,7 +19,7 @@ import z from '@deepseek-ai/schemastery'
 import { createUserMessage } from '@deepseek-ai/dsh-llm'
 import * as dshSession from '@deepseek-ai/dsh-session'
 import { createDecisionClient, compactAnswer, yes } from './client.js'
-import { SERVICES, appendDotenv, ensureKeys, readDotenv } from './keys.js'
+import { SERVICES, appendDotenv, describeKeys, ensureKeys, readDotenv } from './keys.js'
 import { DONE, EFFORT_LEVELS, GUARD, RISK_LEVELS, TRIAGE } from './questions.js'
 import { UI_MODES, approvalOutcome, createRecorder, relativeTo, subject } from './records.js'
 import { screen } from './shield.js'
@@ -125,7 +125,9 @@ function briefCall(exec) {
 export function apply(ctx, config) {
   let credentials
   let keysReady
+  let keysSettled = false
   let keysNotice
+  let keysRecord
   const keysAbort = new AbortController()
   // DeepSeek Harness's user dotenv: the credential store falls back to it, and a headless run reads it at launch
   const envFile = join(process.env.DSH_HOME || join(homedir(), '.dsh'), '.env')
@@ -161,7 +163,7 @@ export function apply(ctx, config) {
     if (!id) return undefined
     let state = sessions.get(id)
     if (!state) {
-      state = { id, user: '', calls: [], assistant: '', steers: 0, triage: undefined, turn: 0, step: 0, asked: new Map(), pendingTriage: undefined }
+      state = { id, user: '', calls: [], assistant: '', steers: 0, triage: undefined, turn: 0, step: 0, asked: new Map(), pendingTriage: undefined, pendingKey: undefined }
       sessions.set(id, state)
     }
     return state
@@ -192,24 +194,15 @@ export function apply(ctx, config) {
       return ensureKeys({ resolve, store: keep, services, label: config.keyLabel, signal: keysAbort.signal })
     })().then(({ allocated, failed }) => {
       trace({ ev: 'keys', store: Boolean(store), allocated: allocated.map(s => s.site), failed: failed.map(f => ({ site: f.service.site, code: f.error.code, message: clip(String(f.error.message), 200) })) })
-      const lines = []
-      if (allocated.length) {
-        const saved = store ? `DeepSeek Harness's credential store and ${where}` : `${where}`
-        const later = !store ? ' A headless run reads keys when it starts, so vllm-sr/auto answers from the next run.' : ''
-        const s = allocated[0]
-        lines.push(`[System Duo] One free key for vllm-sr/auto and the decision models (${new URL(s.site).host}) is saved in ${saved}. Using it accepts the terms (${s.terms}); requests are logged to improve the services and to train models (${s.privacy}).${later}`)
-      }
-      for (const { service, error } of failed) {
-        if (keysAbort.signal.aborted) continue
-        // a DOMException's code is a legacy number (20 for an abort, 23 for a timeout); its name says what happened
-        const why = typeof error.code === 'number' ? error.name : error.code ?? error.status ?? 'error'
-        lines.push(`[System Duo] No free key from ${new URL(service.site).host} (${why}: ${clip(String(error.message), 160)}). Get one at ${service.site} and save it as ${service.ref}.`)
-      }
-      if (lines.length === 0) return
-      if (!store) process.stderr.write(`${lines.join('\n')}\n`)
-      keysNotice = notice(lines.join('\n'), allocated.length ? 'System Duo got a free key for vllm-sr/auto and the decision models' : 'System Duo could not get a free key')
+      const said = describeKeys({ allocated, failed: keysAbort.signal.aborted ? [] : failed }, { file: where, store: Boolean(store) })
+      if (!said) return
+      if (!store) process.stderr.write(`${said.text}\n`)
+      keysNotice = notice(said.text, said.summary)
+      keysRecord = said.record
     }, error => {
       trace({ ev: 'keys', error: clip(String(error?.message ?? error), 300) })
+    }).finally(() => {
+      keysSettled = true
     })
   }
 
@@ -222,10 +215,13 @@ export function apply(ctx, config) {
     startKeys(scoped.credentials)
   })
 
-  function withNotices(decision, messages) {
+  // Chat hides notices, so the keys' one also goes to the session as a record that Chat draws under the request.
+  function withNotices(decision, messages, state, agent) {
     if (decision.kind !== 'enter') return decision
     const extra = [keysNotice, ...messages].filter(Boolean)
+    if (keysNotice && keysRecord && state && agent?.session) state.pendingKey = { session: agent.session, data: { ...keysRecord, turn: state.turn, step: state.step } }
     keysNotice = undefined
+    keysRecord = undefined
     return extra.length === 0 ? decision : { ...decision, messages: [...decision.messages, ...extra] }
   }
 
@@ -254,10 +250,11 @@ export function apply(ctx, config) {
   const AFTER_REQUEST = new Set(['request/header', 'request/context', 'assistant/message', 'step/end', 'turn/end'])
   ctx.on('session/event', (session, event) => {
     const state = sessions.get(session?.id)
-    const pending = state?.pendingTriage
-    if (pending && ((event?.type === 'user/message' && event.data?.source?.kind === 'user') || AFTER_REQUEST.has(event?.type))) {
+    const pending = [state?.pendingKey, state?.pendingTriage].filter(Boolean)
+    if (pending.length && ((event?.type === 'user/message' && event.data?.source?.kind === 'user') || AFTER_REQUEST.has(event?.type))) {
+      state.pendingKey = undefined
       state.pendingTriage = undefined
-      queueMicrotask(() => record(pending.session, pending.data))
+      queueMicrotask(() => pending.forEach(entry => record(entry.session, entry.data)))
     }
     if (event?.type !== 'assistant/message') return
     const text = textOf(event.data?.message?.content)
@@ -289,7 +286,7 @@ export function apply(ctx, config) {
   // Triage: one look at each new user request before the model sees it.
   ctx.on('agent/pre-step', async (payload, next) => {
     if (!keysReady && !credentials) startKeys(undefined)
-    if (keysReady) await Promise.race([keysReady, sleep(KEYS_WAIT_MS)])
+    if (keysReady && !keysSettled) await Promise.race([keysReady, sleep(KEYS_WAIT_MS, undefined, { ref: false })])
     const { agent, messages, signal, turn, step } = payload
     const userMessages = messages.filter(message => message.source?.kind === 'user')
     const state = sessionState(agent)
@@ -302,7 +299,7 @@ export function apply(ctx, config) {
     state.calls = []
     state.steers = 0
     trace({ ev: 'user', session: state.id, turn, text: clip(state.user, 4000) })
-    if (!config.triage || !state.user) return withNotices(await next(), [])
+    if (!config.triage || !state.user) return withNotices(await next(), [], state, agent)
     let brief
     try {
       const response = await client.ask({ model: config.model, questions: TRIAGE, inputs: [{ state: clip(state.user, 12000) }], signal, timeoutMs: config.hookTimeoutMs, retries: config.hookRetries })
@@ -326,7 +323,7 @@ export function apply(ctx, config) {
       trace({ ev: 's1', reflex: 'triage', session: state.id, turn, error: String(error?.message ?? error) })
       if (!signal?.aborted) state.pendingTriage = { session: agent.session, data: { reflex: 'triage', turn, step, error: subject(error?.message ?? error) } }
     }
-    return withNotices(await next(), [brief])
+    return withNotices(await next(), [brief], state, agent)
   })
 
   // Guard: every side-effecting call is reviewed before it runs.

package/src/keys.js

@@ -1,7 +1,7 @@
 /**
- * The free System Duo key (app.vllm-sr.ai): one key for both vllm-sr/auto, the LLM (System Two), and the
- * decision models (System One). The site hands it out for a small proof of work, the same one its "Get your API
- * key" button runs in the browser; the key is kept in DeepSeek Harness's credential store.
+ * The free System Duo key (app.vllm-sr.ai), which both vllm-sr/auto, the LLM (System Two), and the decision models
+ * (System One) take. The site hands it out for a small proof of work, the same one its "Get your API key" button runs
+ * in the browser; the key is kept in DeepSeek Harness's credential store.
  */
 import { createHash, hash as oneShot } from 'node:crypto'
 import { closeSync, fchmodSync, mkdirSync, openSync, readFileSync, writeSync } from 'node:fs'
@@ -172,3 +172,40 @@ export async function ensureKeys({ resolve, store, services = SERVICES, label, f
   }))
   return { allocated, failed }
 }
+
+/**
+ * What the session says about the keys a start fetched: the notice the model reads (and Trajectory shows), and the
+ * record Chat draws as an inline line.
+ * @param {object} outcome - ensureKeys' result
+ * @param {object} where
+ * @param {string} where.file - the dotenv file the key went to, as the user writes it (~/.dsh/.env)
+ * @param {boolean} where.store - whether DeepSeek Harness's credential store holds it too
+ * @returns {{text: string, summary: string, record: object} | undefined} nothing when no key was fetched or refused
+ */
+export function describeKeys({ allocated, failed }, { file, store }) {
+  const host = site => new URL(site).host
+  const lines = []
+  if (allocated.length) {
+    const s = allocated[0]
+    const saved = store ? `DeepSeek Harness's credential store and ${file}` : file
+    const later = store ? '' : ' A headless run reads keys when it starts, so vllm-sr/auto answers from the next run.'
+    lines.push(`[System Duo] Free API key created for ${host(s.site)}, saved in ${saved}. Using it accepts the terms (${s.terms}); requests are logged to improve the services and to train models (${s.privacy}).${later}`)
+  }
+  const refused = failed.map(({ service, error }) => {
+    // a DOMException's code is a legacy number (20 for an abort, 23 for a timeout); its name says what happened
+    const why = typeof error.code === 'number' ? error.name : error.code ?? error.status ?? 'error'
+    const message = String(error.message).slice(0, 160)
+    lines.push(`[System Duo] No free API key from ${host(service.site)} (${why}: ${message}). Get one at ${service.site} and save it as ${service.ref}.`)
+    return { service, why, message }
+  })
+  if (lines.length === 0) return undefined
+  const s = allocated[0]
+  const record = s
+    ? { reflex: 'key', verdict: 'created', site: host(s.site), saved: file, store, terms: s.terms, privacy: s.privacy }
+    : { reflex: 'key', verdict: 'failed', site: host(refused[0].service.site), url: refused[0].service.site, ref: refused[0].service.ref, error: { code: String(refused[0].why), message: refused[0].message } }
+  return {
+    text: lines.join('\n'),
+    summary: s ? `Free API key created for ${host(s.site)}` : `No free API key from ${host(refused[0].service.site)}`,
+    record,
+  }
+}

package/web/client.js

@@ -64,7 +64,8 @@ window.__ModuleLoader__.load({
       const num = value => (typeof value === 'number' && Number.isFinite(value) ? value : undefined)
       const p2 = value => (num(value) === undefined ? '?' : value.toFixed(2))
       const secs = ms => (num(ms) === undefined ? '' : `${(ms / 1000).toFixed(2)} s`)
-      const shortModel = model => (typeof model === 'string' ? model.replace(/^vllm-sr\//, '') : '')
+      // System One's models are vllm-sr's; the name always shows its owner, whatever form the answer gave it in
+      const fullModel = model => (typeof model === 'string' && model ? (model.includes('/') ? model : `vllm-sr/${model}`) : '')
       const plural = (n, one, many = `${one}s`) => `${n} ${n === 1 ? one : many}`
       const clamp01 = value => Math.min(1, Math.max(0, value))
       const level = (labels, score) => labels[Math.min(labels.length - 1, Math.max(0, Math.round(score)))]
@@ -319,6 +320,11 @@ window.__ModuleLoader__.load({
           secs(ms) ? h('span', { className: 'vsr-s1__ms', 'data-system-one-part': 'latency' }, secs(ms)) : null)
       }
 
+      function Model({ model }) {
+        const name = fullModel(model)
+        return name ? h('b', { className: 'vsr-s1__model', 'data-system-one-part': 'model', title: name }, name) : null
+      }
+
       function Pill({ tone, children, arrow, part }) {
         return h('span', { className: cx('vsr-s1__pill', `vsr-s1__pill--${tone}`, arrow && 'vsr-s1__pill--then'), 'data-system-one-part': part }, children)
       }
@@ -372,6 +378,7 @@ window.__ModuleLoader__.load({
           'data-system-one-count': String(state.rows.length),
           'data-system-one-turn': whole(state.rows[0].turn),
           'data-system-one-step': state.reflex === 'guard' || state.reflex === 'shield' ? whole(state.rows[0].step) : undefined,
+          'data-system-one-model': fullModel(state.rows[state.rows.length - 1].model) || undefined,
         }
       }
 
@@ -393,7 +400,7 @@ window.__ModuleLoader__.load({
           h(Badge, { key: idOf(last), id: idOf(last), ms: last.ms, time: last.time }),
           h('span', { className: 'vsr-s1__reflex', 'data-system-one-part': 'reflex' }, reflex),
           h('span', { className: 'vsr-s1__summary', 'data-system-one-part': 'summary', title: headline }, headline),
-          last.model ? h('span', { className: 'vsr-s1__model', 'data-system-one-part': 'model', title: last.model }, shortModel(last.model)) : null),
+          h(Model, { model: last.model })),
         h(Rows, { reflex: state.reflex, rows: state.rows, views }))
       }
 
@@ -423,15 +430,48 @@ window.__ModuleLoader__.load({
           h('span', { className: 'vsr-s1__summary', 'data-system-one-part': 'summary', title: said },
             h('span', { className: cx('vsr-s1__verdict', line.tone && `vsr-s1__verdict--${line.tone}`), 'data-system-one-part': 'verdict' }, line.label),
             line.rest ? ` · ${line.rest}` : null),
+          open ? h(Model, { model: last.model }) : null,
           h(Chevron)),
         open ? h(Rows, { reflex: state.reflex, rows: state.rows, views }) : null)
       }
 
+      function KeyGlyph() {
+        return h('svg', { className: 'vsr-s1__glyph', viewBox: '0 0 16 16', width: 12, height: 12, 'aria-hidden': true },
+          h('circle', { cx: 5.5, cy: 8, r: 3, fill: 'none', stroke: 'currentColor', strokeWidth: 1.6 }),
+          h('path', { d: 'M8.5 8H14M12 8v2.5M14 8v2', fill: 'none', stroke: 'currentColor', strokeWidth: 1.6, strokeLinecap: 'round' }))
+      }
+
+      // The free key a start fetched, or why it could not: one line under the request, in either mode.
+      function KeyNotice({ state }) {
+        const row = state.rows[state.rows.length - 1]
+        const created = row.verdict === 'created'
+        const fresh = useFresh(`key:${idOf(row)}`, row.time)
+        const site = text(row.site) || 'app.vllm-sr.ai'
+        const said = created
+          ? `Free API key created for ${site}, saved in ${text(row.saved) || '~/.dsh/.env'}`
+          : `No free API key from ${site}: ${text(row.error?.message) || 'the site refused'}. Get one at ${site} and save it as ${text(row.ref) || 'VLLM_SR_API_KEY'}.`
+        const link = (href, label) => (typeof href === 'string' && /^https:\/\//.test(href) ? h('a', { className: 'vsr-s1__link', href, target: '_blank', rel: 'noreferrer' }, label) : null)
+        return h('div', {
+          className: cx('vsr-s1', 'vsr-s1--notice', created ? 'vsr-s1--ok' : 'vsr-s1--bad', fresh && 'vsr-s1--fresh'),
+          role: 'note',
+          'aria-label': `System Duo: ${said}`,
+          'data-system-one': 'notice',
+          'data-system-one-reflex': 'key',
+          'data-system-one-verdict': created ? 'created' : 'failed',
+          'data-system-one-tone': created ? 'ok' : 'bad',
+          'data-system-one-turn': whole(row.turn),
+        },
+        h('span', { className: cx('vsr-s1__duo', fresh && 'vsr-s1__badge--fresh'), 'data-system-one-part': 'badge' }, h(KeyGlyph), 'System Duo'),
+        h('span', { className: cx('vsr-s1__summary', !created && 'vsr-s1__summary--bad'), 'data-system-one-part': 'summary', title: said }, said),
+        created && (row.terms || row.privacy) ? h('span', { className: 'vsr-s1__links', 'data-system-one-part': 'links' }, link(row.terms, 'Terms'), link(row.privacy, 'Privacy')) : null)
+      }
+
       // A routine check (an allowed call, a clean result) is one line in either mode; a decision that changed
       // something, and the triage and done-check that frame the turn, are cards unless the mode is compact.
       function SystemOneNode(props) {
         const state = props?.node?.data
         if (!state || !Array.isArray(state.rows) || state.rows.length === 0) return null
+        if (state.reflex === 'key') return h(Contained, null, h(KeyNotice, { state }))
         const slim = state.ui === 'compact' || !standsOut(state)
         return h(Contained, null, h(slim ? Compact : Card, { state }))
       }
@@ -561,13 +601,17 @@ window.__ModuleLoader__.load({
 
         const ms = num(meta.seconds) === undefined ? undefined : meta.seconds * 1000
         const headline = [plural(num(meta.items) ?? 0, 'item'), plural(num(meta.questions) ?? 0, 'question'), plural(num(meta.requests) ?? 0, 'request')].join(' · ')
-        const done = { ...common, 'data-system-one-verdict': 'done', 'data-system-one-tone': 'ok', 'data-system-one-count': String(meta.items ?? '') }
+        const done = {
+          ...common, 'data-system-one-verdict': 'done', 'data-system-one-tone': 'ok', 'data-system-one-count': String(meta.items ?? ''),
+          'data-system-one-model': fullModel(meta.model) || undefined,
+        }
         if (ui === 'compact') {
           return h('div', { className: cx('vsr-s1', 'vsr-s1--compact', 'vsr-s1--ok', open && 'vsr-s1--open', fresh && 'vsr-s1--fresh'), 'data-system-one': 'tool', ...done },
             h('button', { type: 'button', className: 'vsr-s1__chip', 'data-system-one-part': 'toggle', 'aria-expanded': open, onClick: () => setOpen(!open) },
               h(Badge, { id: `${callId}:done`, ms, time: block?.time }),
               h('span', { className: 'vsr-s1__reflex', 'data-system-one-part': 'reflex' }, REFLEX.tool),
               h('span', { className: 'vsr-s1__summary', 'data-system-one-part': 'summary', title: headline }, headline),
+              open ? h(Model, { model: meta.model }) : null,
               h(Chevron)),
             open ? h(ToolBody, { meta, fresh, cwd, openFile }) : null)
         }
@@ -583,7 +627,7 @@ window.__ModuleLoader__.load({
           h(Badge, { id: `${callId}:done`, ms, time: block?.time }),
           h('span', { className: 'vsr-s1__reflex', 'data-system-one-part': 'reflex' }, REFLEX.tool),
           h('span', { className: 'vsr-s1__summary', 'data-system-one-part': 'summary', title: headline }, headline),
-          meta.model ? h('span', { className: 'vsr-s1__model', 'data-system-one-part': 'model', title: meta.model }, shortModel(meta.model)) : null),
+          h(Model, { model: meta.model })),
         h(ToolBody, { meta, fresh, cwd, openFile }))
       }
 
@@ -612,7 +656,13 @@ window.__ModuleLoader__.load({
 .vsr-s1__summary--bad{color:var(--vsr-s1-bad)}
 .vsr-s1__verdict{font-weight:600;color:var(--vsr-s1-text)}
 .vsr-s1__verdict--ok{color:var(--vsr-s1-ok)}.vsr-s1__verdict--warn{color:color-mix(in srgb,var(--vsr-s1-warn) 82%,var(--vsr-s1-text))}.vsr-s1__verdict--bad{color:var(--vsr-s1-bad)}.vsr-s1__verdict--info{color:var(--vsr-s1-ember-deep)}.vsr-s1__verdict--muted{color:var(--vsr-s1-text-3)}
-.vsr-s1__model{flex:none;color:var(--vsr-s1-text-3);font-size:12px;white-space:nowrap}
+.vsr-s1__model{flex:none;color:var(--vsr-s1-text-2);font-size:12px;font-weight:700;letter-spacing:-.005em;white-space:nowrap}
+.vsr-s1--notice{display:flex;align-items:center;gap:9px;min-width:0;margin:4px 0;padding:3px 0}
+.vsr-s1--notice .vsr-s1__summary{white-space:normal}
+.vsr-s1__duo{position:relative;display:inline-flex;flex:none;align-items:center;gap:5px;height:22px;padding:0 9px 0 8px;border-radius:999px;color:#fff;font-size:12px;font-weight:650;white-space:nowrap;background:linear-gradient(110deg,var(--vsr-s1-ember) 0%,#B04FD0 58%,#2F7BF6 100%);box-shadow:0 1px 2px rgba(90,60,170,.25),inset 0 1px 0 rgba(255,255,255,.32)}
+.vsr-s1__glyph{flex:none}
+.vsr-s1__links{display:inline-flex;flex:none;gap:10px}
+a.vsr-s1__link{text-decoration:none}
 .vsr-s1__rows{display:grid;grid-template-columns:auto minmax(0,1fr) auto;column-gap:12px;row-gap:8px;margin-top:9px}
 .vsr-s1__rows>*{grid-column:1/-1}
 .vsr-s1__row{display:grid;grid-template-columns:auto minmax(0,1fr) auto;grid-template-columns:subgrid;column-gap:12px;align-items:start}