feat(nvidia): 5 skills NVIDIA + helper CLI para build.nvidia.com

Instala 5 skills baseadas no pacote NVIDIA/skills (filtradas para M5):

- nvidia-nemotron-speech    (169 linhas)  ASR/TTS/NMT cloud Riva NIM
- nvidia-nemo-retriever     (243 linhas)  RAG multi-formato com NV-Embed-v2
- nvidia-data-designer      (184 linhas)  Dados sinteticos para popular motor
- nvidia-aiq-deploy         (194 linhas)  Agente autonomo multi-step
- nvidia-policy-generator   (241 linhas)  Moderacao de conteudo + LGPD

Total: 1031 linhas de documentacao pratica com:
- Endpoints curl prontos
- Codigo Python helper
- Casos de uso M5 especificos (OpenWA→CRM, Saleor, Twenty, Chatwoot)
- Schema PostgreSQL (pgvector) para RAG
- YAML de configuracao para AI-Q
- Limites e custos do free tier

skills/ nao commitada (sua em ~/.hermes/skills/) — ja' instaladas e ativas.

Adiciona scripts/nvidia-apis.py (CLI wrapper):
- chat PROMPT          chat completion qualquer modelo
- models               lista 121 modelos free
- status               verifica se NVIDIA_API_KEY funciona

Acao para amanha:
1. Obter NVIDIA_API_KEY gratis em https://build.nvidia.com/
2. echo "NVIDIA_API_KEY=nvapi-..." >> /root/ecommerce/scripts/.env
3. python3 scripts/nvidia-apis.py status
4. Habilitar pgvector: docker exec m5_postgres psql ... -c 'CREATE EXTENSION vector;'
5. Testar skill nvidia-nemotron-speech (transcrever audio WhatsApp)

scripts/nvidia-apis.py

@@ -0,0 +1,119 @@
+#!/usr/bin/env python3
+"""
+NVIDIA APIs Helper — wrapper unificado para build.nvidia.com
+Carrega NVIDIA_API_KEY de scripts/.env
+"""
+import os
+import sys
+import json
+import argparse
+import urllib.request
+import urllib.error
+from pathlib import Path
+
+ENV_FILE = Path(__file__).parent / ".env"
+if ENV_FILE.exists():
+    for line in ENV_FILE.read_text().splitlines():
+        line = line.strip()
+        if not line or line.startswith("#"):
+            continue
+        if "=" in line:
+            k, v = line.split("=", 1)
+            v = v.strip().strip('"').strip("'")
+            os.environ.setdefault(k, v)
+
+API_KEY = os.environ.get("NVIDIA_API_KEY", "")
+BASE = "https://integrate.api.nvidia.com/v1"
+
+
+def call_chat(model: str, prompt: str, max_tokens: int = 200, json_mode: bool = False) -> dict:
+    payload = {
+        "model": model,
+        "messages": [{"role": "user", "content": prompt}],
+        "max_tokens": max_tokens,
+    }
+    if json_mode:
+        payload["response_format"] = {"type": "json_object"}
+
+    req = urllib.request.Request(
+        f"{BASE}/chat/completions",
+        data=json.dumps(payload).encode(),
+        headers={
+            "Content-Type": "application/json",
+            "Authorization": f"Bearer {API_KEY}",
+        },
+    )
+    try:
+        with urllib.request.urlopen(req, timeout=60) as resp:
+            return json.loads(resp.read().decode())
+    except urllib.error.HTTPError as e:
+        raw = e.read().decode()
+        try:
+            return json.loads(raw)
+        except Exception:
+            return {"_error": e.code, "_raw": raw[:500]}
+
+
+def cmd_chat(args):
+    if not API_KEY:
+        print("ERRO: NVIDIA_API_KEY nao definida em scripts/.env", file=sys.stderr)
+        print("Obtenha gratis em: https://build.nvidia.com/", file=sys.stderr)
+        sys.exit(1)
+    r = call_chat(args.model, args.prompt, args.max_tokens, args.json)
+    if "_error" in r:
+        print(f"ERRO {r['_error']}: {r['_raw']}")
+        sys.exit(1)
+    print(r["choices"][0]["message"]["content"])
+
+
+def cmd_models(args):
+    """Lista modelos disponiveis (sem auth)."""
+    req = urllib.request.Request(f"{BASE}/models")
+    try:
+        with urllib.request.urlopen(req, timeout=30) as resp:
+            data = json.loads(resp.read().decode())
+        models = [m["id"] for m in data.get("data", [])]
+        print(f"{len(models)} modelos:")
+        for m in sorted(models):
+            if any(k in m.lower() for k in ["nemotron", "parakeet", "magpie", "nv-embed", "cosmos", "deepseek", "kimi", "mistral", "llama"]):
+                print(f"  - {m}")
+    except urllib.error.HTTPError as e:
+        print(f"HTTP {e.code}: {e.read().decode()[:300]}")
+
+
+def cmd_status(args):
+    """Verifica se a key funciona."""
+    if not API_KEY:
+        print("NVIDIA_API_KEY: NAO configurada")
+        print("Acao: obter gratis em https://build.nvidia.com/")
+        return
+    print(f"NVIDIA_API_KEY: configurada (prefixo: {API_KEY[:12]}...)")
+    r = call_chat("nvidia/nemotron-mini-4b-instruct", "Diga 'ok' em portugues", 20)
+    if "_error" in r:
+        print(f"ERRO {r['_error']}: {r['_raw'][:200]}")
+    else:
+        print(f"Status: OK")
+        print(f"Teste: {r['choices'][0]['message']['content']}")
+
+
+def main():
+    parser = argparse.ArgumentParser(description="NVIDIA APIs Helper (build.nvidia.com)")
+    sub = parser.add_subparsers(dest="cmd", required=True)
+
+    p = sub.add_parser("chat", help="chat completion")
+    p.add_argument("prompt")
+    p.add_argument("--model", default="nvidia/nemotron-mini-4b-instruct")
+    p.add_argument("--max-tokens", type=int, default=200)
+    p.add_argument("--json", action="store_true", help="force JSON output")
+    p.set_defaults(func=cmd_chat)
+
+    sub.add_parser("models", help="lista modelos disponiveis").set_defaults(func=cmd_models)
+
+    sub.add_parser("status", help="verifica se key funciona").set_defaults(func=cmd_status)
+
+    args = parser.parse_args()
+    args.func(args)
+
+
+if __name__ == "__main__":
+    main()