Platform
Knowledge Bases
Un modelo de IA por defecto no sabe qué vendes ni cuáles son tus políticas. Una base de conocimiento le da a tu agente el catálogo, políticas de envío, FAQs y documentación. Subes tus datos una vez y Daguito los vectoriza. Cuando un cliente pregunta, el agente busca la respuesta correcta en tu información real.
Retrieval
Se buscan documentos relevantes en tu base de conocimiento usando búsqueda semántica (vectores).
Augmented
Los fragmentos recuperados se añaden al contexto del modelo de IA.
Generation
El modelo genera una respuesta basada en tu información, no en alucinaciones.
How it works
The process has three stages: search for relevant documents, add them to the model's context, and the model generates a response based on that information.
Semantic search
When you ask a question, the system converts it into a vector and compares it with millions of stored vectors. The closest ones are the most relevant. It's like automatically searching for synonyms — "price" and "cost" are found even though they're not the same word.
Example: Semantic Search
When a customer asks "how much does it cost?", the system searches for documents with similar meaning, not just exact words:
Semantic search (Builders)
Query by relevance, not keywords. The system searches the meaning of the question against all your vectorized documents.
const results = await daguito.kb.search({
kb_id: 'kb-uuid-123',
query: '¿cuál es el precio?',
limit: 5
})
// → [
// { text: "Costo: $100", score: 0.98, source: "pricing.pdf" },
// { text: "Total: $100 + envío", score: 0.94, source: "faq.md" },
// ...
// ]Adding knowledge
Upload documents from the platform panel. Upload files, paste text, or connect URLs — Daguito handles chunking, vectorization, and storage. Then connect the base to a Knowledge Search node in your flow so the agent uses it automatically.
Document ingestion (Builders)
Upload text directly, URLs, or files. Daguito vectorizes them in the background.
// Texto directo
await daguito.kb.addText({
kb_id: 'kb-123',
text: 'Envío 3-5 días hábiles',
source: 'shipping-policy'
})
// URL (se rastrea automáticamente)
await daguito.kb.addUrl({
kb_id: 'kb-123',
url: 'https://shop.com/policies'
})
// Archivo
await daguito.kb.upload({
kb_id: 'kb-123',
file: new File([...], 'catalog.pdf')
})Metadata and filters (Builders)
Tag documents. Filter searches by region, type, version, etc.
// Ingesta con metadata
await daguito.kb.addText({
kb_id: 'kb-123',
text: 'Desc 20% primeras compras',
metadata: {
type: 'promo',
valid_until: '2024-12-31',
regions: ['AR', 'BR']
}
})
// Búsqueda con filtro
const results = await daguito.kb.search({
kb_id: 'kb-123',
query: 'descuentos',
filter: { type: 'promo', regions: 'AR' }
})Using knowledge in agents
Connect your knowledge base to an agent node in the flow editor. Add the search_knowledge_base tool to the tools list. From there, the agent automatically decides when it needs to search for context.
When the agent needs information to answer, it triggers a search against your knowledge base. It receives the most relevant fragments ordered by similarity and uses them as context to generate the reply.
Connect to agent in a flow (Builders)
Drag the Knowledge Search node to the canvas and connect it to the AI Agent.
// En el editor de flujos, agrega el nodo Knowledge Search
// y conéctalo al AI Agent:
// 1. Arrastra un nodo "Knowledge Search" al canvas
// 2. Configura el kb_id (selecciona tu base)
// 3. Conecta: output "hits" → input "context" del AI Agent
// El agente ahora busca automáticamente cuando necesita info:
// Cliente: "¿cuánto cuesta el plan Pro?"
// Agente → search_knowledge_base({ query: "precio plan Pro" })
// → hits: [{ text: "Plan Pro: $49/mes", score: 0.96 }]
// → respuesta: "El plan Pro cuesta $49 al mes."