feat(gpu-server): add API key auth, VRAM management, and Piper TTS voices

- Add API key authentication to all GPU services (X-API-Key header)
  - /health and /docs remain public (no key needed)
  - Shared key configured via GPU_API_KEY env variable
- Add VRAM auto-unload for mana-image-gen (5min) and mana-stt (10min)
  - FLUX.2 pipeline freed after idle, recovering ~13GB VRAM
  - WhisperX models freed after idle, recovering ~3GB VRAM
- Install Piper TTS voices (Thorsten + Kerstin) for local German TTS
- Update @manacore/shared-gpu client to support apiKey parameter
- Add GPU_API_KEY to .env.development
- Document API auth and VRAM management in setup guide

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Till JS 2026-03-27 21:54:35 +01:00
parent 97ef728eca
commit c67ed0df14
7 changed files with 65 additions and 6 deletions

View file

@ -417,3 +417,8 @@ MUKKE_DATABASE_URL=postgresql://manacore:devpassword@localhost:5432/mukke
CITYCORNERS_BACKEND_PORT=3025
CITYCORNERS_DATABASE_URL=postgresql://manacore:devpassword@localhost:5432/citycorners
CITYCORNERS_WEB_PORT=5196
# GPU Server (Windows PC with RTX 3090)
GPU_API_KEY=sk-gpu-cf483ede1e05e28fba5e56c94cd3c24e7c245e57816d3e86
GPU_SERVER_URL=https://gpu.mana.how
GPU_SERVER_LAN_URL=http://192.168.178.11

View file

@ -614,13 +614,19 @@ GPU Server (healthcheck.py → log-shipper.py)
Shared Package im Monorepo (`packages/shared-gpu/`) für alle GPU-Services:
```typescript
import { GpuClient, GPU_PUBLIC_URLS } from '@manacore/shared-gpu';
import { GpuClient } from '@manacore/shared-gpu';
// Öffentlich (von überall)
const gpu = new GpuClient({ baseUrl: 'https://gpu.mana.how' });
// Öffentlich (von überall, mit API-Key)
const gpu = new GpuClient({
baseUrl: 'https://gpu.mana.how',
apiKey: process.env.GPU_API_KEY,
});
// Oder LAN (direkt, schneller)
const gpuLan = new GpuClient({ baseUrl: 'http://192.168.178.11' });
const gpuLan = new GpuClient({
baseUrl: 'http://192.168.178.11',
apiKey: process.env.GPU_API_KEY,
});
// Speech-to-Text (mit Word-Timestamps + Speaker Diarization)
const transcript = await gpu.stt.transcribe(audioBuffer, 'recording.wav', {
@ -644,6 +650,37 @@ const health = await gpu.healthCheck();
---
## API-Authentifizierung
Alle GPU-Services erfordern einen API-Key für Zugriff auf geschützte Endpoints.
`/health` und `/docs` sind öffentlich (kein Key nötig).
**API-Key:** In `.env.development` unter `GPU_API_KEY`
**Verwendung:**
```bash
# Mit Header
curl -H "X-API-Key: $GPU_API_KEY" https://gpu-llm.mana.how/v1/models
# Oder als Query-Parameter
curl "https://gpu-stt.mana.how/models?api_key=$GPU_API_KEY"
# Health (kein Key nötig)
curl https://gpu-llm.mana.how/health
```
**Konfiguration auf dem GPU-Server:**
| Service | Env-Variable | Datei |
|---|---|---|
| mana-llm | `GPU_API_KEY` | `C:\mana\services\mana-llm\.env` |
| mana-stt | `API_KEYS`, `INTERNAL_API_KEY` | `C:\mana\services\mana-stt\.env` |
| mana-tts | `API_KEYS`, `INTERNAL_API_KEY` | `C:\mana\services\mana-tts\.env` |
| mana-image-gen | `GPU_API_KEY` | `C:\mana\services\mana-image-gen\.env` |
---
## Fehlerbehebung
### Server nicht erreichbar (kein Ping, kein SSH)

View file

@ -28,8 +28,10 @@ export class GpuClient {
public readonly stt: SttClient;
public readonly tts: TtsClient;
public readonly image: ImageClient;
public readonly apiKey?: string;
constructor(config: GpuServiceConfig) {
this.apiKey = config.apiKey;
this.stt = new SttClient(config);
this.tts = new TtsClient(config);
this.image = new ImageClient(config);

View file

@ -9,10 +9,12 @@ import { resolveServiceUrl } from './resolve-url';
export class ImageClient {
private baseUrl: string;
private timeout: number;
private apiKey?: string;
constructor(config: GpuServiceConfig) {
this.baseUrl = resolveServiceUrl(config, 'image');
this.timeout = config.timeout ?? 120_000;
this.apiKey = config.apiKey;
}
/** Generate an image from a text prompt. */
@ -23,7 +25,10 @@ export class ImageClient {
try {
const response = await fetch(`${this.baseUrl}/generate`, {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
headers: {
'Content-Type': 'application/json',
...(this.apiKey ? { 'X-API-Key': this.apiKey } : {}),
},
body: JSON.stringify({
prompt: options.prompt,
width: options.width ?? 1024,

View file

@ -4,10 +4,12 @@ import { resolveServiceUrl } from './resolve-url';
export class SttClient {
private baseUrl: string;
private timeout: number;
private apiKey?: string;
constructor(config: GpuServiceConfig) {
this.baseUrl = resolveServiceUrl(config, 'stt');
this.timeout = config.timeout ?? 60_000;
this.apiKey = config.apiKey;
}
/** Transcribe audio with optional word timestamps and speaker diarization. */
@ -34,6 +36,7 @@ export class SttClient {
try {
const response = await fetch(`${this.baseUrl}/transcribe`, {
method: 'POST',
headers: this.apiKey ? { 'X-API-Key': this.apiKey } : {},
body: formData,
signal: controller.signal,
});

View file

@ -4,10 +4,12 @@ import { resolveServiceUrl } from './resolve-url';
export class TtsClient {
private baseUrl: string;
private timeout: number;
private apiKey?: string;
constructor(config: GpuServiceConfig) {
this.baseUrl = resolveServiceUrl(config, 'tts');
this.timeout = config.timeout ?? 30_000;
this.apiKey = config.apiKey;
}
/** Synthesize speech. Returns audio as ArrayBuffer. */
@ -23,7 +25,10 @@ export class TtsClient {
try {
const response = await fetch(`${this.baseUrl}/synthesize/auto`, {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
headers: {
'Content-Type': 'application/json',
...(this.apiKey ? { 'X-API-Key': this.apiKey } : {}),
},
body: JSON.stringify({
text: options.text,
voice: options.voice,

View file

@ -119,6 +119,8 @@ export interface GpuServiceConfig {
image?: string;
ollama?: string;
};
/** API key for authenticated access (X-API-Key header) */
apiKey?: string;
/** Request timeout in ms (default: 30000) */
timeout?: number;
}