Skip to main content

​
Enterprise Knowledge Base

The AccelOS Knowledge Base is a central repository for your team’s operational knowledge. It stores runbooks, procedures, and learnings that the AI agent can reference during incident investigations.

​
Why a Knowledge Base?

When an incident occurs, your team’s collective knowledge is often scattered across wikis, Slack threads, and tribal memory. The AccelOS Knowledge Base:
  • Centralizes operational knowledge in one searchable location
  • Enables AI to reference your team’s expertise during investigations
  • Captures learnings from incidents to prevent future issues
  • Standardizes procedures across your team

​
Content Types

Alert Runbooks

Step-by-step procedures for handling specific alerts and incident types.

Integrations

Service-specific troubleshooting guides tied to your integrations.

Generic Instructions

General operational procedures and best practices.

Memories

Learnings from past investigations and incidents.

​
How It Works

​
1. Create Documentation

Write runbooks and procedures using markdown. Each document includes:
  • Title - Clear, searchable name
  • Category - Alert Runbook, Integration, Generic Instruction, or Memory
  • Content - Step-by-step procedures, context, and commands
  • Metadata - Associated alerts, integrations, or services

​
2. AI Agent Access

During investigations, the AI agent automatically:
  • Searches the knowledge base for relevant documents
  • References runbooks that match the incident type
  • Suggests procedures based on your team’s documented knowledge
  • Cites sources so you can verify recommendations

​
3. Continuous Improvement

After each incident:
  • Document learnings in the knowledge base
  • Update runbooks based on what worked
  • Add new procedures for novel issues
  • Build organizational memory over time

​
Document Structure

Each knowledge base document follows a consistent structure: 
---
category: alert_runbooks
alert_name: HighCPUUsage
integration_id: <uuid>
---

# High CPU Usage Runbook

## Symptoms
- CPU usage above 90% for more than 5 minutes
- Increased latency on affected services

## Investigation Steps
1. Check which processes are consuming CPU
2. Review recent deployments
3. Check for traffic spikes

## Remediation
1. Scale horizontally if load-related
2. Restart affected pods if process is stuck
3. Roll back if caused by recent deployment

## Escalation
Contact the platform team if issue persists > 30 minutes.

​
Categories Explained

​
Alert Runbooks

Alert runbooks are tied to specific alerts or incident types. They provide:
  • Clear symptoms to identify the issue
  • Step-by-step investigation procedures
  • Remediation actions
  • Escalation paths
Best for: Recurring alerts, known failure modes, standard operating procedures.

​
Integrations

Integration docs contain service-specific knowledge for your connected tools:
  • How to interpret specific metrics
  • Common queries and commands
  • Service-specific troubleshooting
  • Architecture context
Best for: Service-specific knowledge, tool documentation, integration guides.

​
Generic Instructions

General procedures that apply across services:
  • Incident response protocols
  • Communication templates
  • General debugging techniques
  • Tool usage guides
Best for: Team-wide procedures, best practices, onboarding materials.

​
Memories

Learnings captured from past investigations:
  • Root cause analyses
  • Novel problems and solutions
  • Lessons learned
  • Post-mortem insights
Best for: Building institutional knowledge, preventing repeat incidents.

​
Features

​
Version History

All documents are version-controlled:
  • Track changes over time
  • See who made edits
  • Revert to previous versions if needed

​
AI-Generated Reviews

When you save a document, the system can:
  • Suggest improvements for clarity
  • Identify missing sections
  • Flag potential issues

​
Search & Discovery

Find relevant documents through:
  • Full-text search
  • Category filters
  • Integration/alert associations
  • AI-powered recommendations

​
Best Practices

Begin by documenting your most frequent or critical alerts:
  • Alerts that page on-call regularly
  • Incidents that take longest to resolve
  • Issues that require specialized knowledge
Write for the 3 AM on-call engineer:
  • Use clear, numbered steps
  • Include actual commands (with placeholders)
  • Specify expected outputs
  • Define clear escalation criteria
Treat runbooks as living documents:
  • Update procedures that didn’t work
  • Add new steps discovered during investigation
  • Remove outdated information
  • Capture novel solutions as memories

​
Getting Started

Setup Guide

Step-by-step guide to setting up your knowledge base.