upd

README.md

@@ -1,5 +1,5 @@
 <!-- status: canonical -->
-<!-- lastReviewed: 2026-04-25 -->
+<!-- lastReviewed: 2026-05-10 -->
 
 # PowerOn PORTA -- Dokumentation
 
@@ -59,6 +59,7 @@ b-reference/
 │   ├── workflow.md            Workflow-Engine, Methoden, Aktionen
 │   ├── automation.md          Automation v1 + v2
 │   ├── billing.md             Billing, Subscriptions
+│   ├── voice-google.md        Google STT/TTS (VoiceObjects, Streaming-WS, Feature-Mapping)
 │   └── features/              trustee.md, commcoach.md, chatbot.md, ...
 ├── frontend-nyla/          Frontend-Komponente
 │   └── architecture.md        Seiten, Komponenten, Hooks, Routing

TOPICS.md

@@ -1,5 +1,5 @@
 <!-- status: canonical -->
-<!-- lastReviewed: 2026-05-03 -->
+<!-- lastReviewed: 2026-05-11 -->
 
 # Themen-Index für AI-Kontext
 
@@ -24,10 +24,11 @@ Lade immer zuerst diese Datei. Dann gezielt die passende(n) Referenz-Datei(en).
 | Workflow-Engine | b-reference/gateway/workflow.md | Methoden, Aktionen, WorkflowManager |
 | Automation | b-reference/gateway/automation.md | Graphical Editor, Scheduler, System-Automatisierung (`/automations`, `/api/system/workflow-runs/*`) |
 | Billing & Subscriptions | b-reference/gateway/billing.md | Abrechnung, Prepaid, State Machine |
+| Google Voice (STT/TTS) | b-reference/gateway/voice-google.md | VoiceObjects, ConnectorGoogleSpeech, `/voice-google/stt/*`, CommCoach vs Teamsbot vs Agent |
 | Frontend Nyla | b-reference/frontend-nyla/architecture.md | UI-Seiten, Komponenten, Hooks, Routing |
 | FormGenerator (Table, Form, Tree, Report) | b-reference/frontend-nyla/formgenerator.md | Generische UI-Komponenten, Provider-Pattern, API-Anbindung |
 | Private LLM | b-reference/private-llm/architecture.md | Internes LLM, Neutralisierung |
-| Teams Bot | b-reference/teams-bot/architecture.md | Meeting-Bot, WebSocket, Director Prompts (Hybrid Agent-Routing) |
+| Teams Bot | b-reference/teams-bot/architecture.md | Meeting-Bot, Browser-Bot WebSocket, Director Prompts (Hybrid Agent-Routing), MeetingModule + 5-Tab-UI, Dashboard-SSE, Live-Session-SSE + MFA |
 
 ## Cross-Cutting (repo-übergreifend)
 

b-reference/frontend-nyla/architecture.md

@@ -1,6 +1,6 @@
 <!-- status: canonical -->
-<!-- lastReviewed: 2026-04-24 -->
-<!-- verifiedAgainst: frontend_nyla (codebase audit 2026-04-07, post Automation Unification) + Typed Action Architecture Phasen 1-5 + Vitest+RTL Setup -->
+<!-- lastReviewed: 2026-05-11 -->
+<!-- verifiedAgainst: frontend_nyla (codebase audit 2026-04-07, post Automation Unification) + Typed Action Architecture Phasen 1-5 + Vitest+RTL Setup + useVoiceStream STT open options (2026-05-10) + Teams Bot dashboard SSE + Module deep-link (2026-05-11) -->
 
 # Frontend Nyla -- Architektur
 
@@ -16,7 +16,7 @@ Technologie-Stack (Stand UI-Doku): React 19.x, Vite 5.x, TypeScript 5.8.x, React
 |--------|--------|
 | `pages/` | Seiten: `admin/`, `basedata/`, `billing/`, `settings/`, `views/` (workspace, commcoach, chatbot, trustee, graphicalEditor, realestate, neutralization, teamsbot) |
 | `components/` | Wiederverwendbar: FormGenerator (Table, Form, Tree, List, Report), Navigation, UnifiedDataBar, FlowEditor (Graphical Editor), OnboardingAssistant |
-| `hooks/` | useApiRequest (useApi.ts), useFiles, useNavigation, useConfirm, usePrompt, useResizablePanels, etc. |
+| `hooks/` | useApiRequest (useApi.ts), useFiles, useNavigation, useConfirm, usePrompt, useResizablePanels, `useSpeechAudioCapture` (`useVoiceStream`: Mic + WebSocket `/voice-google/stt/stream`, optionale STT-`open`-Parameter), etc. |
 | `contexts/` | FileContext, PekContext, ToastContext, WorkflowSelectionContext |
 | `api/` | API-Client (`api.ts`) und Feature-spezifische API-Module |
 | `core/` | PageManager |
@@ -43,6 +43,7 @@ Ergänzend typische Root-Dateien und Bereiche im Repo: `main.tsx`, `App.tsx`, `a
 | `TemplatePicker` | Modal zur Auswahl von Workflow-Vorlagen beim Erstellen neuer Workflows |
 | `WorkspacePage` | AI-Workspace mit Chat, UDB, Agent-Streaming |
 | `TrusteeDataTablesView` | Konsolidierte Daten-Tabellen-Seite (Treuhand): 13 Tabs (Stammdaten, Lokale Daten, Konfiguration, Buchhaltungs-Daten) ueber generischen `TrusteeDataTab`-Wrapper, Lazy-Mount pro Tab, URL-State `?tab=<key>`, read-only fuer Sync-Tabellen |
+| Teams Bot Views (`pages/views/teamsbot/`) | Dashboard (KPI + SSE), Assistent (Wizard), Module (CRUD + `?moduleId`), Live-Session (SSE + MFA), Settings — API `api/teamsbotApi.ts`; Architektur [teams-bot/architecture.md](../teams-bot/architecture.md) |
 
 ## Routing
 
@@ -50,6 +51,7 @@ Ergänzend typische Root-Dateien und Bereiche im Repo: `main.tsx`, `App.tsx`, `a
 - **Geschützte Bereiche:** Route-Guards (z. B. `ProtectedRoute`) prüfen Authentifizierung; Redirect bei fehlender Session.
 - **Haupt-App:** Nach Login **`MainLayout.tsx`** mit **`MandateNavigation`** (Sidebar) und `<Outlet />` (React Router).
 - **Seiten-Mapping:** `pageRegistry.tsx` definiert `PAGE_REGISTRY` und `FEATURE_REGISTRY`; Seiten werden lazy geladen. `core/PageManager/` enthält ergänzende Infrastruktur (State Preservation, Lifecycle-Hooks).
+- **Teams Bot:** `pages/views/teamsbot/` — fünf Tabs (Dashboard, Assistent, Module, Live-Session, Einstellungen). Dashboard: `teamsbotApi.createDashboardStream` (SSE). Live-Session: `createSessionStream`. Module-Liste: URL `?moduleId=<uuid>` klappt ein Modul auf. Detail: [teams-bot/architecture.md](../teams-bot/architecture.md).
 - **i18n:** DB-backed via `LanguageContext` (`t()`-Hook). Sprachsets werden dynamisch via public API geladen (`GET /api/i18n/sets/{code}`). Key-Konvention: **Deutscher Klartext = Key**. Jeder neue/geaenderte UI-Text MUSS mit `t('Deutscher Klartext')` getaggt werden. **`t()` darf NUR String-Literale enthalten** — `t(variable)` ist verboten. Werte vom Backend (Labels, Descriptions) sind bereits uebersetzt und werden direkt gerendert. Variable Interpolation: `t('Text {var}', {var: 'Wert'})`. Fallback: Ziel-Set → `de`-Set → `[key]` (eckige Klammern machen fehlende Uebersetzungen sichtbar). Keine statischen Locale-Files. Fuer statische Frontend-Maps (z.B. Wochentage, Monatsnamen, Status-Labels) wird `t()` per `switch`/Funktion mit Literalen aufgerufen, nicht ueber Map-Lookup.
 - **TextMultilingual:** Felder vom Typ `TextMultilingual` rendern dynamisch Eingabefelder fuer alle verfuegbaren Sprachen (aus `availableLanguages`). `xx` ist das Pflichtfeld (Quelltext). Keine hardcodierten Sprach-Codes. Sprachcodes folgen ISO 639-1 (en, de, fr, it, …).
 - **Theme:** global über Context; API-Requests mit Auth-Header (Interceptor in zentralem API-Client).
@@ -68,6 +70,7 @@ Ergänzend typische Root-Dateien und Bereiche im Repo: `main.tsx`, `App.tsx`, `a
 | `main.tsx` | Application Entry Point |
 | `App.tsx` | Root-Komponente, Router-Setup, Provider |
 | `api.ts` (Root) | Axios-Instanz, Base URL, Token-Interceptor, 401-Handling |
+| `hooks/useSpeechAudioCapture.ts` | `useVoiceStream`: STT-Streaming; `start(language, sttOpenOptions?)`; CommCoach setzt `model`/`lightweight`/`singleUtterance`, Workspace nutzt Defaults |
 | `layouts/MainLayout.tsx` | Haupt-Shell mit MandateNavigation + Outlet |
 | `config/pageRegistry.tsx` | PAGE_REGISTRY, FEATURE_REGISTRY, Lazy Imports |
 | `core/PageManager/` | State Preservation, Lifecycle, Caching |

b-reference/gateway/ai-agent.md

@@ -1,6 +1,6 @@
 <!-- status: canonical -->
-<!-- lastReviewed: 2026-04-28 (OpenAI temperature handling for GPT-5.x / o-series) -->
-<!-- verifiedAgainst: gateway (codebase audit 2026-04-07, post Automation Unification); gateway/modules/features/teamsbot/service.py (Hybrid Agent Escalation 2026-04-24); Typed Action Architecture Phasen 1-5; featureDataAgent domain hints hook 2026-04-27; central parameterValidation + DatabaseQueryError 2026-04-28; OpenAI temperature contract for GPT-5.x / o-series 2026-04-28 -->
+<!-- lastReviewed: 2026-05-10 -->
+<!-- verifiedAgainst: gateway (codebase audit 2026-04-07, post Automation Unification); gateway/modules/features/teamsbot/service.py (Hybrid Agent Escalation 2026-04-24); Typed Action Architecture Phasen 1-5; featureDataAgent domain hints hook 2026-04-27; central parameterValidation + DatabaseQueryError 2026-04-28; OpenAI temperature contract for GPT-5.x / o-series 2026-04-28; Voice STT speechToText params note 2026-05-10 -->
 
 # AI Agent & Knowledge Store
 
@@ -143,7 +143,7 @@ Zusätzlich zu den unten genannten **Kern-Tools** existieren **dynamische Tools*
 | `readUrl` | Inhalt einer bekannten URL laden |
 | `translateText` | Übersetzung (Voice/Translation-Pipeline) |
 | `textToSpeech` | TTS |
-| `speechToText` | Transkription Audio-Datei |
+| `speechToText` | Transkription Audio-Datei (Gateway: `VoiceObjects.speechToText`; optionale Connector-Parameter `model`/`lightweight`/`audioFormat` — Agent-Tool nutzt Defaults) |
 | `detectLanguage` | Spracherkennung für Text |
 
 **Externe Datenquellen / Mail**

b-reference/gateway/architecture.md

@@ -1,6 +1,6 @@
 <!-- status: canonical -->
-<!-- lastReviewed: 2026-04-28 -->
-<!-- verifiedAgainst: gateway (codebase audit 2026-04-07, post Automation Unification) + Typed Action Architecture Phasen 1-5 + Infomaniak-Connector (2026-04-28) + MSFT/Google Calendar+Contacts + Reconnect (2026-04-28) -->
+<!-- lastReviewed: 2026-05-10 -->
+<!-- verifiedAgainst: gateway (codebase audit 2026-04-07, post Automation Unification) + Typed Action Architecture Phasen 1-5 + Infomaniak-Connector (2026-04-28) + MSFT/Google Calendar+Contacts + Reconnect (2026-04-28) + Google Voice STT/TTS (2026-05-10) -->
 
 # Gateway -- Architektur
 
@@ -99,6 +99,13 @@ Weitere Interface-Dateien im Ordner (z. B. Voice, Tickets, Messaging, Bootstrap)
 | `gateway/modules/workflows/automation2/executionEngine.py` | Graph-Execution-Engine: topoSort, Transit-Routing, `_normalizeToSchema` nach Execute, `flow.merge`-Wait, Resume-Schema-Validierung |
 | `gateway/modules/workflows/scheduler/mainScheduler.py` | Konsolidierter Workflow-Scheduler |
 | `gateway/modules/interfaces/interfaceBootstrap.py` | System-Bootstrap (Templates, Billing, Stripe) |
+| `gateway/modules/interfaces/interfaceVoiceObjects.py` | Fassade Google STT/TTS, Billing-Callback Streaming |
+| `gateway/modules/connectors/connectorVoiceGoogle.py` | Google Speech v1 + Translation + TTS-Client |
+| `gateway/modules/routes/routeVoiceGoogle.py` | `/voice-google/*` inkl. STT-Streaming-WebSocket |
+
+## Google Voice (STT / TTS)
+
+Siehe Kanon-Seite **[voice-google.md](voice-google.md)** (Batch- vs Streaming-Pfad, WS-Protokoll `open`/`end_of_single_utterance`, Parameter `model` / `lightweight` / `audioFormat`, Zuordnung CommCoach / Teamsbot / Workspace / Agent-Tools).
 
 ## i18n (Mehrsprachigkeit)
 

b-reference/gateway/voice-google.md

@@ -0,0 +1,44 @@
+<!-- status: canonical -->
+<!-- lastReviewed: 2026-05-10 -->
+<!-- verifiedAgainst: gateway/modules/{connectors/connectorVoiceGoogle.py,interfaces/interfaceVoiceObjects.py,routes/routeVoiceGoogle.py}; gateway/modules/features/{commcoach/serviceCommcoach.py,teamsbot/service.py}; frontend_nyla/src/hooks/useSpeechAudioCapture.ts; frontend_nyla/src/pages/views/commcoach/useVoiceController.ts -->
+
+# Google Voice (STT / TTS)
+
+Zentrale Sprach-Pipeline im Gateway: **Google Cloud Speech-to-Text (v1)** und **Text-to-Speech** über die Fassade `VoiceObjects` (`interfaceVoiceObjects.py`) und den Connector `ConnectorGoogleSpeech` (`connectorVoiceGoogle.py`). Kein direkter Feature-Code soll den Connector umgehen; Billing-Hooks hängen an der Interface-Schicht (u. a. Streaming-STT bei finalen Segmenten).
+
+## Batch-Erkennung (`speechToText`)
+
+- **API:** `SpeechClient.recognize` (ein Request pro Audioclip).
+- **Parameter (optional, abwaertskompatibel):**
+  - **`model`:** z. B. `latest_long` (Default), `latest_short`.
+  - **`lightweight`:** `True` schaltet u. a. ab: Word-Timestamps, Word-Confidence, `max_alternatives>1`, `use_enhanced`. Ziel: geringere Latenz/Kosten bei Kurzäusserungen.
+  - **`audioFormat`:** z. B. `webm_opus`, `linear16`, `mp3`, `flac`, `wav` — wenn gesetzt, entfällt `validateAudioFormat()` (Fast-Path); fehlende Sample-Rate wird formattypisch defaulted (z. B. `webm_opus` → 48 kHz).
+- **Fallbacks:** Bei leeren Ergebnissen (wenn nicht `skipFallbacks`) weiterhin Encoding-/Modell-Matrix; erstes Fallback-Modell ist der vom Caller gewählte `model`-Wert.
+
+## Streaming-Erkennung
+
+- **Route:** `POST /voice-google/stt/token` (kurzlebiges `wsToken`), WebSocket **`/voice-google/stt/stream?wsToken=...`** (`routeVoiceGoogle.py`).
+- **Client-Protokoll (JSON):**
+  - `open`: `language`, optional `model`, `lightweight`, `singleUtterance` (Booleans).
+  - `audio`: `chunk` (Base64), typisch WebM/Opus-Fragmente vom `MediaRecorder`.
+  - `close` / `ping` / `pong` unveraendert.
+- **Server-Antworten:** `interim`, `final`, optional **`end_of_single_utterance`** (mit `audioDurationSec`) wenn Google Single-Utterance-Modus endet — Client sendet daraufhin erneut `open` mit denselben STT-Optionen (Recorder laeuft weiter).
+- **Connector:** `streamingRecognize` — `StreamingRecognitionConfig.single_utterance` steuerbar; **Abrechnungs-Dauer** fuer Finals bevorzugt aus `result_end_time` der Streaming-Results (Fallback: Byte-Heuristik).
+
+## Feature-Zuordnung (Ist-Stand)
+
+| Bereich | STT-Pfad | Konfiguration (Kurz) |
+|---------|----------|----------------------|
+| **CommCoach Live** | Streaming WS → Text → Coach | Frontend (`useVoiceStream`): `latest_short`, `lightweight`, `singleUtterance: true` (`useVoiceController`). |
+| **CommCoach Batch** | `processAudioMessage` → `speechToText` | `latest_short`, `lightweight: true` (Format weiterhin Auto-Detect wenn kein `audioFormat`). |
+| **Workspace** | Gleicher generischer WS | `start(language)` ohne Extra-Optionen → Defaults wie frueher (`latest_long`, kein Single-Utterance). |
+| **Teamsbot** | Pro Audio-Chunk `speechToText` (Batch) | Gemeinsamer Connector; **`audioFormat: linear16`** (bekannte PCM-Chunks); Modell/Default wie bisher `latest_long` + volle Features (`lightweight: false`). |
+| **Agent-Tool** `_mediaTools.speechToText` | Batch | Keine zusaetzlichen Argumente → Default-Verhalten. |
+
+## TTS
+
+Unveraendert ueber `textToSpeech` / Gemini-TTS-Pfad im Connector; diese Seite fokussiert STT.
+
+## Weiterfuehrung
+
+- Evaluation **Speech-to-Text v2 / Chirp:** [`wiki/c-work/1-plan/stt-chirp-v2-evaluation.md`](../../c-work/1-plan/stt-chirp-v2-evaluation.md).

b-reference/teams-bot/architecture.md

@@ -1,6 +1,6 @@
 <!-- status: canonical -->
-<!-- lastReviewed: 2026-04-24 -->
-<!-- verifiedAgainst: service-teams-browser-bot (documentation review 2026-02-18); gateway/modules/features/teamsbot/{service,routeFeatureTeamsbot,interfaceFeatureTeamsbot,datamodelTeamsbot}.py + gateway/tests/unit/teamsbot/test_directorPrompts.py (Director Prompts review 2026-04-24) -->
+<!-- lastReviewed: 2026-05-11 -->
+<!-- verifiedAgainst: service-teams-browser-bot (documentation review 2026-02-18); gateway/modules/features/teamsbot/{service,routeFeatureTeamsbot,interfaceFeatureTeamsbot,datamodelTeamsbot}.py + `GET /api/teamsbot/{instanceId}/dashboard/stream` (2026-05-11); gateway/tests/unit/teamsbot/test_directorPrompts.py (Director Prompts 2026-04-24); gateway Voice STT batch + linear16 (2026-05-10); frontend_nyla TeamsbotDashboardView / TeamsbotModulesView / TeamsbotSessionView (IA + SSE 2026-05-11) -->
 
 # Teams Meeting Bot -- Architektur
 
@@ -22,7 +22,21 @@ AI-gesteuerter Meeting-Bot für Microsoft Teams. Tritt Meetings als regulärer T
 |------------|----------|-------|
 | Gateway ↔ Bot | WebSocket | Echtzeit: Transkripte, Chat, Audio, Status |
 | Gateway → Bot | HTTP | Session-Steuerung (join, leave, status) |
-| Frontend ← Gateway | SSE | Live-Transkript-Stream für UI |
+| Frontend ← Gateway | SSE | Live-Transkript-Stream für UI; Dashboard-Push (`dashboard/stream`) |
+
+## Nyla UI (MeetingModule-IA)
+
+Die Feature-Oberfläche ist in **fünf Tabs** strukturiert (siehe auch `wiki/c-work/2-build/2026-04-teamsbot-greenfield-ia-and-live-update.md`):
+
+| Tab | Route-Segment | Zweck |
+|-----|----------------|--------|
+| Dashboard | `dashboard` | KPIs, Modul-Aktivität, Quick-Actions; Daten per **SSE** `GET /api/teamsbot/{instanceId}/dashboard/stream` (Snapshots `sessions` + `modules`, Intervall 3 s bei aktiver eigener Sitzung, sonst 20 s) |
+| Assistent | `assistant` | Wizard: Modul wählen/anlegen, Meeting-Link, Join-Modus (Systembot / Gast / Mein Account), optional Sitzungskontext → startet Session und navigiert zur Live-Ansicht |
+| Module | `modules` | CRUD `TeamsbotMeetingModule`, aufklappbare Session-Liste pro Modul; **Deep-Link** `?moduleId=<uuid>` klappt das Modul auf und scrollt es ins Sichtfeld |
+| Live-Session | `sessions` | Regie-Panel, UDB, Transkript, SSE `sessions/{id}/stream`; **MFA** (`mfaChallenge` / `mfaResolved`) wird hier über dieselbe SSE-Verbindung abgewickelt |
+| Einstellungen | `settings` | Bot-Stimme, User-Settings, System-Bots (SysAdmin) |
+
+**Datenmodell:** `TeamsbotMeetingModule` gruppiert Reihen; `TeamsbotSession.moduleId` verknüpft optional (Adhoc ohne Modul). Standard-Meeting-Link und Standard-Bot-Name pro Modul für Prefill im Assistenten.
 
 ## Kernfähigkeiten
 
@@ -50,12 +64,15 @@ Der Gateway (Feature `teamsbot`) verwaltet Sessions und stellt die AI-Pipeline b
 - WebSocket-Verbindung pro Session
 - AI-Analyse der Transkript-Segmente via `serviceAi`
 - TTS-Generierung für Voice-Responses
+- **Dashboard-SSE:** `GET /api/teamsbot/{instanceId}/dashboard/stream` — wiederholte JSON-Events `{ "type": "dashboardState", "sessions": [...], "modules": [...] }` mit gleicher Sichtbarkeit wie `GET /sessions` (eigene Sessions, ausser Platform-Admin sieht alle).
 
 ## Schlüssel-Dateien
 
 | Datei / Bereich | Rolle |
 |-----------------|-------|
-| `gateway/modules/features/teamsbot/` | Gateway-seitiges Feature-Modul |
+| `gateway/modules/features/teamsbot/` | Gateway-seitiges Feature-Modul (inkl. `dashboard/stream`, Session-SSE, Module-CRUD) |
+| `frontend_nyla/src/pages/views/teamsbot/` | Dashboard (SSE), Assistent, Module, Live-Session, Einstellungen |
+| `frontend_nyla/src/api/teamsbotApi.ts` | u.a. `createDashboardStream`, `createSessionStream`, Module-API |
 | `service-teams-browser-bot/` | Eigenständiger Bot-Service (separates Repository) |
 
 ## Regeln / Invarianten
@@ -65,6 +82,8 @@ Der Gateway (Feature `teamsbot`) verwaltet Sessions und stellt die AI-Pipeline b
 - Authentifizierter Join (mit Microsoft-Account) oder Anonymous Guest -- je nach Konfiguration
 - Gateway ist die **einzige** Schnittstelle für AI-Aufrufe und TTS -- der Bot-Service selbst hat keine AI-Logik
 
+**STT auf dem Gateway:** Meeting-Audio-Chunks (WebSocket `audioChunk`, PCM) werden pro Chunk mit `VoiceObjects.speechToText` transkribiert (Batch `recognize`, gemeinsamer Connector mit CommCoach). Konfiguration u. a. `audioFormat=linear16`, `skipFallbacks=True`; Details und Modell-Defaults: [voice-google.md](../gateway/voice-google.md).
+
 ## Hybrid-Routing: SPEECH_TEAMS + Agent
 
 Der Teamsbot läuft auf zwei kooperierenden Pfaden:

c-work/1-plan/stt-chirp-v2-evaluation.md

@@ -0,0 +1,22 @@
+# STT: Google Speech-to-Text v2 / Chirp Evaluation (follow-up)
+
+Status: planned (not implemented). Related: gateway `connectorVoiceGoogle.py` uses Speech v1 `SpeechClient` only.
+
+## Goal
+
+Benchmark STT v2 (e.g. Chirp / Chirp 2) for `de-DE` vs current v1 `latest_short` / `latest_long` on:
+
+- Latency (time-to-first-token, final latency)
+- WER / subjective quality in meeting + coaching scenarios
+- Cost and quota
+
+## Steps
+
+1. Add optional v2 client path (`google.cloud.speech_v2` or REST) behind a feature flag.
+2. Run A/B on CommCoach streaming and Teamsbot batch paths with identical audio fixtures.
+3. Document decision in `wiki/b-reference/` and remove flag or make v2 default.
+
+## Notes
+
+- Streaming and batch config differ between v1 and v2; keep `VoiceObjects` as the single facade.
+- Billing hooks (`calculateSttCostCHF`) must use measured duration (see streaming `result_end_time`), not compressed byte heuristics.

c-work/_CHANGELOG.md

@@ -12,8 +12,16 @@ type: `feat` `fix` `refactor` `docs` `test` `chore` `build`  ·  scope: `gateway
 
 Skip: reine Refactors, Formatting, Lint, Dep-Bumps, Test-only, Wiki-Tippfehler.
 
+## 2026-05-11
+
+- 2026-05-11 | feat | gateway, frontend-nyla, wiki | TeamsBot: `GET /api/teamsbot/{instanceId}/dashboard/stream` (SSE dashboardState 3s/20s); dashboard consumes EventSource + reconnect; module tiles navigate to `modules?moduleId=` with expand + scroll highlight; canonical `b-reference/teams-bot/architecture.md`, TOPICS, `b-reference/frontend-nyla/architecture.md` (Teams Bot UI row). (c-work: 2-build/2026-04-teamsbot-greenfield-ia-and-live-update.md)
+
 ## 2026-05-10
 
+- 2026-05-10 | feat | frontend-nyla | TeamsBot UI aligned to greenfield IA: dashboard as KPI + module-activity + quick actions (no duplicate start form); MFA on live session SSE; assistant step adds join mode, MS account block, session context; modules header `Modul anlegen` dialog + `Meeting starten`; nav icons for assistant/modules; `FEATURE_REGISTRY` tab label Dashboard. (c-work: 2-build/2026-04-teamsbot-greenfield-ia-and-live-update.md)
+- 2026-05-10 | feat, fix | gateway, frontend-nyla, teams-bot | TeamsBot: `startSession` binds `moduleId` with validation; DB migration M2 `TeamsbotMeetingModule.defaultMeetingLink` / `defaultBotName`; assistant prefill + searchable multi-row module list; modules edit saves defaults; session view polls while pending/joining even with SSE and syncs switcher row; browser bot chat compose marker + safe `AudioContext.close`. (c-work: 2-build/2026-04-teamsbot-greenfield-ia-and-live-update.md)
+- 2026-05-10 | docs | wiki | Kanon: `b-reference/gateway/voice-google.md`; TOPICS + README Ordnerbaum; Querverweise in gateway/architecture, teams-bot/architecture, frontend-nyla/architecture, ai-agent.
+- 2026-05-10 | feat | gateway, frontend-nyla | STT: configurable model/lightweight/audioFormat on batch `speechToText`; streaming supports model/lightweight/singleUtterance; duration from `result_end_time`; WS `open` + `end_of_single_utterance` replumbing. CommCoach uses `latest_short` + lightweight + single-utterance; Teamsbot passes `audioFormat=linear16`. Chirp/v2 evaluation tracked in `wiki/c-work/1-plan/stt-chirp-v2-evaluation.md`.
 - 2026-05-10 | fix | gateway | CommCoach: conversational turns use `operationType=DATA_QUERY` + `priority=SPEED` for faster model selection (gpt-5.4-mini instead of gpt-5.5); model selector speed priority weight scaled from `/10` to `*100` to be meaningful.
 - 2026-05-10 | fix | gateway | CommCoach: persona sessions now distinguish first vs follow-up; opening prompt references prior sessions instead of re-introducing persona every time.
 - 2026-05-10 | feat | gateway | Document generation: unified `documentTitle` style separate from `headings.h1`; BASE/renderer converts independently; HTML/CSS, PDF title block, Markdown/Text heading offsets updated; DEFAULT_STYLE h1 lowered vs title.