feat: Add transcript engine API and connect ConversationHistory to real data

- Add transcript-engine service that parses Claude Code JSONL transcripts
  with session listing, message extraction, token/stats analysis, and caching
- Add transcript REST routes (sessions list, latest, by session ID, section filtering)
- Rewrite ConversationHistory to fetch from /api/transcript/* instead of mock data
- Add session pills for switching between conversation sessions
- Add stats bar footer with model, duration, tokens, and tool count
- Add TranscriptSession/TranscriptMessage types, ChatInput, InputSettings,
  PromptBar updates, TranscriptCard, and useVoiceCapture composable
This commit is contained in:
2026-02-15 20:05:27 -06:00
parent 68edc01d44
commit f3ac7986ec
10 changed files with 2246 additions and 97 deletions

View File

@@ -5,6 +5,7 @@ const props = defineProps<{
placeholder?: string
recording?: boolean
historyActive?: boolean
settingsActive?: boolean
autofocus?: boolean
}>()
@@ -12,6 +13,7 @@ const emit = defineEmits<{
submit: [text: string]
mic: []
'toggle-history': []
'toggle-settings': []
}>()
const inputText = ref('')
@@ -82,6 +84,17 @@ defineExpose({ focus })
<polyline points="12 6 12 12 16 14"/>
</svg>
</button>
<button
class="ci-btn ci-settings"
:class="{ active: settingsActive }"
title="Voice settings"
@click="emit('toggle-settings')"
>
<svg xmlns="http://www.w3.org/2000/svg" width="16" height="16" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round">
<circle cx="12" cy="12" r="3"/>
<path d="M19.4 15a1.65 1.65 0 0 0 .33 1.82l.06.06a2 2 0 0 1-2.83 2.83l-.06-.06a1.65 1.65 0 0 0-1.82-.33 1.65 1.65 0 0 0-1 1.51V21a2 2 0 0 1-4 0v-.09A1.65 1.65 0 0 0 9 19.4a1.65 1.65 0 0 0-1.82.33l-.06.06a2 2 0 0 1-2.83-2.83l.06-.06A1.65 1.65 0 0 0 4.68 15a1.65 1.65 0 0 0-1.51-1H3a2 2 0 0 1 0-4h.09A1.65 1.65 0 0 0 4.6 9a1.65 1.65 0 0 0-.33-1.82l-.06-.06a2 2 0 0 1 2.83-2.83l.06.06A1.65 1.65 0 0 0 9 4.68a1.65 1.65 0 0 0 1-1.51V3a2 2 0 0 1 4 0v.09a1.65 1.65 0 0 0 1 1.51 1.65 1.65 0 0 0 1.82-.33l.06-.06a2 2 0 0 1 2.83 2.83l-.06.06A1.65 1.65 0 0 0 19.4 9a1.65 1.65 0 0 0 1.51 1H21a2 2 0 0 1 0 4h-.09a1.65 1.65 0 0 0-1.51 1z"/>
</svg>
</button>
</div>
</template>
@@ -148,7 +161,8 @@ defineExpose({ focus })
color: rgba(139, 92, 246, 0.9);
}
.ci-history.active {
.ci-history.active,
.ci-settings.active {
background: rgba(255, 255, 255, 0.12);
color: rgba(255, 255, 255, 0.7);
}

View File

@@ -1,85 +1,205 @@
<script setup lang="ts">
import type { Agent, ConversationEntry } from '../../types/agent'
import { ref, watch, onMounted } from 'vue'
import type { Agent, TranscriptSession, TranscriptMessage } from '../../types/agent'
defineProps<{
agent: Agent
}>()
const mockEntries: ConversationEntry[] = [
{
id: '1',
role: 'user',
content: 'Revisa el módulo de autenticación y encuentra los errores de validación de tokens.',
timestamp: '14:32',
method: 'voice'
},
{
id: '2',
role: 'agent',
content: 'Encontré 3 issues en el validador de tokens JWT: expiración no verificada en refresh, falta sanitización del header Authorization, y el middleware no maneja tokens revocados.',
timestamp: '14:33',
method: 'text'
},
{
id: '3',
role: 'user',
content: 'Corrige los tres problemas y agrega tests unitarios para cada caso.',
timestamp: '14:35',
method: 'text'
},
{
id: '4',
role: 'agent',
content: 'Corregidos los 3 issues. Se agregaron 8 tests unitarios cubriendo cada caso: token expirado, header malformado, token revocado, y sus variantes edge-case. Todos los tests pasan.',
timestamp: '14:38',
method: 'text'
},
{
id: '5',
role: 'user',
content: 'Ahora implementa un sistema de rate limiting para la API de login con un máximo de 5 intentos por minuto.',
timestamp: '15:01',
method: 'voice'
},
{
id: '6',
role: 'agent',
content: 'Rate limiter implementado usando sliding window en Redis. Configurado a 5 intentos/minuto por IP. Retorna HTTP 429 con header Retry-After cuando se excede el límite.',
timestamp: '15:04',
method: 'text'
const sessions = ref<TranscriptSession[]>([])
const activeSessionId = ref<string | null>(null)
const messages = ref<TranscriptMessage[]>([])
const stats = ref<{
model: string
duration: number
toolCallCount: number
totalInput: number
totalOutput: number
} | null>(null)
const loading = ref(false)
const error = ref<string | null>(null)
// ── Helpers ──
function formatSessionTime(iso: string): string {
if (!iso) return '--:--'
const d = new Date(iso)
return d.toLocaleTimeString([], { hour: '2-digit', minute: '2-digit', hour12: false })
}
function formatDuration(ms: number): string {
if (!ms || ms <= 0) return '0s'
const totalSec = Math.floor(ms / 1000)
const m = Math.floor(totalSec / 60)
const s = totalSec % 60
if (m === 0) return `${s}s`
return `${m}m ${s}s`
}
function formatTokens(n: number): string {
if (!n || n <= 0) return '0'
if (n >= 1_000_000) return `${(n / 1_000_000).toFixed(1)}M`
if (n >= 1_000) return `${(n / 1_000).toFixed(1)}k`
return String(n)
}
function shortModel(model: string): string {
if (!model) return '?'
if (model.includes('opus')) return 'opus-4'
if (model.includes('sonnet')) return 'sonnet-4'
if (model.includes('haiku')) return 'haiku-4'
// Fallback: last segment
const parts = model.split('-')
return parts.slice(-2).join('-')
}
function truncateContent(text: string, max = 200): string {
if (!text || text.length <= max) return text || ''
return text.slice(0, max) + '...'
}
// ── Fetch helpers ──
function getBaseUrl(sessionId: string): string {
const firstSession = sessions.value[0]
if (firstSession && sessionId === firstSession.id) {
return '/api/transcript/latest'
}
]
return `/api/transcript/${sessionId}`
}
async function fetchSessions() {
try {
const res = await fetch('/api/transcript/sessions')
if (!res.ok) throw new Error(`HTTP ${res.status}`)
sessions.value = await res.json()
} catch (e: any) {
error.value = `Error cargando sesiones: ${e.message}`
}
}
async function fetchSessionData(sessionId: string) {
loading.value = true
error.value = null
try {
const base = getBaseUrl(sessionId)
const [msgRes, statsRes, tokensRes] = await Promise.all([
fetch(`${base}?section=messages`),
fetch(`${base}?section=stats`),
fetch(`${base}?section=tokens`)
])
if (!msgRes.ok || !statsRes.ok || !tokensRes.ok) {
throw new Error('Error fetching session data')
}
const [msgData, statsData, tokensData] = await Promise.all([
msgRes.json(),
statsRes.json(),
tokensRes.json()
])
messages.value = msgData.messages || []
stats.value = {
model: statsData.model || '',
duration: statsData.duration || 0,
toolCallCount: statsData.stats?.toolCallCount || 0,
totalInput: tokensData.tokens?.totalInput || 0,
totalOutput: tokensData.tokens?.totalOutput || 0
}
} catch (e: any) {
error.value = `Error cargando datos: ${e.message}`
messages.value = []
stats.value = null
} finally {
loading.value = false
}
}
// ── Lifecycle ──
watch(activeSessionId, (id) => {
if (id) fetchSessionData(id)
})
onMounted(async () => {
await fetchSessions()
if (sessions.value.length > 0) {
activeSessionId.value = sessions.value[0].id
}
})
</script>
<template>
<div class="conversation-history">
<!-- Header -->
<div class="history-header">
<span class="history-title">Historial</span>
<span class="history-count">{{ mockEntries.length }}</span>
<span class="history-count">{{ messages.length }}</span>
</div>
<div class="history-list">
<!-- Session Pills -->
<div v-if="sessions.length > 1" class="session-pills">
<button
v-for="session in sessions"
:key="session.id"
class="session-pill"
:class="{ active: session.id === activeSessionId }"
@click="activeSessionId = session.id"
>
{{ formatSessionTime(session.startTime) }}
<span class="pill-count">({{ session.messageCount }})</span>
</button>
</div>
<!-- Loading -->
<div v-if="loading" class="history-loading">
<span class="loading-dots">...</span>
</div>
<!-- Error -->
<div v-else-if="error" class="history-error">{{ error }}</div>
<!-- Empty -->
<div v-else-if="messages.length === 0 && !loading" class="history-empty">
Sin mensajes en esta sesión
</div>
<!-- Message List -->
<div v-else class="history-list">
<div
v-for="entry in mockEntries"
:key="entry.id"
v-for="msg in messages"
:key="msg.uuid"
class="history-entry"
:class="entry.role"
:class="msg.role === 'assistant' ? 'agent' : 'user'"
>
<div class="entry-meta">
<span class="role-badge" :class="entry.role">
{{ entry.role === 'user' ? 'Tú' : agent.uiConfig?.shortLabel || 'AG' }}
<span
class="role-badge"
:class="msg.role === 'assistant' ? 'agent' : 'user'"
>
{{ msg.role === 'user' ? 'Tú' : agent.uiConfig?.shortLabel || 'AG' }}
</span>
<span class="entry-time">{{ formatSessionTime(msg.timestamp) }}</span>
<span v-if="msg.toolCalls && msg.toolCalls.length > 0" class="tool-badge">
{{ msg.toolCalls.length }} tools
</span>
<span class="entry-time">{{ entry.timestamp }}</span>
<!-- Mic icon for voice entries -->
<svg v-if="entry.method === 'voice'" class="method-icon" xmlns="http://www.w3.org/2000/svg" width="12" height="12" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round">
<path d="M12 1a3 3 0 0 0-3 3v8a3 3 0 0 0 6 0V4a3 3 0 0 0-3-3z"/>
<path d="M19 10v2a7 7 0 0 1-14 0v-2"/>
<line x1="12" y1="19" x2="12" y2="23"/>
<line x1="8" y1="23" x2="16" y2="23"/>
</svg>
</div>
<div class="entry-content">{{ entry.content }}</div>
<div class="entry-content">{{ truncateContent(msg.content) }}</div>
</div>
</div>
<!-- Stats Bar -->
<div v-if="stats && !loading && messages.length > 0" class="stats-bar">
<span>{{ shortModel(stats.model) }}</span>
<span class="stats-sep">&middot;</span>
<span>{{ formatDuration(stats.duration) }}</span>
<span class="stats-sep">&middot;</span>
<span>{{ formatTokens(stats.totalInput + stats.totalOutput) }}</span>
<span class="stats-sep">&middot;</span>
<span>{{ stats.toolCallCount }} tools</span>
</div>
</div>
</template>
@@ -115,8 +235,53 @@ const mockEntries: ConversationEntry[] = [
border-radius: 8px;
}
/* Session Pills */
.session-pills {
display: flex;
gap: 6px;
overflow-x: auto;
padding-bottom: 8px;
margin-bottom: 8px;
scrollbar-width: none;
}
.session-pills::-webkit-scrollbar {
display: none;
}
.session-pill {
flex-shrink: 0;
font-size: 10px;
font-weight: 600;
padding: 3px 8px;
border-radius: 10px;
border: 1px solid rgba(255, 255, 255, 0.08);
background: rgba(255, 255, 255, 0.04);
color: rgba(255, 255, 255, 0.5);
cursor: pointer;
transition: all 0.15s ease;
white-space: nowrap;
}
.session-pill:hover {
background: rgba(255, 255, 255, 0.08);
color: rgba(255, 255, 255, 0.7);
}
.session-pill.active {
background: rgba(139, 92, 246, 0.15);
border-color: rgba(139, 92, 246, 0.3);
color: rgba(139, 92, 246, 0.9);
}
.pill-count {
opacity: 0.6;
margin-left: 2px;
}
/* Message List */
.history-list {
max-height: 300px;
max-height: 220px;
overflow-y: auto;
display: flex;
flex-direction: column;
@@ -168,8 +333,13 @@ const mockEntries: ConversationEntry[] = [
color: rgba(255, 255, 255, 0.25);
}
.method-icon {
color: rgba(255, 255, 255, 0.25);
.tool-badge {
font-size: 9px;
font-weight: 600;
padding: 1px 5px;
border-radius: 4px;
color: rgba(245, 158, 11, 0.9);
background: rgba(245, 158, 11, 0.12);
}
.entry-content {
@@ -178,8 +348,57 @@ const mockEntries: ConversationEntry[] = [
color: rgba(255, 255, 255, 0.7);
}
/* Stats Bar */
.stats-bar {
display: flex;
align-items: center;
gap: 6px;
margin-top: 8px;
padding-top: 8px;
border-top: 1px solid rgba(255, 255, 255, 0.06);
font-family: monospace;
font-size: 10px;
color: rgba(255, 255, 255, 0.35);
}
.stats-sep {
opacity: 0.4;
}
/* States */
.history-loading {
padding: 16px 0;
text-align: center;
color: rgba(255, 255, 255, 0.3);
font-size: 12px;
}
.loading-dots {
animation: pulse 1s ease-in-out infinite;
}
.history-error {
padding: 12px 0;
text-align: center;
color: rgba(239, 68, 68, 0.7);
font-size: 11px;
}
.history-empty {
padding: 16px 0;
text-align: center;
color: rgba(255, 255, 255, 0.25);
font-size: 11px;
font-style: italic;
}
@keyframes slide-in {
from { opacity: 0; transform: translateY(-8px); }
to { opacity: 1; transform: translateY(0); }
}
@keyframes pulse {
0%, 100% { opacity: 0.3; }
50% { opacity: 1; }
}
</style>

View File

@@ -0,0 +1,328 @@
<script setup lang="ts">
import type { VoiceCapture } from '../../composables/useVoiceCapture'
const props = defineProps<{
voice: VoiceCapture
}>()
const emit = defineEmits<{
close: []
}>()
function handleMicSelect(e: Event) {
const target = e.target as HTMLSelectElement
props.voice.selectMicrophone(target.value)
}
</script>
<template>
<div class="input-settings">
<div class="is-header">
<span class="is-title">Voice Settings</span>
<button class="is-close" @click="emit('close')">
<svg width="10" height="10" viewBox="0 0 10 10">
<line x1="1" y1="1" x2="9" y2="9" stroke="currentColor" stroke-width="1.5" stroke-linecap="round"/>
<line x1="9" y1="1" x2="1" y2="9" stroke="currentColor" stroke-width="1.5" stroke-linecap="round"/>
</svg>
</button>
</div>
<!-- Voice Mode Toggle -->
<div class="is-section">
<label class="is-label">Mode</label>
<div class="is-mode-row">
<button
class="is-mode-btn"
:class="{ active: voice.voiceMode.value === 'webspeech' }"
:disabled="voice.isRecording.value"
@click="voice.voiceMode.value !== 'webspeech' && voice.toggleWhisperMode()"
>
<span class="is-mode-icon">Web</span>
<span class="is-mode-label">Speech API</span>
</button>
<button
class="is-mode-btn"
:class="{
active: voice.voiceMode.value === 'whisper',
loading: voice.whisperStatus.value === 'loading'
}"
:disabled="voice.isRecording.value"
@click="voice.voiceMode.value !== 'whisper' && voice.toggleWhisperMode()"
>
<span class="is-mode-icon">GPU</span>
<span class="is-mode-label">Whisper</span>
</button>
</div>
<div class="is-status">
<span
class="is-status-dot"
:class="{
ready: voice.whisperStatus.value === 'ready',
loading: voice.whisperStatus.value === 'loading',
offline: voice.whisperStatus.value === 'offline'
}"
></span>
<span class="is-status-text">
{{ voice.voiceMode.value === 'whisper'
? (voice.whisperStatus.value === 'ready' ? 'Whisper ready' : voice.whisperStatus.value === 'loading' ? 'Starting...' : 'Offline')
: 'Web Speech API'
}}
</span>
</div>
</div>
<!-- Microphone Selection -->
<div class="is-section">
<label class="is-label">Microphone</label>
<select
class="is-select"
:value="voice.selectedDeviceId.value"
@change="handleMicSelect"
>
<option
v-for="device in voice.audioDevices.value"
:key="device.deviceId"
:value="device.deviceId"
>
{{ device.label || `Microphone ${voice.audioDevices.value.indexOf(device) + 1}` }}
</option>
</select>
</div>
<!-- Debug: Last Audio Playback -->
<div class="is-section">
<label class="is-label">Debug</label>
<button
class="is-debug-btn"
:class="{ playing: voice.isPlayingAudio.value }"
:disabled="!voice.lastAudioUrl.value"
@click="voice.playLastAudio()"
>
<svg v-if="!voice.isPlayingAudio.value" xmlns="http://www.w3.org/2000/svg" width="14" height="14" viewBox="0 0 24 24" fill="currentColor" stroke="none">
<polygon points="5 3 19 12 5 21 5 3"/>
</svg>
<svg v-else xmlns="http://www.w3.org/2000/svg" width="14" height="14" viewBox="0 0 24 24" fill="currentColor" stroke="none">
<rect x="6" y="4" width="4" height="16" rx="1"/>
<rect x="14" y="4" width="4" height="16" rx="1"/>
</svg>
<span>{{ voice.isPlayingAudio.value ? 'Stop' : 'Play last audio' }}</span>
</button>
</div>
</div>
</template>
<style scoped>
.input-settings {
background: rgba(255, 255, 255, 0.04);
border: 1px solid rgba(255, 255, 255, 0.08);
border-radius: 10px;
padding: 12px;
margin-top: 8px;
animation: settings-in 0.2s ease-out;
}
.is-header {
display: flex;
align-items: center;
justify-content: space-between;
margin-bottom: 12px;
}
.is-title {
font-size: 11px;
font-weight: 600;
color: rgba(255, 255, 255, 0.5);
text-transform: uppercase;
letter-spacing: 0.5px;
}
.is-close {
width: 20px;
height: 20px;
display: flex;
align-items: center;
justify-content: center;
background: transparent;
border: none;
color: rgba(255, 255, 255, 0.3);
cursor: pointer;
border-radius: 4px;
transition: all 0.15s;
}
.is-close:hover {
background: rgba(255, 255, 255, 0.1);
color: rgba(255, 255, 255, 0.7);
}
.is-section {
margin-bottom: 10px;
}
.is-section:last-child {
margin-bottom: 0;
}
.is-label {
display: block;
font-size: 11px;
font-weight: 500;
color: rgba(255, 255, 255, 0.4);
margin-bottom: 6px;
}
.is-mode-row {
display: flex;
gap: 6px;
}
.is-mode-btn {
flex: 1;
display: flex;
flex-direction: column;
align-items: center;
gap: 2px;
padding: 8px 6px;
background: rgba(255, 255, 255, 0.04);
border: 1px solid rgba(255, 255, 255, 0.08);
border-radius: 8px;
cursor: pointer;
transition: all 0.15s;
color: rgba(255, 255, 255, 0.4);
}
.is-mode-btn:hover:not(:disabled) {
background: rgba(255, 255, 255, 0.08);
color: rgba(255, 255, 255, 0.7);
}
.is-mode-btn.active {
background: rgba(139, 92, 246, 0.15);
border-color: rgba(139, 92, 246, 0.3);
color: rgba(139, 92, 246, 0.9);
}
.is-mode-btn.loading {
border-color: rgba(251, 191, 36, 0.3);
color: rgba(251, 191, 36, 0.8);
}
.is-mode-btn:disabled {
opacity: 0.4;
cursor: not-allowed;
}
.is-mode-icon {
font-size: 11px;
font-weight: 700;
letter-spacing: 0.5px;
}
.is-mode-label {
font-size: 10px;
font-weight: 500;
}
.is-status {
display: flex;
align-items: center;
gap: 6px;
margin-top: 6px;
}
.is-status-dot {
width: 6px;
height: 6px;
border-radius: 50%;
background: rgba(255, 255, 255, 0.2);
}
.is-status-dot.ready {
background: #22c55e;
box-shadow: 0 0 6px rgba(34, 197, 94, 0.4);
}
.is-status-dot.loading {
background: #fbbf24;
animation: status-pulse 1s ease-in-out infinite;
}
.is-status-dot.offline {
background: rgba(255, 255, 255, 0.2);
}
.is-status-text {
font-size: 11px;
color: rgba(255, 255, 255, 0.4);
}
.is-select {
width: 100%;
background: rgba(255, 255, 255, 0.06);
border: 1px solid rgba(255, 255, 255, 0.08);
border-radius: 8px;
padding: 7px 10px;
font-size: 12px;
color: rgba(255, 255, 255, 0.8);
outline: none;
cursor: pointer;
font-family: system-ui, sans-serif;
appearance: none;
-webkit-appearance: none;
background-image: url("data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' width='12' height='12' viewBox='0 0 24 24' fill='none' stroke='rgba(255,255,255,0.3)' stroke-width='2' stroke-linecap='round' stroke-linejoin='round'%3E%3Cpolyline points='6 9 12 15 18 9'%3E%3C/polyline%3E%3C/svg%3E");
background-repeat: no-repeat;
background-position: right 8px center;
padding-right: 28px;
}
.is-select:focus {
border-color: rgba(255, 255, 255, 0.2);
}
.is-select option {
background: #1a1025;
color: rgba(255, 255, 255, 0.85);
}
.is-debug-btn {
display: flex;
align-items: center;
gap: 6px;
width: 100%;
padding: 7px 10px;
background: rgba(255, 255, 255, 0.04);
border: 1px solid rgba(255, 255, 255, 0.08);
border-radius: 8px;
color: rgba(255, 255, 255, 0.4);
font-size: 12px;
font-family: system-ui, sans-serif;
cursor: pointer;
transition: all 0.15s;
}
.is-debug-btn:hover:not(:disabled) {
background: rgba(255, 255, 255, 0.08);
color: rgba(255, 255, 255, 0.7);
}
.is-debug-btn:disabled {
opacity: 0.3;
cursor: not-allowed;
}
.is-debug-btn.playing {
background: rgba(59, 130, 246, 0.12);
border-color: rgba(59, 130, 246, 0.25);
color: rgba(59, 130, 246, 0.9);
}
@keyframes settings-in {
from { opacity: 0; transform: translateY(4px); }
to { opacity: 1; transform: translateY(0); }
}
@keyframes status-pulse {
0%, 100% { opacity: 1; }
50% { opacity: 0.4; }
}
</style>

View File

@@ -1,8 +1,11 @@
<script setup lang="ts">
import { ref, reactive, computed, watch, nextTick, onMounted, onBeforeUnmount } from 'vue'
import type { Agent } from '../../types/agent'
import { useVoiceCapture } from '../../composables/useVoiceCapture'
import { useCanvasStore } from '../../stores/canvas'
import ChatInput from './ChatInput.vue'
import TranscriptCard from './TranscriptCard.vue'
import InputSettings from './InputSettings.vue'
import ConversationHistory from './ConversationHistory.vue'
interface ChatMessage {
@@ -35,11 +38,17 @@ const emit = defineEmits<{
submit: [text: string]
}>()
const canvasStore = useCanvasStore()
const voice = useVoiceCapture({
onNotification: (msg, type, duration) => canvasStore.showNotification(msg, type, duration)
})
const contentEl = ref<HTMLDivElement | null>(null)
const chatInputEl = ref<InstanceType<typeof ChatInput> | null>(null)
const isRecording = ref(false)
const showTranscript = ref(false)
const showHistory = ref(false)
const showSettings = ref(false)
const messages = reactive<ChatMessage[]>([])
const panelStyle = computed(() => {
@@ -60,7 +69,7 @@ const panelStyle = computed(() => {
})
const hasContent = computed(() =>
messages.length > 0 || showTranscript.value || showHistory.value
messages.length > 0 || showTranscript.value || showHistory.value || showSettings.value
)
async function scrollToBottom() {
@@ -100,6 +109,8 @@ function handleTranscriptDone(text: string) {
isRecording.value = false
showTranscript.value = false
if (!text.trim()) return
messages.push({ id: ++idCounter, role: 'user', content: text, status: 'sent' })
const agentMsg: ChatMessage = { id: ++idCounter, role: 'agent', content: '', status: 'thinking' }
@@ -108,6 +119,10 @@ function handleTranscriptDone(text: string) {
pushAgentResponse(agentMsg)
}
function toggleSettings() {
showSettings.value = !showSettings.value
}
function toggleHistory() {
showHistory.value = !showHistory.value
if (showHistory.value) scrollToBottom()
@@ -122,14 +137,22 @@ watch(() => props.visible, async (v) => {
isRecording.value = false
showTranscript.value = false
showHistory.value = false
showSettings.value = false
messages.length = 0
idCounter = 0
await voice.init()
await nextTick()
if (props.startRecording) {
handleMic()
} else {
chatInputEl.value?.focus()
}
} else {
// Cleanup when panel closes
if (voice.isRecording.value) {
voice.stopRecording()
}
voice.cleanup()
}
})
@@ -142,6 +165,7 @@ onMounted(() => {
onBeforeUnmount(() => {
document.removeEventListener('keydown', handleKeydown)
if (thinkTimer) clearTimeout(thinkTimer)
voice.cleanup()
})
</script>
@@ -187,7 +211,8 @@ onBeforeUnmount(() => {
</template>
</div>
<TranscriptCard v-if="showTranscript" @done="handleTranscriptDone" />
<TranscriptCard v-if="showTranscript" :voice="voice" @done="handleTranscriptDone" />
<InputSettings v-if="showSettings" :voice="voice" @close="showSettings = false" />
<ConversationHistory v-if="showHistory" :agent="agent" />
</div>
@@ -197,10 +222,12 @@ onBeforeUnmount(() => {
:placeholder="`Mensaje a ${agent.uiConfig?.label || agent.name}...`"
:recording="isRecording"
:history-active="showHistory"
:settings-active="showSettings"
:autofocus="visible"
@submit="handleSubmit"
@mic="handleMic"
@toggle-history="toggleHistory"
@toggle-settings="toggleSettings"
/>
</div>
</div>

View File

@@ -1,52 +1,62 @@
<script setup lang="ts">
import { ref, onMounted, onBeforeUnmount } from 'vue'
import { onMounted, onBeforeUnmount } from 'vue'
import type { VoiceCapture } from '../../composables/useVoiceCapture'
const props = withDefaults(defineProps<{
typeSpeed?: number
}>(), {
typeSpeed: 30
})
const props = defineProps<{
voice: VoiceCapture
}>()
const emit = defineEmits<{
done: [text: string]
}>()
const PLACEHOLDER_TEXT = 'Necesito que revises el componente de autenticación en el módulo de usuarios. ' +
'Hay un problema con la validación de tokens JWT cuando el usuario tiene sesiones múltiples activas. ' +
'El token se invalida correctamente en el servidor pero el cliente sigue usando el token anterior ' +
'hasta que expira naturalmente. Quiero que implementes una verificación en tiempo real usando WebSocket ' +
'para notificar al cliente cuando su token ha sido revocado desde otra sesión.'
const displayedText = ref('')
let intervalId: number | null = null
let charIndex = 0
function handleStop() {
props.voice.stopRecording()
// Brief delay for final Whisper result
const delay = props.voice.voiceMode.value === 'whisper' ? 800 : 200
setTimeout(() => {
emit('done', props.voice.transcript.value.trim())
}, delay)
}
onMounted(() => {
intervalId = window.setInterval(() => {
if (charIndex < PLACEHOLDER_TEXT.length) {
displayedText.value += PLACEHOLDER_TEXT[charIndex]
charIndex++
} else {
if (intervalId) clearInterval(intervalId)
intervalId = null
emit('done', displayedText.value)
}
}, props.typeSpeed)
props.voice.clearTranscript()
props.voice.startRecording()
})
onBeforeUnmount(() => {
if (intervalId) clearInterval(intervalId)
if (props.voice.isRecording.value) {
props.voice.stopRecording()
}
})
</script>
<template>
<div class="transcript-card">
<div class="transcript-header">
<div class="transcript-header-left">
<span class="rec-dot"></span>
<span class="rec-label">Transcribiendo...</span>
<span class="mode-badge" :class="voice.voiceMode.value">
{{ voice.voiceMode.value === 'whisper' ? 'GPU' : 'Web' }}
</span>
</div>
<div class="transcript-body">
<span class="transcript-text">{{ displayedText }}</span>
<button class="stop-btn" @click="handleStop" title="Stop recording">
<svg width="10" height="10" viewBox="0 0 10 10">
<rect x="1" y="1" width="8" height="8" rx="1" fill="currentColor"/>
</svg>
</button>
</div>
<!-- Error state -->
<div v-if="voice.error.value" class="transcript-error">
{{ voice.error.value }}
</div>
<!-- Transcript body -->
<div v-else class="transcript-body">
<span class="transcript-text">{{ voice.animatedTranscript.value }}</span>
<span v-if="voice.interimTranscript.value" class="interim-text">{{ voice.interimTranscript.value }}</span>
<span class="blink-cursor">|</span>
</div>
</div>
@@ -64,10 +74,16 @@ onBeforeUnmount(() => {
.transcript-header {
display: flex;
align-items: center;
gap: 8px;
justify-content: space-between;
margin-bottom: 8px;
}
.transcript-header-left {
display: flex;
align-items: center;
gap: 8px;
}
.rec-dot {
width: 8px;
height: 8px;
@@ -85,16 +101,67 @@ onBeforeUnmount(() => {
letter-spacing: 0.5px;
}
.mode-badge {
font-size: 9px;
font-weight: 700;
padding: 2px 5px;
border-radius: 4px;
text-transform: uppercase;
letter-spacing: 0.3px;
}
.mode-badge.whisper {
background: rgba(139, 92, 246, 0.2);
color: rgba(139, 92, 246, 0.9);
border: 1px solid rgba(139, 92, 246, 0.3);
}
.mode-badge.webspeech {
background: rgba(59, 130, 246, 0.2);
color: rgba(59, 130, 246, 0.9);
border: 1px solid rgba(59, 130, 246, 0.3);
}
.stop-btn {
width: 24px;
height: 24px;
display: flex;
align-items: center;
justify-content: center;
background: rgba(239, 68, 68, 0.15);
border: 1px solid rgba(239, 68, 68, 0.3);
border-radius: 6px;
color: #ef4444;
cursor: pointer;
transition: all 0.15s;
}
.stop-btn:hover {
background: rgba(239, 68, 68, 0.25);
}
.transcript-body {
font-size: 13px;
line-height: 1.6;
color: rgba(255, 255, 255, 0.85);
min-height: 20px;
}
.transcript-text {
white-space: pre-wrap;
}
.interim-text {
color: rgba(255, 255, 255, 0.4);
font-style: italic;
}
.transcript-error {
font-size: 12px;
color: rgba(239, 68, 68, 0.8);
padding: 4px 0;
}
.blink-cursor {
color: rgba(255, 255, 255, 0.7);
animation: cursor-blink 0.8s step-end infinite;

View File

@@ -0,0 +1,806 @@
import { ref, watch, type Ref } from 'vue'
import { endpoints } from '../config/endpoints'
// Web Speech API types (not in default TS lib)
interface SpeechRecognitionEvent extends Event {
resultIndex: number
results: SpeechRecognitionResultList
}
interface SpeechRecognitionErrorEvent extends Event {
error: string
message?: string
}
interface SpeechRecognition extends EventTarget {
continuous: boolean
interimResults: boolean
lang: string
onresult: ((event: SpeechRecognitionEvent) => void) | null
onerror: ((event: SpeechRecognitionErrorEvent) => void) | null
onend: (() => void) | null
start(): void
stop(): void
abort(): void
}
export type VoiceMode = 'webspeech' | 'whisper'
export type WhisperStatus = 'offline' | 'loading' | 'ready'
export interface VoiceCapture {
// State
isRecording: Ref<boolean>
transcript: Ref<string>
interimTranscript: Ref<string>
animatedTranscript: Ref<string>
error: Ref<string>
voiceMode: Ref<VoiceMode>
whisperStatus: Ref<WhisperStatus>
audioDevices: Ref<MediaDeviceInfo[]>
selectedDeviceId: Ref<string>
isAndroid: Ref<boolean>
lastAudioUrl: Ref<string>
isPlayingAudio: Ref<boolean>
// Actions
startRecording: () => void
stopRecording: () => void
toggleWhisperMode: () => Promise<void>
checkWhisperStatus: () => Promise<any>
loadAudioDevices: () => Promise<void>
selectMicrophone: (deviceId: string) => void
playLastAudio: () => void
init: () => Promise<void>
cleanup: () => void
clearTranscript: () => void
}
export function useVoiceCapture(options?: {
onNotification?: (message: string, type: 'info' | 'success' | 'error', duration?: number) => void
}): VoiceCapture {
const notify = options?.onNotification || (() => {})
// ====== State ======
const isRecording = ref(false)
const transcript = ref('')
const interimTranscript = ref('')
const animatedTranscript = ref('')
const error = ref('')
const voiceMode = ref<VoiceMode>('webspeech')
const whisperStatus = ref<WhisperStatus>('offline')
const audioDevices = ref<MediaDeviceInfo[]>([])
const selectedDeviceId = ref('')
const isAndroid = ref(false)
// Audio debug & save
const lastAudioUrl = ref('')
const isPlayingAudio = ref(false)
// ====== Internal state ======
let recognition: SpeechRecognition | null = null
let lastProcessedResult = ''
// Typing animation
let typingTimeout: number | null = null
let lastAnimatedLength = 0
// Whisper
const WHISPER_WS_URL = endpoints.whisper
let whisperSocket: WebSocket | null = null
let mediaRecorder: MediaRecorder | null = null
let audioChunks: Blob[] = []
let chunkInterval: number | null = null
const CHUNK_INTERVAL_MS = 3000
let mediaStream: MediaStream | null = null
let supportedMimeType = 'audio/webm;codecs=opus'
// Audio playback debug
let audioElement: HTMLAudioElement | null = null
let recordingStartTime = 0
// ====== Mobile / Audio Format ======
function checkMobile() {
const ua = navigator.userAgent
isAndroid.value = /Android/i.test(ua)
}
function detectAudioFormat(): string {
const formats = [
'audio/webm;codecs=opus',
'audio/webm',
'audio/mp4',
'audio/mp4;codecs=mp4a.40.2',
'audio/aac',
'audio/ogg;codecs=opus',
'audio/wav'
]
for (const format of formats) {
if (MediaRecorder.isTypeSupported(format)) {
console.log(`[VoiceCapture] Using audio format: ${format}`)
return format
}
}
console.warn('[VoiceCapture] No preferred format supported, using default')
return ''
}
// ====== Web Speech API ======
function initRecognition(): SpeechRecognition | null {
const SpeechRecognitionCtor = (window as any).SpeechRecognition || (window as any).webkitSpeechRecognition
if (!SpeechRecognitionCtor) {
error.value = 'Speech recognition not supported in this browser'
return null
}
const rec: SpeechRecognition = new SpeechRecognitionCtor()
rec.continuous = !isAndroid.value
rec.interimResults = true
rec.lang = 'es-419'
if (isAndroid.value) {
console.log('[VoiceCapture] Android detected - using non-continuous mode')
}
rec.onresult = (event: SpeechRecognitionEvent) => {
let interim = ''
let final = ''
for (let i = event.resultIndex; i < event.results.length; i++) {
const result = event.results[i]
if (!result || !result[0]) continue
if (result.isFinal) {
final += result[0].transcript + ' '
} else {
interim += result[0].transcript
}
}
if (final) {
const trimmedFinal = final.trim()
if (isAndroid.value && lastProcessedResult && trimmedFinal.startsWith(lastProcessedResult.trim())) {
const newPart = trimmedFinal.slice(lastProcessedResult.trim().length).trim()
if (newPart) {
transcript.value += newPart + ' '
lastProcessedResult = trimmedFinal
}
} else {
transcript.value += final
lastProcessedResult = trimmedFinal
}
}
interimTranscript.value = interim
}
rec.onerror = (event: SpeechRecognitionErrorEvent) => {
// no-speech and aborted are transient — don't kill the session
if (event.error === 'no-speech' || event.error === 'aborted') {
console.log('[VoiceCapture] Transient error:', event.error, '(will auto-restart)')
return
}
console.error('[VoiceCapture] Recognition error:', event.error)
if (event.error === 'not-allowed') {
error.value = 'Microphone access denied'
} else {
error.value = `Error: ${event.error}`
}
isRecording.value = false
}
rec.onend = () => {
if (isRecording.value && voiceMode.value === 'webspeech') {
if (isAndroid.value) {
isRecording.value = false
console.log('[VoiceCapture] Android session ended - tap mic to continue')
} else {
rec.start()
}
}
}
return rec
}
// ====== Whisper Functions ======
async function checkWhisperStatusFn(updateLoading = true): Promise<any> {
try {
const res = await fetch('/api/whisper/status')
const data = await res.json()
if (data.enabled) {
voiceMode.value = 'whisper'
}
if (data.running) {
whisperStatus.value = 'ready'
} else if (updateLoading && (data.starting || false)) {
whisperStatus.value = 'loading'
} else if (!data.running) {
if (voiceMode.value === 'whisper' && !data.starting) {
whisperStatus.value = 'offline'
}
}
return data
} catch {
voiceMode.value = 'webspeech'
whisperStatus.value = 'offline'
return null
}
}
async function pollWhisperStatus(): Promise<void> {
const maxAttempts = 60
let attempts = 0
while (attempts < maxAttempts) {
await new Promise(resolve => setTimeout(resolve, 2000))
attempts++
try {
const status = await checkWhisperStatusFn(false)
if (!status) continue
if (status.starting) {
console.log(`[VoiceCapture] Still starting... (${attempts * 2}s)`)
continue
}
if (status.running && status.enabled) {
console.log('[VoiceCapture] Server ready!')
notify('Whisper GPU ready!', 'success')
connectWhisperSocket()
whisperStatus.value = 'ready'
return
}
console.log('[VoiceCapture] Server failed to start')
notify('Whisper server failed to start', 'error')
whisperStatus.value = 'offline'
return
} catch (e) {
console.error('[VoiceCapture] Polling error:', e)
}
}
notify('Whisper server timeout', 'error')
whisperStatus.value = 'offline'
}
function connectWhisperSocket() {
if (whisperStatus.value !== 'ready') {
console.log('[VoiceCapture] Whisper not ready, skipping connection')
return
}
if (whisperSocket?.readyState === WebSocket.OPEN) return
console.log('[VoiceCapture] Connecting to Whisper at:', WHISPER_WS_URL)
whisperSocket = new WebSocket(WHISPER_WS_URL)
const connectionTimeout = setTimeout(() => {
if (whisperSocket && whisperSocket.readyState !== WebSocket.OPEN) {
console.error('[VoiceCapture] Whisper connection timeout (10s)')
whisperSocket.close()
whisperStatus.value = 'offline'
}
}, 10000)
whisperSocket.onopen = () => {
clearTimeout(connectionTimeout)
console.log('[VoiceCapture] Whisper WebSocket connected')
whisperStatus.value = 'ready'
}
whisperSocket.onmessage = (event) => {
try {
const msg = JSON.parse(event.data)
if (msg.type === 'ready') {
console.log('[VoiceCapture] Whisper ready:', msg.model, msg.device)
whisperStatus.value = 'ready'
} else if (msg.type === 'transcription') {
if (msg.success && msg.text) {
const fullText = msg.text.trim()
if (msg.partial) {
transcript.value = fullText + ' '
interimTranscript.value = ''
} else {
transcript.value = fullText + ' '
interimTranscript.value = ''
console.log(`[VoiceCapture] WHISPER-GPU (${msg.model}/${msg.device}):`, fullText)
}
} else if (msg.error) {
error.value = msg.error
console.error('[VoiceCapture] Whisper error:', msg.error)
}
}
} catch (e) {
console.error('[VoiceCapture] Whisper message error:', e)
}
}
whisperSocket.onclose = () => {
console.log('[VoiceCapture] Whisper WebSocket closed')
whisperStatus.value = 'offline'
}
whisperSocket.onerror = (e) => {
console.error('[VoiceCapture] Whisper WebSocket error:', e)
whisperStatus.value = 'offline'
}
}
function disconnectWhisperSocket() {
if (whisperSocket) {
whisperSocket.close()
whisperSocket = null
}
whisperStatus.value = 'offline'
}
async function startWhisperRecording() {
if (!whisperSocket || whisperSocket.readyState !== WebSocket.OPEN) {
console.warn('[VoiceCapture] Whisper socket not connected, attempting to connect...')
connectWhisperSocket()
await new Promise(resolve => setTimeout(resolve, 500))
if (!whisperSocket || whisperSocket.readyState !== WebSocket.OPEN) {
error.value = 'Whisper server not connected'
notify('Whisper not connected. Try toggling GPU mode.', 'error')
return
}
}
try {
const audioConstraints: MediaTrackConstraints = {
echoCancellation: true,
noiseSuppression: true,
autoGainControl: true,
...(selectedDeviceId.value ? { deviceId: { exact: selectedDeviceId.value } } : {})
}
mediaStream = await navigator.mediaDevices.getUserMedia({ audio: audioConstraints })
const recorderOptions: MediaRecorderOptions = {}
if (supportedMimeType) {
recorderOptions.mimeType = supportedMimeType
}
mediaRecorder = new MediaRecorder(mediaStream, recorderOptions)
console.log(`[VoiceCapture] MediaRecorder using: ${mediaRecorder.mimeType}`)
audioChunks = []
mediaRecorder.ondataavailable = (event) => {
if (event.data.size > 0) {
audioChunks.push(event.data)
}
}
audioChunks = []
mediaRecorder.start(100)
isRecording.value = true
recordingStartTime = Date.now()
console.log(`[VoiceCapture] Whisper recording started`)
// Permission granted via user gesture — reload devices with labels
loadAudioDevices(true)
chunkInterval = window.setInterval(() => {
if (audioChunks.length > 0 && whisperSocket?.readyState === WebSocket.OPEN) {
sendAudioChunk(false)
}
}, CHUNK_INTERVAL_MS)
} catch (e: any) {
error.value = `Microphone error: ${e.message}`
console.error('[VoiceCapture] Microphone error:', e)
}
}
function sendAudioChunk(isFinal: boolean) {
if (audioChunks.length === 0) return
const mimeType = mediaRecorder?.mimeType || supportedMimeType || 'audio/webm'
const audioBlob = new Blob(audioChunks, { type: mimeType })
if (audioBlob.size < 5000) {
if (isFinal) audioChunks = []
return
}
if (isFinal) {
audioChunks = []
saveAudioForPlayback(audioBlob)
}
const reader = new FileReader()
reader.onloadend = () => {
const base64 = (reader.result as string).split(',')[1]
if (whisperSocket?.readyState === WebSocket.OPEN) {
whisperSocket.send(JSON.stringify({
type: 'transcribe',
audio: base64,
language: 'es',
partial: !isFinal
}))
}
}
reader.readAsDataURL(audioBlob)
}
function stopWhisperRecording() {
if (chunkInterval) {
clearInterval(chunkInterval)
chunkInterval = null
}
if (audioChunks.length > 0) {
sendAudioChunk(true)
}
if (mediaRecorder && mediaRecorder.state !== 'inactive') {
mediaRecorder.stop()
}
if (mediaStream) {
mediaStream.getTracks().forEach(track => track.stop())
mediaStream = null
}
isRecording.value = false
}
// ====== Audio Save & Debug Playback ======
function currentMicName(): string {
if (!selectedDeviceId.value) return 'Default'
const device = audioDevices.value.find(d => d.deviceId === selectedDeviceId.value)
return device?.label || 'Microphone'
}
function saveAudioForPlayback(blob: Blob) {
if (lastAudioUrl.value) {
URL.revokeObjectURL(lastAudioUrl.value)
}
lastAudioUrl.value = URL.createObjectURL(blob)
saveRecordingToBackend(blob)
}
async function saveRecordingToBackend(blob: Blob) {
try {
const duration_ms = Date.now() - recordingStartTime
const reader = new FileReader()
reader.onloadend = async () => {
const base64 = (reader.result as string).split(',')[1]
const response = await fetch('/api/recordings', {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({
audio: base64,
transcription: transcript.value.trim(),
microphone: currentMicName(),
duration_ms
})
})
const data = await response.json()
if (data.success) {
console.log(`[VoiceCapture] Recording saved: ${data.filename} (${(data.size / 1024).toFixed(1)} KB)`)
} else {
console.error('[VoiceCapture] Failed to save recording:', data.error)
}
}
reader.readAsDataURL(blob)
} catch (e) {
console.error('[VoiceCapture] Error saving recording:', e)
}
}
function playLastAudio() {
if (!lastAudioUrl.value) return
if (isPlayingAudio.value && audioElement) {
audioElement.pause()
audioElement.currentTime = 0
isPlayingAudio.value = false
return
}
audioElement = new Audio(lastAudioUrl.value)
audioElement.onplay = () => { isPlayingAudio.value = true }
audioElement.onended = () => { isPlayingAudio.value = false }
audioElement.onpause = () => { isPlayingAudio.value = false }
audioElement.play().catch(e => {
console.error('[VoiceCapture] Failed to play audio:', e)
isPlayingAudio.value = false
})
}
// ====== Parallel Audio Capture (for Web Speech mode) ======
async function startAudioCapture() {
try {
const audioConstraints: MediaTrackConstraints = {
echoCancellation: true,
noiseSuppression: true,
autoGainControl: true,
...(selectedDeviceId.value ? { deviceId: { exact: selectedDeviceId.value } } : {})
}
mediaStream = await navigator.mediaDevices.getUserMedia({ audio: audioConstraints })
const recorderOptions: MediaRecorderOptions = {}
if (supportedMimeType) {
recorderOptions.mimeType = supportedMimeType
}
mediaRecorder = new MediaRecorder(mediaStream, recorderOptions)
audioChunks = []
mediaRecorder.ondataavailable = (event) => {
if (event.data.size > 0) {
audioChunks.push(event.data)
}
}
mediaRecorder.start(100)
recordingStartTime = Date.now()
console.log(`[VoiceCapture] Audio capture started (${mediaRecorder.mimeType})`)
// Permission is now granted via user gesture — reload devices with labels
loadAudioDevices(true)
} catch (e: any) {
console.error('[VoiceCapture] Audio capture error:', e)
}
}
function stopAudioCapture() {
if (mediaRecorder && mediaRecorder.state !== 'inactive') {
mediaRecorder.stop()
}
// Build final blob and save
if (audioChunks.length > 0) {
const mimeType = mediaRecorder?.mimeType || supportedMimeType || 'audio/webm'
const audioBlob = new Blob(audioChunks, { type: mimeType })
audioChunks = []
if (audioBlob.size > 1000) {
saveAudioForPlayback(audioBlob)
}
}
if (mediaStream) {
mediaStream.getTracks().forEach(track => track.stop())
mediaStream = null
}
}
// ====== Public Recording API ======
function startRecording() {
error.value = ''
if (voiceMode.value === 'whisper' && whisperStatus.value === 'ready') {
startWhisperRecording()
} else {
if (!recognition) {
recognition = initRecognition()
}
if (recognition) {
try {
recognition.start()
isRecording.value = true
// Capture raw audio in parallel for save/debug playback
startAudioCapture()
if (isAndroid.value) {
notify('Android: Tap mic again to continue recording', 'info', 3000)
}
} catch (e) {
console.error('[VoiceCapture] Failed to start:', e)
}
}
}
}
function stopRecording() {
if (voiceMode.value === 'whisper') {
stopWhisperRecording()
} else {
if (recognition) {
recognition.stop()
}
stopAudioCapture()
isRecording.value = false
}
interimTranscript.value = ''
}
async function toggleWhisperMode() {
if (whisperStatus.value === 'loading') return
whisperStatus.value = 'loading'
error.value = ''
if (voiceMode.value !== 'whisper') {
notify('Starting Whisper GPU server...', 'info', 10000)
}
try {
const res = await fetch('/api/whisper/toggle', { method: 'POST' })
const data = await res.json()
if (data.starting) {
console.log('[VoiceCapture] Server starting, polling...')
voiceMode.value = 'whisper'
await pollWhisperStatus()
return
}
if (data.enabled) {
voiceMode.value = 'whisper'
whisperStatus.value = data.running ? 'ready' : 'offline'
if (data.running) {
notify('Whisper GPU ready!', 'success')
connectWhisperSocket()
}
} else {
voiceMode.value = 'webspeech'
whisperStatus.value = 'offline'
notify('Using Web Speech API', 'info')
disconnectWhisperSocket()
}
} catch (e: any) {
error.value = 'Failed to toggle Whisper'
notify('Error starting Whisper server', 'error')
console.error('[VoiceCapture] Whisper toggle error:', e)
whisperStatus.value = 'offline'
}
}
// ====== Microphone ======
async function loadAudioDevices(skipPermissionRequest = false) {
try {
if (!skipPermissionRequest) {
// Request permission to get device labels
const tempStream = await navigator.mediaDevices.getUserMedia({ audio: true })
tempStream.getTracks().forEach(track => track.stop())
}
const devices = await navigator.mediaDevices.enumerateDevices()
audioDevices.value = devices.filter(d => d.kind === 'audioinput')
if (!selectedDeviceId.value && audioDevices.value.length > 0) {
selectedDeviceId.value = audioDevices.value[0]?.deviceId || ''
}
console.log(`[VoiceCapture] Found ${audioDevices.value.length} audio devices`)
} catch (e) {
console.error('[VoiceCapture] Failed to enumerate devices:', e)
}
}
function selectMicrophone(deviceId: string) {
selectedDeviceId.value = deviceId
if (isRecording.value) {
stopRecording()
setTimeout(() => startRecording(), 100)
}
}
// ====== Typing Animation ======
function animateTyping(targetText: string) {
if (typingTimeout) {
clearTimeout(typingTimeout)
typingTimeout = null
}
if (targetText.length < animatedTranscript.value.length) {
animatedTranscript.value = targetText
lastAnimatedLength = targetText.length
return
}
const startIndex = lastAnimatedLength
function typeNext(index: number) {
if (index <= targetText.length) {
animatedTranscript.value = targetText.substring(0, index)
lastAnimatedLength = index
if (index < targetText.length) {
const delay = 15 + Math.random() * 10
typingTimeout = window.setTimeout(() => typeNext(index + 1), delay)
}
}
}
typeNext(startIndex)
}
watch(transcript, (newVal) => {
animateTyping(newVal)
})
// ====== Transcript ======
function clearTranscript() {
transcript.value = ''
interimTranscript.value = ''
animatedTranscript.value = ''
lastAnimatedLength = 0
lastProcessedResult = ''
if (typingTimeout) {
clearTimeout(typingTimeout)
typingTimeout = null
}
}
// ====== Lifecycle ======
async function init() {
recognition = initRecognition()
checkMobile()
supportedMimeType = detectAudioFormat()
// Only enumerate without getUserMedia — no user gesture here
// Devices will get full labels after first recording (user gesture)
await loadAudioDevices(true)
const status = await checkWhisperStatusFn()
if (status?.starting) {
console.log('[VoiceCapture] Server is starting, resuming polling...')
pollWhisperStatus()
} else if (voiceMode.value === 'whisper' && whisperStatus.value === 'ready') {
connectWhisperSocket()
} else if (voiceMode.value === 'whisper' && whisperStatus.value !== 'ready') {
console.log('[VoiceCapture] Whisper was enabled but server not running, disabling')
voiceMode.value = 'webspeech'
}
}
function cleanup() {
stopRecording()
recognition = null
disconnectWhisperSocket()
if (chunkInterval) clearInterval(chunkInterval)
if (typingTimeout) clearTimeout(typingTimeout)
if (mediaStream) {
mediaStream.getTracks().forEach(track => track.stop())
mediaStream = null
}
if (audioElement) {
audioElement.pause()
audioElement = null
}
if (lastAudioUrl.value) {
URL.revokeObjectURL(lastAudioUrl.value)
lastAudioUrl.value = ''
}
isPlayingAudio.value = false
}
return {
// State
isRecording,
transcript,
interimTranscript,
animatedTranscript,
error,
voiceMode,
whisperStatus,
audioDevices,
selectedDeviceId,
isAndroid,
lastAudioUrl,
isPlayingAudio,
// Actions
startRecording,
stopRecording,
toggleWhisperMode,
checkWhisperStatus: checkWhisperStatusFn,
loadAudioDevices,
selectMicrophone,
playLastAudio,
init,
cleanup,
clearTranscript
}
}

View File

@@ -46,3 +46,21 @@ export interface ConversationEntry {
timestamp: string
method: 'text' | 'voice'
}
export interface TranscriptSession {
id: string
startTime: string
messageCount: number
model: string
}
export interface TranscriptMessage {
uuid: string
role: 'user' | 'assistant'
content: string
timestamp: string
isMeta: boolean
tokens?: { input: number; output: number }
toolCalls?: string[]
hasThinking: boolean
}

View File

@@ -20,6 +20,7 @@ import {
handleAgentsPlugins, handleAgentsMcpJson,
handleAgentsConfigPermissions, handleAgentsConfigHooks, handleAgentsConfigMcp
} from './agents'
import { handleTranscript, handleTranscriptSessions } from './transcript'
export async function handleRequest(req: Request): Promise<Response> {
const url = new URL(req.url)
@@ -277,6 +278,20 @@ export async function handleRequest(req: Request): Promise<Response> {
return handleGitFile(url)
}
// Transcript
if (path === '/api/transcript/sessions' && req.method === 'GET') {
return handleTranscriptSessions()
}
if (path === '/api/transcript/latest' && req.method === 'GET') {
return handleTranscript(req, url, 'latest')
}
const transcriptMatch = path.match(/^\/api\/transcript\/([a-f0-9-]+)$/)
if (transcriptMatch && req.method === 'GET') {
return handleTranscript(req, url, transcriptMatch[1])
}
// Agents
if (path === '/api/agents' && req.method === 'GET') {
return handleAgents(req)

View File

@@ -0,0 +1,81 @@
import { jsonResponse, errorResponse } from '../utils/cors'
import { getTranscriptAnalysis, listSessions } from '../services/transcript-engine'
import type { TranscriptAnalysis } from '../services/transcript-engine'
export function handleTranscriptSessions(): Response {
const sessions = listSessions()
return jsonResponse(sessions)
}
export function handleTranscript(req: Request, url: URL, sessionId: string): Response {
if (req.method !== 'GET') return errorResponse('Method not allowed', 405)
const resolvedId = sessionId === 'latest' ? undefined : sessionId
const analysis = getTranscriptAnalysis(resolvedId)
if (!analysis) {
return errorResponse('Transcript not found', 404)
}
// Section filtering
const section = url.searchParams.get('section')
if (section) {
return handleSection(analysis, section)
}
// Full response (exclude thinking by default)
const includeThinking = url.searchParams.get('includeThinking') === 'true'
if (!includeThinking) {
return jsonResponse({
...analysis,
messages: analysis.messages.filter(m => !m.isMeta)
})
}
return jsonResponse({
...analysis,
messages: analysis.messages.filter(m => !m.isMeta)
})
}
function handleSection(analysis: TranscriptAnalysis, section: string): Response {
switch (section) {
case 'messages':
return jsonResponse({
sessionId: analysis.sessionId,
messages: analysis.messages.filter(m => !m.isMeta)
})
case 'tokens':
return jsonResponse({
sessionId: analysis.sessionId,
tokens: analysis.tokens
})
case 'tools':
return jsonResponse({
sessionId: analysis.sessionId,
tools: analysis.tools
})
case 'stats':
return jsonResponse({
sessionId: analysis.sessionId,
stats: analysis.stats,
model: analysis.model,
version: analysis.version,
duration: analysis.duration,
startTime: analysis.startTime,
endTime: analysis.endTime
})
case 'files':
return jsonResponse({
sessionId: analysis.sessionId,
filesModified: analysis.filesModified
})
case 'subagents':
return jsonResponse({
sessionId: analysis.sessionId,
subagents: analysis.subagents
})
default:
return errorResponse(`Unknown section: ${section}. Valid: messages, tokens, tools, stats, files, subagents`, 400)
}
}

View File

@@ -0,0 +1,574 @@
// Transcript Engine - Parses Claude Code JSONL transcripts
// Module-level state pattern (like terminal.ts, torch-handler.ts)
import { existsSync, readFileSync, statSync, readdirSync } from 'fs'
import { join } from 'path'
import { homedir } from 'os'
import { WORKING_DIR } from '../config'
// ── Types ──
export interface TranscriptAnalysis {
sessionId: string
model: string
version: string
gitBranch: string
cwd: string
startTime: string
endTime: string
duration: number
messages: TranscriptMessage[]
tokens: {
totalInput: number
totalOutput: number
totalCacheRead: number
totalCacheCreation: number
byTurn: TurnTokens[]
}
tools: {
summary: Record<string, number>
calls: ToolCall[]
}
filesModified: string[]
subagents: SubagentInfo[]
summaries: string[]
stats: {
messageCount: number
userMessageCount: number
assistantMessageCount: number
toolCallCount: number
thinkingBlocks: number
errors: number
}
}
export interface TranscriptMessage {
uuid: string
role: 'user' | 'assistant'
content: string
timestamp: string
isMeta: boolean
tokens?: { input: number; output: number }
toolCalls?: string[]
hasThinking: boolean
}
export interface ToolCall {
name: string
input: unknown
output?: string
timestamp: string
isError: boolean
}
export interface TurnTokens {
turnIndex: number
input: number
output: number
cacheRead: number
cacheCreation: number
model: string
}
export interface SubagentInfo {
agentId: string
prompt: string
timestamp: string
}
export interface SessionInfo {
id: string
startTime: string
messageCount: number
model: string
}
// ── Module-level cache ──
const cache = new Map<string, {
analysis: TranscriptAnalysis
lastModified: number
}>()
// ── Project hash ──
function getProjectHash(): string {
// C:\Users\jodar\agent-ui → C--Users-jodar-agent-ui
return WORKING_DIR.replace(/[\\/]/g, '-').replace(/:/g, '-')
}
function getProjectDir(): string {
return join(homedir(), '.claude', 'projects', getProjectHash())
}
// ── Path resolution ──
function resolveTranscriptPath(sessionId?: string): string | null {
const projectDir = getProjectDir()
if (!existsSync(projectDir)) return null
if (sessionId && sessionId !== 'latest') {
const filePath = join(projectDir, `${sessionId}.jsonl`)
return existsSync(filePath) ? filePath : null
}
// Find most recent by mtime
try {
const files = readdirSync(projectDir)
.filter(f => f.endsWith('.jsonl'))
.map(f => {
const fullPath = join(projectDir, f)
return { name: f, path: fullPath, mtime: statSync(fullPath).mtimeMs }
})
.sort((a, b) => b.mtime - a.mtime)
return files.length > 0 ? files[0].path : null
} catch {
return null
}
}
function sessionIdFromPath(filePath: string): string {
const basename = filePath.split(/[\\/]/).pop() || ''
return basename.replace('.jsonl', '')
}
// ── Helpers ──
function truncate(str: string, maxLen: number): string {
if (!str || str.length <= maxLen) return str || ''
return str.slice(0, maxLen) + '...'
}
function extractText(content: any): string {
if (typeof content === 'string') {
return content.replace(/<[^>]+>/g, '').trim()
}
if (Array.isArray(content)) {
return content
.filter((c: any) => c.type === 'text')
.map((c: any) => c.text || '')
.join('\n')
.replace(/<[^>]+>/g, '')
.trim()
}
return ''
}
// ── JSONL parsing ──
interface ParsedLine {
type: string
data: any
}
function parseTranscriptFile(filePath: string): ParsedLine[] {
const content = readFileSync(filePath, 'utf8')
const rawLines = content.trim().split('\n')
const lines: ParsedLine[] = []
for (const line of rawLines) {
if (!line.trim()) continue
try {
const obj = JSON.parse(line)
lines.push({ type: obj.type, data: obj })
} catch {
// Skip unparseable lines
}
}
return lines
}
// ── Build analysis from parsed lines ──
function buildAnalysis(lines: ParsedLine[], fileSessionId: string): TranscriptAnalysis {
let sessionId = fileSessionId
let model = ''
let version = ''
let gitBranch = ''
let cwd = ''
let startTime = ''
let endTime = ''
const messages: TranscriptMessage[] = []
const toolCalls: ToolCall[] = []
const filesModified = new Set<string>()
const subagents: SubagentInfo[] = []
const summaries: string[] = []
const turnTokens: TurnTokens[] = []
let totalInput = 0
let totalOutput = 0
let totalCacheRead = 0
let totalCacheCreation = 0
let thinkingBlocks = 0
let errors = 0
// Track assistant message chunks by message.id (streaming chunks share the same id)
const assistantChunks = new Map<string, {
uuid: string
timestamp: string
model: string
textParts: string[]
toolNames: string[]
hasThinking: boolean
usage: any
pendingToolCalls: { name: string; input: any; id: string; timestamp: string }[]
}>()
// Track tool results by tool_use_id
const toolResults = new Map<string, { content: string; isError: boolean }>()
// ── First pass: collect all data ──
for (const { type, data } of lines) {
// Extract metadata from first message that has it
if (data.sessionId && !sessionId) sessionId = data.sessionId
if (data.version && !version) version = data.version
if (data.gitBranch && !gitBranch) gitBranch = data.gitBranch
if (data.cwd && !cwd) cwd = data.cwd
// Track time bounds
if (data.timestamp) {
if (!startTime || data.timestamp < startTime) startTime = data.timestamp
if (!endTime || data.timestamp > endTime) endTime = data.timestamp
}
switch (type) {
case 'user': {
const msg = data.message
if (!msg) break
// Collect tool results (user messages contain tool_result blocks)
if (Array.isArray(msg.content)) {
for (const block of msg.content) {
if (block.type === 'tool_result') {
const resultText = typeof block.content === 'string'
? block.content
: Array.isArray(block.content)
? block.content.map((c: any) => c.text || '').join('\n')
: ''
toolResults.set(block.tool_use_id, {
content: truncate(resultText, 300),
isError: !!block.is_error
})
if (block.is_error) errors++
}
}
}
// Add as user message (skip meta, skip tool-result-only messages)
const isMeta = !!data.isMeta
const text = extractText(msg.content)
const hasToolResult = Array.isArray(msg.content) &&
msg.content.some((c: any) => c.type === 'tool_result')
if (text && !hasToolResult) {
messages.push({
uuid: data.uuid || '',
role: 'user',
content: text,
timestamp: data.timestamp || '',
isMeta,
hasThinking: false
})
}
break
}
case 'assistant': {
const msg = data.message
if (!msg || msg.role !== 'assistant') break
const msgId = msg.id || data.uuid
if (!model && msg.model) model = msg.model
let chunk = assistantChunks.get(msgId)
if (!chunk) {
chunk = {
uuid: data.uuid || '',
timestamp: data.timestamp || '',
model: msg.model || '',
textParts: [],
toolNames: [],
hasThinking: false,
usage: null,
pendingToolCalls: []
}
assistantChunks.set(msgId, chunk)
}
// Take latest usage (streaming chunks repeat usage, last is most accurate)
if (msg.usage) chunk.usage = msg.usage
// Process content blocks (each JSONL line typically has one block)
if (Array.isArray(msg.content)) {
for (const block of msg.content) {
if (block.type === 'text' && block.text?.trim()) {
chunk.textParts.push(block.text)
} else if (block.type === 'thinking') {
chunk.hasThinking = true
thinkingBlocks++
} else if (block.type === 'tool_use') {
chunk.toolNames.push(block.name)
chunk.pendingToolCalls.push({
name: block.name,
input: block.input,
id: block.id,
timestamp: data.timestamp || ''
})
}
}
}
break
}
case 'progress': {
if (data.data?.type === 'agent_progress' && data.data.agentId) {
const existing = subagents.find(s => s.agentId === data.data.agentId)
if (!existing) {
subagents.push({
agentId: data.data.agentId,
prompt: truncate(data.data.prompt || '', 200),
timestamp: data.timestamp || ''
})
}
}
break
}
case 'file-history-snapshot': {
const backups = data.snapshot?.trackedFileBackups
if (backups && typeof backups === 'object') {
for (const filePath of Object.keys(backups)) {
filesModified.add(filePath)
}
}
break
}
case 'summary': {
const summaryText = data.summary || data.message?.content
if (summaryText) {
summaries.push(truncate(
typeof summaryText === 'string' ? summaryText : JSON.stringify(summaryText),
1000
))
}
break
}
}
}
// ── Second pass: assemble assistant messages and finalize tool calls ──
let turnIndex = 0
for (const [, chunk] of assistantChunks) {
const text = chunk.textParts.join('\n').trim()
if (text || chunk.toolNames.length > 0) {
const msgTokens = chunk.usage
? { input: chunk.usage.input_tokens || 0, output: chunk.usage.output_tokens || 0 }
: undefined
messages.push({
uuid: chunk.uuid,
role: 'assistant',
content: text || `[Tool calls: ${chunk.toolNames.join(', ')}]`,
timestamp: chunk.timestamp,
isMeta: false,
tokens: msgTokens,
toolCalls: chunk.toolNames.length > 0 ? chunk.toolNames : undefined,
hasThinking: chunk.hasThinking
})
}
// Finalize tool calls with results
for (const tc of chunk.pendingToolCalls) {
const result = toolResults.get(tc.id)
const inputStr = typeof tc.input === 'string' ? tc.input : JSON.stringify(tc.input)
toolCalls.push({
name: tc.name,
input: truncate(inputStr, 500),
output: result?.content,
timestamp: tc.timestamp,
isError: result?.isError || false
})
}
// Token tracking per turn
if (chunk.usage) {
const u = chunk.usage
const input = u.input_tokens || 0
const output = u.output_tokens || 0
const cacheRead = u.cache_read_input_tokens || 0
const cacheCreation = u.cache_creation_input_tokens || 0
totalInput += input
totalOutput += output
totalCacheRead += cacheRead
totalCacheCreation += cacheCreation
turnTokens.push({
turnIndex: turnIndex++,
input,
output,
cacheRead,
cacheCreation,
model: chunk.model
})
}
}
// Sort messages chronologically
messages.sort((a, b) => a.timestamp.localeCompare(b.timestamp))
// Build tool summary
const toolSummary: Record<string, number> = {}
for (const tc of toolCalls) {
toolSummary[tc.name] = (toolSummary[tc.name] || 0) + 1
}
// Extract files from Edit/Write tool calls
for (const tc of toolCalls) {
if (['Edit', 'Write', 'NotebookEdit'].includes(tc.name) && tc.input) {
try {
const input = typeof tc.input === 'string' ? JSON.parse(tc.input) : tc.input
if (input.file_path) filesModified.add(input.file_path)
if (input.notebook_path) filesModified.add(input.notebook_path)
} catch { /* skip */ }
}
}
const duration = startTime && endTime
? new Date(endTime).getTime() - new Date(startTime).getTime()
: 0
const userMsgCount = messages.filter(m => m.role === 'user').length
const assistantMsgCount = messages.filter(m => m.role === 'assistant').length
return {
sessionId,
model,
version,
gitBranch,
cwd,
startTime,
endTime,
duration,
messages,
tokens: {
totalInput,
totalOutput,
totalCacheRead,
totalCacheCreation,
byTurn: turnTokens
},
tools: {
summary: toolSummary,
calls: toolCalls
},
filesModified: [...filesModified],
subagents,
summaries,
stats: {
messageCount: messages.length,
userMessageCount: userMsgCount,
assistantMessageCount: assistantMsgCount,
toolCallCount: toolCalls.length,
thinkingBlocks,
errors
}
}
}
// ── Exported API ──
export function getTranscriptAnalysis(sessionId?: string): TranscriptAnalysis | null {
const filePath = resolveTranscriptPath(sessionId)
if (!filePath) return null
const sid = sessionIdFromPath(filePath)
try {
const stat = statSync(filePath)
const mtime = stat.mtimeMs
// Return cached if file hasn't changed
const cached = cache.get(sid)
if (cached && cached.lastModified === mtime) {
return cached.analysis
}
// Full parse
const lines = parseTranscriptFile(filePath)
const analysis = buildAnalysis(lines, sid)
cache.set(sid, { analysis, lastModified: mtime })
return analysis
} catch (e) {
console.error('[transcript-engine] Error parsing transcript:', e)
return null
}
}
export function listSessions(): SessionInfo[] {
const projectDir = getProjectDir()
if (!existsSync(projectDir)) return []
try {
const files = readdirSync(projectDir)
.filter(f => f.endsWith('.jsonl'))
.map(f => {
const fullPath = join(projectDir, f)
const stat = statSync(fullPath)
return { name: f, path: fullPath, mtime: stat.mtimeMs }
})
.sort((a, b) => b.mtime - a.mtime)
return files.map(f => {
const sid = f.name.replace('.jsonl', '')
// Try cache first for quick metadata
const cached = cache.get(sid)
if (cached && cached.lastModified === f.mtime) {
return {
id: sid,
startTime: cached.analysis.startTime,
messageCount: cached.analysis.stats.messageCount,
model: cached.analysis.model
}
}
// Quick scan: read first few lines for metadata without full parse
try {
const content = readFileSync(f.path, 'utf8')
const firstLines = content.split('\n').slice(0, 20)
let startTime = ''
let model = ''
let lineCount = content.split('\n').filter(l => l.trim()).length
for (const line of firstLines) {
if (!line.trim()) continue
try {
const obj = JSON.parse(line)
if (obj.timestamp && !startTime) startTime = obj.timestamp
if (obj.type === 'assistant' && obj.message?.model && !model) {
model = obj.message.model
}
} catch { /* skip */ }
}
return {
id: sid,
startTime,
messageCount: lineCount,
model
}
} catch {
return { id: sid, startTime: '', messageCount: 0, model: '' }
}
})
} catch {
return []
}
}