Ai architecture
Skill henriquescastilho/my-claude/.claude/skills/ai-architecture
Arquitetura de plataformas AI, integracao de modelos, MLOps, edge AI. Usar quando o projeto envolve IA, modelos de linguagem, Gemma, fine-tuning, ou inferencia.From its SKILL.md
npx -y skills add henriquescastilho/my-claude --skill ai-architectureAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
1.5 KB, 389 tokens by cl100k_base, as published. Nobody here has run it
Arquitetura AI -- Padrao DME
Stack AI da DME
- Modelo local: Gemma 4 (no computador do Henrique)
- API: Claude API (Anthropic SDK)
- Embeddings: OpenAI ou local
- Vector DB: pgvector (Supabase) ou Pinecone
- Framework: LangChain ou direto com SDK
Padroes de integracao
Claude API (principal)
import Anthropic from '@anthropic-ai/sdk';
const client = new Anthropic();
const message = await client.messages.create({
model: "claude-sonnet-4-6",
max_tokens: 1024,
messages: [{ role: "user", content: prompt }]
});
Gemma 4 (local, gratuito)
Usar para pre-processamento, resumo, classificacao -- tarefas que nao precisam do Claude. Rodar via Ollama ou direto no Python.
Decisoes de arquitetura AI
| Tarefa | Modelo | Por que |
|---|---|---|
| Resumo de documentos | Gemma 4 (local) | Gratuito, rapido |
| Geracao de codigo | Claude Sonnet | Qualidade + velocidade |
| Decisoes complexas | Claude Opus | Raciocinio profundo |
| Embeddings | text-embedding-3-small | Barato, bom o suficiente |
| Classificacao simples | Gemma 4 (local) | Zero custo |
Seguranca AI
- Nunca enviar PII para APIs externas sem anonimizar
- Rate limiting em endpoints AI
- Timeout em chamadas de inferencia (30s max)
- Fallback quando modelo esta indisponivel
- Logging de prompts sem dados sensiveis
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.