Knowledge base retrieval

Overview

Tool identifier load_knowledgebase. System tools are built-in tools that VeADK mounts automatically based on the agent’s configuration — you don’t add them to the tools list yourself. When you pass a knowledgebase to the agent, VeADK mounts the load_knowledgebase tool. The agent decides on its own when to query the knowledge base and which keywords to search with, then uses the retrieved knowledge as context for its answer. For backend types, initialization, and data ingestion, see the knowledge base docs.

Prerequisites

Install python -m pip install "veadk-python[extensions]", then configure the model and local backend. Local storage still sends text to the configured embedding service. Ingest data before the first search.

Usage

import asyncio
from veadk import Agent, Runner
from veadk.knowledgebase import KnowledgeBase

knowledgebase = KnowledgeBase(backend="local", index="company_faq")
assert knowledgebase.add_from_text("Employees have 15 days of annual leave after one year of service")
agent = Agent(
    name="kb_agent",
    instruction="Call load_knowledgebase to retrieve the leave policy before answering.",
    knowledgebase=knowledgebase,
)
runner = Runner(agent=agent)

async def main():
    print(await runner.run("How many days of annual leave do employees receive?"))

asyncio.run(main())

Notes

System tools are managed by the framework: just provide a knowledgebase — you don’t add load_knowledgebase to the tools list yourself. The agent decides when to call it based on its instruction and user input.

Retrieval input and checks

Parameter	Type	Default	Description
`query`	`str`	Required	Search terms or question generated by the agent
`tool_context`	`ToolContext`	Injected	Current invocation context, not supplied by the model

The knowledges list contains retrieved knowledge entries. The expected answer includes the 15-day annual leave policy. Passing a user ID does not automatically enforce document permissions; multitenant applications must restrict which knowledge base and data are bound to each request. The model still decides whether to call the tool. If the answer does not reflect the data, inspect tool events and retrieval results. If direct retrieval fails, check ingestion, embedding configuration, and backend permissions first.

Overview

Get started

Core components

Observability

Extensions

Frontend

Security

Deployment

Overview

Prerequisites

Usage

Notes

Retrieval input and checks

​Overview

​Prerequisites

​Usage

​Notes

​Retrieval input and checks

Overview

Prerequisites

Usage

Notes

Retrieval input and checks