K-10 pipeline transcripciones + settings IA + cache-aside + session doc

Nuevos modulos:
- services/ai.ts: cliente IA provider-agnostico (OpenRouter, MiniMax)
- services/db.ts: Dexie core con tabla settings + project_docs
- services/storage.ts: Cache-Aside + Write-Through (L1 Map → L2 Dexie → L3 localStorage)
- services/parse-transcription.ts: parser .docx/.vtt/.txt/.md
- services/session-analyzer.ts: extraccion IA de sesiones (resumen, tareas, decisiones)
- services/project-doc.ts: documento maestro MD (Bloque 1 resumen + Bloque 2 sesiones)
- stores/settings.ts: proveedores IA, modelos, API keys separadas por provider
- stores/transcriptions.ts: pipeline upload → analyze → create HU en KAPPA
- views/SettingsView.vue: configuracion IA (OpenRouter, MiniMax, OpenCode bridge)
- views/TranscriptionsView.vue: subida multiple + analisis sesion + visor MD + calendario
- components/AiProjectChat.vue: chat contextual por proyecto con selector de modelo

Cambios en existentes:
- stores/auth.ts, kappa-api.ts, upload-hu.ts: migrados a storage service (Dexie + localStorage)
- stores/projects.ts, workitems.ts: kappa_last_project via storage
- DashboardView.vue: descripcion reemplazada por AiProjectChat
- NewDashboardView.vue: tabs transcriptions + settings + navigate-settings events
- NavMain.vue: items Transcripciones + Configuracion
- SiteHeader.vue: labels tabs + language via storage
- LoginView.vue: remember_email via storage
- i18n: +80 keys español/ingles
- vite.config.ts: proxy CORS para MiniMax
- package.json: +mammoth.js

src/services/parse-transcription.ts

@@ -0,0 +1,69 @@
+import * as mammoth from 'mammoth'
+
+export type TranscriptionFileType = 'docx' | 'vtt' | 'txt' | 'md'
+
+export interface ParsedTranscription {
+  fileName: string
+  fileType: TranscriptionFileType
+  text: string
+  size: number
+}
+
+const VTT_HEADER_RE = /^WEBVTT\s/mi
+const VTT_TIMING_RE = /^\d{2}:\d{2}:\d{2}\.\d{3}\s*-->\s*\d{2}:\d{2}:\d{2}\.\d{3}/m
+const VTT_CUE_RE = /\d{2}:\d{2}:\d{2}\.\d{3}\s*-->\s*\d{2}:\d{2}:\d{2}\.\d{3}/g
+
+function isVTT(text: string): boolean {
+  return VTT_HEADER_RE.test(text) || VTT_TIMING_RE.test(text)
+}
+
+function parseVtt(raw: string): string {
+  return raw
+    .replace(VTT_HEADER_RE, '')
+    .replace(/Kind:.*\n?/gi, '')
+    .replace(/Language:.*\n?/gi, '')
+    .split(/\r?\n/)
+    .filter(line => {
+      const trimmed = line.trim()
+      if (!trimmed) return false
+      if (/^\d+$/.test(trimmed)) return false
+      if (VTT_TIMING_RE.test(trimmed)) return false
+      return true
+    })
+    .join(' ')
+    .replace(/\s+/g, ' ')
+    .trim()
+}
+
+function detectType(fileName: string): TranscriptionFileType {
+  const ext = fileName.split('.').pop()?.toLowerCase()
+  if (ext === 'docx') return 'docx'
+  if (ext === 'vtt') return 'vtt'
+  if (ext === 'md') return 'md'
+  return 'txt'
+}
+
+export async function parseFile(file: File): Promise<ParsedTranscription> {
+  const fileName = file.name
+  const fileType = detectType(fileName)
+  const size = file.size
+
+  console.log(`[Alpha] Parsing file: ${fileName} (${fileType}, ${size} bytes)`)
+
+  let text: string
+
+  if (fileType === 'docx') {
+    const arrayBuffer = await file.arrayBuffer()
+    const result = await mammoth.extractRawText({ arrayBuffer })
+    text = result.value.trim()
+  } else {
+    const raw = await file.text()
+    if (fileType === 'vtt' || isVTT(raw)) {
+      text = parseVtt(raw)
+    } else {
+      text = raw.trim()
+    }
+  }
+
+  return { fileName, fileType, text, size }
+}