Overview

Ground synthetic data generation and evaluations in your organization's content so outputs stay relevant, consistent, and under your control.

What it is

A Knowledge Base (KB) is a store of organization content—FAQs, documentation, SOPs, manuals—that is semantically processed and indexed. Generation and evaluation runs use this content so outputs stay grounded in real source material instead of drifting into wrong terminology, invented procedures, or generic answers. The result is context-aware synthetic data and evals that reflect the organization’s domain, reduce hallucinations, and keep behavior consistent.

Purpose

  • Relevance — Outputs match your specific context and domain instead of generic or off-domain content.
  • Reliability — Consistent behavior aligned with your organization’s knowledge and needs.
  • Control — Scalable customization by adding, updating, or removing content in your KB.
  • Synthetic data generation — Create training examples from your documents so generated data reflects your domain; optional KB selection when creating synthetic datasets.
  • Evaluation and hallucination detection — Run evals that use your KB so you can check whether outputs are grounded in your source materials.

Getting started with knowledge base

Create a Knowledge Base using UI

Create and manage a KB from the Future AGI platform with drag-and-drop or bulk file upload.

Create a Knowledge Base using SDK

Create and update knowledge bases programmatically with the Python SDK for automation and large file ingestion.
Was this page helpful?

Questions & Discussion