> ## Documentation Index
> Fetch the complete documentation index at: https://docs.agentium.in/llms.txt
> Use this file to discover all available pages before exploring further.

# Knowledge Base Overview

> RAG concept, KnowledgeBase class, and asTool() for equipping agents with document retrieval in Agentium.

# Knowledge Base Overview

Agentium provides a **KnowledgeBase** abstraction for **Retrieval-Augmented Generation (RAG)**. Store documents in a vector store, search by semantic similarity, and expose retrieval as a tool so agents can answer questions using your private data.

***

## What is RAG?

<CardGroup cols={2}>
  <Card title="Retrieve" icon="search">
    Convert documents to embeddings, store in a vector database, and search by semantic similarity.
  </Card>

  <Card title="Augment" icon="plus">
    Inject retrieved chunks into the LLM context before generating a response.
  </Card>

  <Card title="Generate" icon="message">
    The model produces answers grounded in the retrieved content instead of relying only on its training data.
  </Card>
</CardGroup>

RAG reduces hallucinations and keeps responses aligned with your documents—ideal for internal docs, support knowledge bases, and domain-specific Q\&A.

***

## KnowledgeBase Class

```typescript theme={null}
import { KnowledgeBase } from "@agentium/core";

const kb = new KnowledgeBase({
  name: "Product Docs",
  vectorStore: myVectorStore,
  collection: "product_docs", // optional, defaults to sanitized name
});

await kb.initialize();
await kb.add({ id: "doc-1", content: "Our API supports webhooks..." });
const results = await kb.search("How do I set up webhooks?");
```

<ParamField path="name" type="string" required>
  Display name for the knowledge base. Used in tool descriptions.
</ParamField>

<ParamField path="vectorStore" type="VectorStore" required>
  A vector store implementation (InMemory, PgVector, Qdrant, MongoDB).
</ParamField>

<ParamField path="collection" type="string" required={false}>
  Collection/index name inside the vector store. Defaults to a sanitized version of `name`.
</ParamField>

<ParamField path="searchMode" type="&#x22;vector&#x22; | &#x22;keyword&#x22; | &#x22;hybrid&#x22;" required={false} default="vector">
  Default search strategy. `"hybrid"` combines vector + keyword search via Reciprocal Rank Fusion for the best results. See [Hybrid Search](/knowledge/hybrid-search).
</ParamField>

<ParamField path="hybridConfig" type="HybridSearchConfig" required={false}>
  Fine-tune hybrid search weights and RRF constant. See [Hybrid Search](/knowledge/hybrid-search).
</ParamField>

***

## asTool() — Expose KB to Agents

The most powerful feature: turn a KnowledgeBase into a **ToolDef** that agents can call automatically.

```typescript theme={null}
const kb = new KnowledgeBase({
  name: "Support KB",
  vectorStore: vectorStore,
});

await kb.initialize();
await kb.addDocuments(docs);

const agent = new Agent({
  name: "SupportBot",
  model: openai("gpt-4o"),
  tools: [kb.asTool({ toolName: "search_kb", topK: 5 })],
  instructions: "Answer using the knowledge base when relevant.",
});

const result = await agent.run("How do I reset my password?");
// Agent calls search_kb internally and uses retrieved docs in its answer
```

<ParamField path="toolName" type="string" required={false}>
  Tool name exposed to the LLM. Default: `search_<collection>`.
</ParamField>

<ParamField path="description" type="string" required={false}>
  Custom tool description. Default: auto-generated from KB name.
</ParamField>

<ParamField path="topK" type="number" required={false} default="5">
  Number of results to return per search.
</ParamField>

<ParamField path="minScore" type="number" required={false}>
  Minimum similarity score to include a result.
</ParamField>

<ParamField path="filter" type="Record<string, unknown>" required={false}>
  Metadata filter applied to every search.
</ParamField>

<ParamField path="searchMode" type="&#x22;vector&#x22; | &#x22;keyword&#x22; | &#x22;hybrid&#x22;" required={false}>
  Override search mode for this tool. Inherits from KB config if not set.
</ParamField>

<ParamField path="formatResults" type="(results) => string" required={false}>
  Custom formatter for search results. Default: numbered list with scores.
</ParamField>

***

## Flow Diagram

<Steps>
  <Step title="Create KnowledgeBase">
    Configure name, vector store, and optional collection.
  </Step>

  <Step title="Initialize">
    Call `kb.initialize()` to ensure the vector store is ready.
  </Step>

  <Step title="Add Documents">
    Use `add()` or `addDocuments()` to ingest content. Embeddings are computed if not provided.
  </Step>

  <Step title="Expose as Tool">
    Call `kb.asTool()` and pass the result to `Agent({ tools: [...] })`.
  </Step>

  <Step title="Query">
    When the user asks a question, the agent calls the search tool and uses results in its response.
  </Step>
</Steps>

***

## Next Steps

* [Vector Stores](/knowledge/vector-stores) — InMemory, PgVector, Qdrant, MongoDB
* [Embeddings](/knowledge/embeddings) — OpenAI and Google embedding providers
* [Hybrid Search](/knowledge/hybrid-search) — BM25, RRF, and hybrid search modes
* [RAG Example](/knowledge/rag-example) — End-to-end walkthrough