46 lines
782 B
Markdown
46 lines
782 B
Markdown
---
|
|
layout: default
|
|
---
|
|
|
|
# createAgent: Creando el Agente
|
|
|
|
<div class="mt-6 text-xs">
|
|
|
|
<v-click>
|
|
|
|
<div class="mb-4">
|
|
|
|
**createAgent** es la forma estándar de crear agentes en LangChain v1
|
|
|
|
</div>
|
|
|
|
</v-click>
|
|
|
|
<v-click>
|
|
|
|
```typescript
|
|
import { createAgent } from "langchain";
|
|
|
|
const agent = createAgent({
|
|
model: "mistral:mistral-large-latest", // Modelo como string
|
|
tools: [weatherTool], // Array de tools
|
|
systemPrompt: "Eres un asistente útil.",
|
|
});
|
|
|
|
const result = await agent.invoke({
|
|
messages: [{ role: "user", content: "¿Qué tiempo hace en Madrid?" }]
|
|
});
|
|
```
|
|
|
|
</v-click>
|
|
|
|
<v-click>
|
|
|
|
<div class="mt-4 text-xs opacity-70">
|
|
El agente automatiza el ciclo: llamar al LLM -> ejecutar tools -> procesar resultados -> repetir
|
|
</div>
|
|
|
|
</v-click>
|
|
|
|
</div>
|