Skip to main content

Overview

The Knowledge Base tool enables your Utari workers to store, organize, and intelligently retrieve information from a personal knowledge library. Using semantic search and natural language queries, agents can find relevant information across documents, maintain organized file structures, and access contextual knowledge to inform their responses and actions.

Knowledge Base Capabilities

Semantic Search

Find information using natural language queries across all knowledge base files

File Management

Upload, organize, enable/disable, and delete files and folders

Sync & Index

Synchronize knowledge base to local workspace for fast access

List Contents

View all folders and files in the knowledge base hierarchy

Enable/Disable

Control which files are active and searchable for specific agents

Maintenance

Cleanup and optimization operations for knowledge base health
The Knowledge Base uses semantic search, meaning it understands the meaning and context of queries rather than just matching keywords. Ask questions naturally, and the system finds relevant information even if exact words don’t match.

Understanding the Knowledge Base

Semantic Search Explained

Unlike traditional keyword search, semantic search understands:
  • Meaning and Context: “How do I improve sales?” finds information about “increasing revenue” and “customer conversion”
  • Related Concepts: Searching for “pricing strategy” finds documents about “monetization” and “value-based pricing”
  • Natural Language: Ask questions as you would ask a person
  • Intent Recognition: Understands what you’re really looking for
Example:
  • Query: “How should we position our product?”
  • Finds: Documents about “market positioning,” “competitive advantage,” “value proposition,” and “target audience definition”

Global vs. Local Knowledge Base

Location: Cloud-based, accessible across all agentsPurpose:
  • Centralized knowledge repository
  • Shared across workspace
  • Persistent storage
  • Version controlled
Management:
  • Upload files from sandbox
  • Create folder structure
  • Enable/disable for specific agents
  • Delete when no longer needed
Use Cases:
  • Company SOPs and guidelines
  • Reusable templates
  • Shared reference materials
  • Brand guidelines
  • Product documentation
1

Ask Natural Language Questions

2

Worker Performs Semantic Search

The system:
  • Understands the semantic meaning of your query
  • Searches all indexed files in knowledge base
  • Ranks results by relevance
  • Returns most pertinent information
3

Review Results

4

Use Retrieved Information

Worker uses found information to:
  • Answer your questions
  • Inform recommendations
  • Create content aligned with guidelines
  • Reference established procedures
Search within specific files or folders:

Advanced Search Techniques

Ask about concepts, not just keywords:
Ask direct questions:
The system finds procedural documents and guidelines that answer these questions.
Provide context for better results:
Search for multiple related aspects:

Managing Knowledge Base Files

Listing Knowledge Base Contents

1

View All Files

2

Review Structure

3

Check File Status

Uploading Files to Knowledge Base

1

Ensure File in Sandbox

Files must exist in the workspace before uploading:
2

Create Folder (Optional)

3

Upload to Knowledge Base

4

Verify Upload

5

Enable for Agent

Organizing with Folders

1

Create Folder Structure

2

Upload to Folders

3

Maintain Organization

Enabling and Disabling Files

1

Control Agent Access

2

Strategic Enabling

3

Performance Optimization

Deleting Files and Folders

1

Identify Items to Delete

2

Delete Files

3

Confirm Deletion

4

Cleanup Complete

Syncing Knowledge Base

Global KB Sync

1

Initiate Sync

2

Sync Process

3

Access Synced Files

4

Re-sync as Needed

Knowledge Base Maintenance

Cleanup Operations

1

Perform Cleanup

2

Cleanup Tasks

3

Regular Maintenance

Use Cases and Workflows

Reference-Based Content Creation

Creating Brand-Aligned Content

Procedure-Guided Task Execution

Following SOPs

Question Answering with Context

Informed Responses

Multi-Document Synthesis

Comprehensive Research

Best Practices

Organized Structure

Maintain clear folder hierarchy mirroring your organizational structure

Descriptive Naming

Use clear, searchable file names: “Brand_Voice_Guide_2024.pdf” not “doc1.pdf”

Regular Updates

Keep knowledge base current by updating or replacing outdated files

Strategic Enabling

Only enable relevant files for each agent to improve search quality

Periodic Cleanup

Regularly remove obsolete files and run maintenance

Version Control

Include version/date in file names for evolving documents

Format Consistency

Use consistent file formats (PDF for final, DOCX for editable)

Document Metadata

Include README files in folders explaining contents and purpose

Supported File Types

  • PDF (.pdf) - Best for final documents, reports, guides
  • Word (.docx, .doc) - Editable documents
  • Text (.txt) - Plain text files
  • Markdown (.md) - Formatted text documentation
  • RTF (.rtf) - Rich text format

Troubleshooting

Check:
  • Files are enabled for the agent
  • Files have finished indexing (check with ls_kb)
  • Try broader or different search terms
  • Verify files actually contain relevant information
  • Run sync to update local copy
  • Check if files are in correct format
Verify:
  • File was successfully uploaded
  • Check with list contents to see all files
  • File is not being filtered from view
  • File finished processing
  • Refresh or re-list contents
Possible causes:
  • Large file size (may take longer)
  • Complex document structure
  • High system load
  • Wait 5-10 minutes for large files
  • Check file isn’t corrupted
  • Try re-uploading if stuck
Ensure:
  • File exists in workspace sandbox first
  • File path is correct
  • File format is supported
  • Sufficient storage space
  • File isn’t corrupted
  • Try smaller file or different format
Improve by:
  • Using more specific, natural language queries
  • Providing context in your search
  • Checking if right files are enabled
  • Ensuring files contain expected information
  • Trying question-based queries
  • Updating knowledge base with better documents
Try:
  • Check network connection
  • Verify permissions on files
  • Run sync again
  • Check available disk space
  • List contents to see what should sync
  • Contact support if persists

Summary

You’ve successfully learned how to:
Perform semantic search using natural language queries
Upload and organize files in the knowledge base
Create folder structures for organization
Enable and disable files strategically for agents
Sync knowledge base to local workspace
List and review knowledge base contents
Delete files and folders when no longer needed
Perform maintenance and cleanup operations
Apply best practices for knowledge management
The Knowledge Base tool transforms static documents into an intelligent, searchable knowledge library that agents can query naturally to find relevant information, follow established procedures, and create content aligned with organizational guidelines—making institutional knowledge accessible and actionable.

Next Steps

Activate Knowledge

Learn to activate knowledge for specific workers

Add Knowledge

Upload SOPs and documents to knowledge base

Agent Configuration

Configure agents with knowledge access

Jeremy SOPs

Access pre-built knowledge resources