| name | AI Service Expert |
| description | Integrate and manage AI services (Grok, OpenAI, Claude, Gemini).
Trigger keywords: ai service, grok, openai, claude, llm, fallback, streaming
Not for: AI architecture decisions (-> ai-architecture-layering), AI Teams (-> ai-teams-expert)
|
| allowed-tools | ["Bash","Read","Write","Edit","Grep","Glob"] |
| tags | ["ai","grok","openai","llm","claude","gemini"] |
| boundaries | {"includes":["AI provider client configuration","Fallback and retry logic","Streaming implementation","Error handling patterns"],"excludes":["Architecture layer decisions","Multi-agent collaboration"],"handoff":[{"skill":"ai-architecture-layering","when":"Deciding capability placement"},{"skill":"ai-teams-expert","when":"Multi-agent implementation"}]} |
AI Service Expert
Integrate and manage AI services for gens.team.
Architecture
┌─────────────────────────────────────────────────────────┐
│ AI Orchestrator │
│ (NestJS Backend) │
├─────────────────────────────────────────────────────────┤
│ Primary: Grok API (x.AI) - Speed optimized │
│ Fallback: OpenAI → Claude → Gemini → DeepSeek │
└─────────────────────────────────────────────────────────┘
│ │ │
┌───▼───┐ ┌─────▼─────┐ ┌─────▼─────┐
│ Grok │ │ OpenAI │ │ Claude │
│ (x.AI)│ │ GPT-4o │ │ Sonnet │
└───────┘ └───────────┘ └───────────┘
Orchestration Pattern
@Injectable()
export class AIOrchestrator {
private providers = ["grok", "openai", "claude", "gemini"];
async complete(request: AIRequest): Promise<AIResponse> {
for (const provider of this.providers) {
try {
const result = await this.callProvider(provider, request);
return { ...result, provider };
} catch (error) {
this.logger.warn(`Provider ${provider} failed, trying next...`);
continue;
}
}
throw new ServiceUnavailableException("All AI providers failed");
}
}
Error Handling
class RateLimitError extends AIServiceError {}
class TokenLimitError extends AIServiceError {}
async function safeCompletion(messages: Message[]): Promise<Response> {
try {
return await client.chat(messages);
} catch (e) {
if (e.status === 429) throw new RateLimitError();
if (e.status === 400) throw new TokenLimitError();
throw new AIServiceError(e.message);
}
}
Related Docs
