Knowledge Base

What is a Knowledge Base?

A Knowledge Base (KB) is a centralized repository of documents and content that can be ingested, processed, and queried using natural language. It enables businesses to organize unstructured information — such as PDFs, Markdown files, JSON data, CSVs — into a searchable, AI-powered knowledge store.

Unlike traditional datasets that operate on structured, tabular data, a Knowledge Base is designed for unstructured and semi-structured content. Once documents are added to a Knowledge Base, the platform automatically processes and indexes them, making the information available for retrieval-augmented generation (RAG) and conversational querying through the built-in Knowledge Base Chat.

Common Use Cases:

• Internal Documentation: Centralize company policies, SOPs, and onboarding materials for instant retrieval.
• Resume Screening: Upload candidate resumes and query them for specific skills, experience, or qualifications.
• Research Repositories: Aggregate research papers, reports, and whitepapers into a queryable knowledge store.
• Customer Support: Build a KB from support articles and FAQs to power intelligent search and chatbots.
• Compliance & Legal: Organize regulatory documents, contracts, and audit reports for quick lookup.

Key Concepts

Knowledge Entities (Sources)

A Knowledge Entity (also called a Source) is an individual piece of content linked to a Knowledge Base. Each entity represents a single document or URL that has been ingested and processed.

Each source has the following attributes:

• Name: A descriptive label for the source (e.g., "Q4 Sales Report").
• Source Type: The ingestion method — either Upload File or URL.
• Description: A brief summary of the source content.
• Type Badge: Indicates the file format (e.g., Pdf, Md, Csv).
• Records: The number of processed chunks/records extracted from the source. For PDFs, this indicates the number of pages.
• Status: The processing state — typically ACTIVE once successfully ingested, ERROR if ingestion fails.
• Updated Date: Timestamp of the last modification.

Supported Source Types

Knowledge Bases support two methods of content ingestion:

1. Upload File: Drag-and-drop or browse to upload files directly. Supported formats:

   • .pdf — PDF documents
   • .md / .mdx — Markdown files
   • .text — Plain text files
   • .json / .jsonl — JSON and JSON Lines
   • .csv — Comma-separated values

2. URL: Provide an accessible / downloadable URL of the above formats.

Records

When a source is ingested, its content is automatically split into records (also referred to as chunks). These records represent discrete units of information that the platform indexes and uses for retrieval during queries. The number of records depends on the length and structure of the source document. For PDFs, this indicates the number of pages ingested.

Knowledge Base Chat

Every Knowledge Base includes a built-in chat interface for querying its contents. The chat panel uses retrieval-augmented generation (RAG) to find relevant information from linked sources and generate natural language responses.

Key features of KB Chat:

• Session Management: Create multiple chat sessions with the + New Session button.
• Natural Language Queries: Ask questions in plain language and receive answers grounded in your KB content.
• Conversation History: Past sessions are preserved for reference.

How Knowledge Base Works

The Knowledge Base follows an Ingest → Process → Query pipeline:

1. Ingest: Users upload documents or provide URLs as knowledge entities. The platform accepts multiple file formats and web content.

2. Process: Uploaded content is automatically parsed, chunked into records, and indexed. Each record becomes a retrievable unit of information.

3. Query: Users interact with the Knowledge Base through the built-in chat interface. The system retrieves relevant records based on the query and generates an AI-powered response.

Where to Find Knowledge Base

Knowledge Base is located within the Dataset Manager section of the Data Hub:

Data Hub → Dataset Manager → Knowledge Base

From the Knowledge Base listing page, users can:

• View all existing Knowledge Bases in a tabular format
• Create new Knowledge Bases
• Search, sort, and filter Knowledge Bases
• Edit or delete existing Knowledge Bases