@kuralle-agents/tools

Pre-built AI SDK tools for grounded retrieval and answering in Kuralle agents.

Install

npm install @kuralle-agents/tools

Peers: @kuralle-agents/core @kuralle-agents/rag ai@^6 zod.

What it does

Three tools you attach directly to a defineAgent call: CAG retrieve, CAG answer, and vector similarity retrieval.

Key exports:

• createCagTool — retrieves relevant chunks from a KnowledgeSource (Chunk-and-Generate retrieve step).
• createCagAnswerTool — retrieves chunks and synthesizes a grounded answer in one tool call.
• createVectorRetrievalTool — similarity search over a VectorStoreCore with optional metadata filtering.

Usage

import { defineAgent, createRuntime, buildToolSet } from '@kuralle-agents/core';
import { createCagTool, createCagAnswerTool } from '@kuralle-agents/tools';
import { createStaticKnowledgeSource, createMarkdownChunker } from '@kuralle-agents/rag';
import { openai } from '@ai-sdk/openai';

const source = createStaticKnowledgeSource({
  content: '# Refund Policy\nRefunds are available within 30 days.',
  chunker: createMarkdownChunker(),
});

const retrieveTool = createCagTool({ source, topK: 5 });
const answerTool = createCagAnswerTool({ source, model: openai('gpt-4o-mini') });

const agent = defineAgent({
  id: 'support',
  instructions: 'Answer questions using the provided knowledge.',
  model: openai('gpt-4o-mini'),
  tools: { retrieve: retrieveTool, answer: answerTool },
});

Vector retrieval

import { createVectorRetrievalTool } from '@kuralle-agents/tools';
import { InMemoryVectorStore, AiSdkEmbedder } from '@kuralle-agents/rag';
import { openai } from '@ai-sdk/openai';

const vectorStore = new InMemoryVectorStore();
const embedder = new AiSdkEmbedder({ model: openai.embedding('text-embedding-3-small') });

const vectorTool = createVectorRetrievalTool({
  vectorStore,
  embedder,
  topK: 5,
  indexName: 'docs',
});

Related

• @kuralle-agents/core — runtime and agent primitives.
• @kuralle-agents/rag — chunkers, retrievers, vector stores.