host: cut busy-ack, tool lines and the untrusted fence name on code point boundaries

changelog.d/1930-host-line-bounds.fixed.md

@@ -0,0 +1 @@
+The host's own bounded lines no longer end in half of an emoji ([#1930](https://github.com/ThinkFlowLab/FlyBot/issues/1930)): the busy acknowledgment's 600-unit task text and 1,500-unit message text, `cleanName`'s 60-unit peer name (the one that lands in `fenceUntrusted`'s `from="…"` attribute), the `list_bots` group line's 100-unit notice and 120-unit job line, the group answer echo at 300 and a rework card's note at 200 now cut on a code point boundary, so a value whose cap falls inside a surrogate pair keeps the whole character instead of U+FFFD. Every cap is unchanged.

core/src/host/busy-ack.js

@@ -14,6 +14,7 @@
 // otherwise the chat's own model is used.
 import { BlockAssembler } from '@deepseek-ai/dsh-llm'
 import { splitRef } from './text.js'
+import { cut } from '../shared/text-boundary.js'
 
 export const COALESCE_MS = 3_000
 export const ACK_TIMEOUT_MS = 3_000
@@ -56,11 +57,11 @@ export const ACK_SYSTEM = [
 /** The user prompt for the ack call: the new message and what the turn is doing. */
 export function ackPrompt({ message, tool, working }) {
   const doing = typeof tool === 'string' && tool.trim() !== '' && tool !== 'ask_user' ? `running the tool \`${tool}\`` : 'working on the previous request'
-  const task = String(working ?? '').trim().slice(0, 600)
+  const task = cut(String(working ?? '').trim(), 600).text
   return [
     `Currently: ${doing}.`,
     ...(task === '' ? [] : [`The request being worked on: ${task}`]),
-    `New message: ${String(message ?? '').trim().slice(0, 1_500)}`,
+    `New message: ${cut(String(message ?? '').trim(), 1_500).text}`,
   ].join('\n')
 }
 

core/src/host/delivery.js

@@ -3,6 +3,7 @@ import { randomUUID } from 'node:crypto'
 import { createUserMessage } from '@deepseek-ai/dsh-llm'
 import { SOURCE_KIND } from './constants.js'
 import { OWNER, OTHER, SCHEDULED } from './turn-origin.js'
+import { cut } from '../shared/text-boundary.js'
 
 // Text from outside the team, marked so the Bot treats it as data. Marker fragments are
 // removed until none is left, so a nested "</untrusted-</untrusted-message>message>"
@@ -35,7 +36,7 @@ export const unwrapUntrusted = (text) => {
 // A name from outside the team, safe on one line of trusted text: no control characters,
 // newlines, brackets or quotes, at most 60 characters.
 export const cleanName = (name, fallback = 'a remote Bot') => {
-  const text = String(name ?? '').replace(/[\u0000-\u001f\u007f-\u009f\u2028\u2029]/g, ' ').replace(/[[\]"<>`]/g, '').replace(/\s+/g, ' ').trim().slice(0, 60)
+  const text = cut(String(name ?? '').replace(/[\u0000-\u001f\u007f-\u009f\u2028\u2029]/g, ' ').replace(/[[\]"<>`]/g, '').replace(/\s+/g, ' ').trim(), 60).text
   return text || fallback
 }
 export const fenceUntrusted = (name, text) => `<untrusted-message from="${cleanName(name)}">\n${stripUntrusted(text ?? '')}\n</untrusted-message>`

core/src/host/tools.js

@@ -86,7 +86,7 @@ export function install(rt) {
     const parts = [
       `members: ${[...members, ...remotes].join(', ')}`,
       room.mode === 'admin' && room.admin ? 'plain messages go to the admin first' : 'every member replies in turn',
-      room.notice ? `notice: ${room.notice.split('\n')[0].slice(0, 100)}` : '',
+      room.notice ? `notice: ${cut(room.notice.split('\n')[0], 100).text}` : '',
       room.workDir ? `members work in ${room.workDir} here` : '',
     ].filter(Boolean)
     return `- "${room.name}" — ${parts.join('; ')}`
@@ -174,7 +174,7 @@ export function install(rt) {
       const lines = Object.values(state.bots).filter(listed).map(bot => {
         const model = bot.appliedModel ?? bot.model
         const tags = [bot.role, model ? `model ${model}` : '', heldSends[bot.id] ? `${heldSends[bot.id]} to approve` : '', !foreign(bot) && bot.workDir ? `works in ${bot.workDir}` : '', rt.isMain(bot.id) ? 'Main Bot' : '', bot.id === self ? 'you' : ''].filter(Boolean).join(', ')
-        const job = !foreign(bot) && bot.instructions ? ` — ${bot.instructions.split('\n')[0].slice(0, 120)}` : ''
+        const job = !foreign(bot) && bot.instructions ? ` — ${cut(bot.instructions.split('\n')[0], 120).text}` : ''
         return `- ${bot.name}${tags ? ` (${tags})` : ''}${job}`
       })
       if (lines.length === 0) return 'No Bots yet.'
@@ -740,7 +740,7 @@ export function install(rt) {
     await rt.save()
     const who = cleanName(actor?.displayName, 'You')
     const given = JSON.parse(reply.text).answers[0]
-    const values = [...given.selected, ...(given.custom ? [given.custom] : [])].join(', ').replace(/\s+/g, ' ').slice(0, 300)
+    const values = cut([...given.selected, ...(given.custom ? [given.custom] : [])].join(', ').replace(/\s+/g, ' '), 300).text
     void rt.postToRoom?.(room, { id: 'face-notice', name: 'Room' }, `${who} answered ${bot.name}: ${question.question} → ${values}`, { quiet: true, notice: true })
       .catch(error => rt.warn('ds-bot: the answer notice in %s was not shown: %s', room.name, error))
     return { ok: true }
@@ -1028,7 +1028,7 @@ ${task.writeBack.body}`)
         }
         if (args?.rework !== undefined) {
           const child = await rt.trackBotTasks(self, args ?? {}, { sessionId: exec.agent?.id, roomId })
-          return `Rework ${child.id} on ${child.reworkOf} for the ${child.handoff?.to ?? 'producer'} pool: ${String(child.note ?? '').slice(0, 200)}\nThe producer takes it from its own pool. Set that card's column to in-progress when you pick it up. Do not post a status in the group.`
+          return `Rework ${child.id} on ${child.reworkOf} for the ${child.handoff?.to ?? 'producer'} pool: ${cut(String(child.note ?? ''), 200).text}\nThe producer takes it from its own pool. Set that card's column to in-progress when you pick it up. Do not post a status in the group.`
         }
         if (args?.release !== undefined) {
           const returned = await rt.trackBotTasks(self, args ?? {}, { sessionId: exec.agent?.id, roomId })

core/test/busy-ack.test.js

@@ -284,3 +284,20 @@ test('the own-chat ack is unchanged beside the room path (#1646)', async (t) =>
   await settleAck(host.rt, 'gb')
   assert.deepEqual(Object.keys(host.rt.busyAcksView()).sort(), ['b', 'gb'])
 })
+
+test('the prompt bounds keep whole characters at 600 and 1,500 (#P282b)', () => {
+  const sourceTask = 'w'.repeat(599) + '😀tail'
+  const sourceMessage = 'a'.repeat(1_499) + '😀tail'
+  const prompt = ackPrompt({ message: sourceMessage, tool: 'bash', working: sourceTask })
+  assert.ok(prompt.isWellFormed(), `the prompt is not well formed: ${JSON.stringify(prompt.slice(-8))}`)
+  const lines = prompt.split('\n')
+  const keptTask = lines.find(line => line.startsWith('The request being worked on: ')).slice('The request being worked on: '.length)
+  const keptMessage = lines[lines.length - 1].slice('New message: '.length)
+  assert.equal(keptTask, 'w'.repeat(599))
+  assert.equal(keptMessage, 'a'.repeat(1_499))
+  assert.ok(keptTask.length <= 600)
+  assert.ok(keptMessage.length <= 1_500)
+  const bytesOf = value => Buffer.from(value, 'utf8')
+  assert.ok(bytesOf(sourceTask).subarray(0, bytesOf(keptTask).length).equals(bytesOf(keptTask)))
+  assert.ok(bytesOf(sourceMessage).subarray(0, bytesOf(keptMessage).length).equals(bytesOf(keptMessage)))
+})

core/test/fence.test.js

@@ -37,3 +37,14 @@ test('#130: previews unwrap the stored untrusted-message wrap', () => {
   assert.equal(unwrapUntrusted(fenceUntrusted('User', 'hello from the owner')), 'hello from the owner')
   assert.equal(unwrapUntrusted('plain text'), 'plain text')
 })
+
+test('a peer name at the 60 bound keeps the whole character (#P282b)', () => {
+  const source = 'a'.repeat(59) + '😀tail'
+  const name = cleanName(source)
+  assert.equal(name, 'a'.repeat(59))
+  assert.ok(name.isWellFormed(), `the name is not well formed: ${JSON.stringify(name.slice(-4))}`)
+  assert.ok(name.length <= 60)
+  const fenced = fenceUntrusted(source, 'hello')
+  assert.ok(fenced.isWellFormed(), 'the fence header is not well formed')
+  assert.ok(fenced.startsWith(`<untrusted-message from="${'a'.repeat(59)}">`))
+})

core/test/flybot.test.js

@@ -882,6 +882,37 @@ test('list_bots keeps whole characters in a peer display name and owner name (#1
   assert.ok(Buffer.from(keptOwner, 'utf8').equals(Buffer.from(ownerName, 'utf8').subarray(0, Buffer.byteLength(keptOwner))), 'the owner name is a prefix of what the peer published, byte for byte')
 })
 
+// #P282b: the group line's notice is bounded at 100, and a cut inside an emoji left half a pair.
+test('list_bots keeps whole characters in a group notice at the 100 bound (#P282b)', async (t) => {
+  const notice = 'a'.repeat(99) + '😀tail'
+  const { tool, stop } = await start({
+    rooms: { g: { id: 'g', name: 'Group', members: ['a'], admin: 'a', mode: 'admin', notice, createdAt: 1, createdBy: null } },
+  })
+  t.after(stop)
+  const listed = await tool('list_bots', {}, 'a')
+  assert.equal(listed.isWellFormed(), true, `the listing is not well formed: ${JSON.stringify(listed.slice(0, 80))}`)
+  const line = listed.split('\n').find(l => l.includes('notice: '))
+  assert.ok(line, `the notice is listed: ${JSON.stringify(listed)}`)
+  const kept = line.split('notice: ')[1].split('; ')[0]
+  assert.equal(kept, 'a'.repeat(99), `the notice keeps the 100 bound: ${JSON.stringify(kept.slice(-4))}`)
+  assert.ok(Buffer.from(kept, 'utf8').equals(Buffer.from(notice, 'utf8').subarray(0, Buffer.byteLength(kept))), 'the notice is a prefix, byte for byte')
+})
+
+// #P282b: a Bot's job line is bounded at 120, and a cut inside an emoji left half a pair.
+test('list_bots keeps whole characters in a Bot job line at the 120 bound (#P282b)', async (t) => {
+  const { tool, stop } = await start()
+  t.after(stop)
+  const instructions = 'a'.repeat(119) + '😀tail'
+  assert.match(await tool('update_bot', { bot: 'b', instructions }, 'a'), /Updated Writer/)
+  const listed = await tool('list_bots', {}, 'a')
+  assert.equal(listed.isWellFormed(), true, `the listing is not well formed: ${JSON.stringify(listed.slice(0, 80))}`)
+  const line = listed.split('\n').find(l => l.includes(' — '))
+  assert.ok(line, `the job line is listed: ${JSON.stringify(listed)}`)
+  const kept = line.split(' — ')[1]
+  assert.equal(kept, 'a'.repeat(119), `the job line keeps the 120 bound: ${JSON.stringify(kept.slice(-4))}`)
+  assert.ok(Buffer.from(kept, 'utf8').equals(Buffer.from(instructions, 'utf8').subarray(0, Buffer.byteLength(kept))), 'the job line is a prefix, byte for byte')
+})
+
 test('cleanRemote sanitizes peer name, homeName and owner at the record boundary (#75)', async (t) => {
   const { ctx, fb, tool, received, stop, grantTo } = await start()
   t.after(stop)

core/test/group-work.test.js

@@ -370,6 +370,27 @@ test('the answer given in the group reaches the asking Bot\'s waiting work and i
   assert.equal((await host.call('answer-question', { sessionId: 'r1', questionId: asked.id, answer: { selected: ['dev'] } })).error.code, 'bot/stale')
 })
 
+// #P282b: the answer echo is bounded at 300, and a cut inside an emoji left half a pair in the
+// room line every member reads.
+test('the answer echo in a group keeps whole characters at the 300 bound (#P282b)', async (t) => {
+  const host = await start(t)
+  await host.tool('message_bot', { to: 'Tester', message: '把测试跑一遍' }, 'gb')
+  host.begin('c')
+  await askBranch(host)
+  host.finish('c', 'Need to know the branch first.')
+  await settle(() => host.agentOf('gb').calls.length === 1)
+  const asked = (await host.call('state', {})).value.questions.r1
+  const custom = 'a'.repeat(299) + '😀tail'
+  assert.equal((await host.call('answer-question', { sessionId: 'r1', questionId: asked.id, answer: { selected: [], custom } })).ok, true)
+  await settle(() => host.agentOf('r1').calls.some(call => call.source?.notice === true))
+  const notice = host.agentOf('r1').calls.find(call => call.source?.notice === true)
+  const said = text(notice)
+  assert.equal(said.isWellFormed(), true, `the answer echo is not well formed: ${JSON.stringify(said.slice(-8))}`)
+  const values = said.split(' → ').at(-1)
+  assert.equal(values, 'a'.repeat(299), `the answer echo keeps the 300 bound: ${JSON.stringify(values.slice(-4))}`)
+  assert.ok(Buffer.from(values, 'utf8').equals(Buffer.from(custom, 'utf8').subarray(0, Buffer.byteLength(values))), 'the answer echo is a prefix of the answer, byte for byte')
+})
+
 test('a question in the Bot\'s own chat stays there, and a request answered to the user keeps its card there (#550)', async (t) => {
   const host = await start(t)
   host.emit('c', 'turn/start', { turn: 1 })

core/test/task-handoff-route.test.js

@@ -178,6 +178,24 @@ test('a role returns the card to its producer as a rework card the producer can
   assert.equal(app.rt.state.tasks[childId].claimedBy, 'd1')
 })
 
+// #P282b: the rework line carries the note at 200, and a cut inside an emoji left half a pair.
+test('the rework line keeps whole characters in the note at the 200 bound (#P282b)', async (t) => {
+  const app = await start(t)
+  const id = app.idOf(await app.track('a', { tasks: [{ title: 'Build the route', handoff: { from: 'Lead', to: 'Dev' } }] }))
+  await app.drive()
+  await app.track('d1', { claim: true, id })
+  await app.track('d1', { tasks: [{ id, handoff: { from: 'Dev', to: 'Reviewer', reason: 'diff ready' } }] })
+  await app.drive()
+  await app.track('r1', { claim: true, id })
+  await app.track('r1', { tasks: [{ id, column: 'in-progress' }] })
+  const reason = 'a'.repeat(199) + '😀tail'
+  const returned = await app.track('r1', { rework: { task: id, reason } })
+  assert.equal(returned.isWellFormed(), true, `the rework line is not well formed: ${JSON.stringify(returned.slice(-8))}`)
+  const shown = returned.split(' pool: ')[1].split('\n')[0]
+  assert.equal(shown, 'a'.repeat(199), `the rework line keeps the 200 bound: ${JSON.stringify(shown.slice(-4))}`)
+  assert.ok(Buffer.from(shown, 'utf8').equals(Buffer.from(reason, 'utf8').subarray(0, Buffer.byteLength(shown))), 'the note is a prefix of the reason, byte for byte')
+})
+
 test('a card waiting for its rework is not re-offered when its lease expires (#1804)', async (t) => {
   const app = await start(t)
   const id = app.idOf(await app.track('a', { tasks: [{ title: 'Build the route', handoff: { from: 'Lead', to: 'Dev' } }] }))