Skip to main content
Teach Foggy about your specific infrastructure and processes. All enabled entries are automatically included as context in every investigation — so Foggy uses your team’s knowledge, not just generic infrastructure patterns.

​
How it works

  1. Write entries in Markdown — architecture docs, runbooks, troubleshooting procedures
  2. Enable or disable individual entries with a toggle
  3. Foggy uses them automatically — enabled entries are sent to the agent as context before every investigation
Knowledge Base list view with toggle states

​
Two categories

Product DocsRunbooks
PurposeHelps Foggy understand what your system isHelps Foggy understand how to investigate failures
ContentServices, architecture, deployment topology, team ownershipTroubleshooting procedures with a “use when” trigger
Example”Checkout Service runs Go, uses PostgreSQL, owned by #platform-team""When DNS fails → check CoreDNS pods, verify service DNS, escalate to #platform-team”

​
Example entries

Product Doc: “Checkout Service” 
## Checkout Service
- Language: Go
- Database: PostgreSQL (primary + read replica)
- Dependencies: payment-service, inventory-service, user-service
- Deployment: 3 replicas in production, 1 in staging
- Owner: Platform Team (#platform-team in Slack)
Runbook: “DNS Troubleshooting”
  • Use when: DNS resolution fails or services can’t reach each other
## Steps
1. Check CoreDNS pods: kubectl get pods -n kube-system -l k8s-app=kube-dns
2. Check CoreDNS logs for errors
3. Verify service DNS: nslookup <service>.<namespace>.svc.cluster.local
4. Check if the issue is namespace-specific or cluster-wide
5. Escalate to: #platform-team

​
Writing effective entries

  • One topic per entry — a focused “Payment Service Architecture” is more useful than a massive “Everything About Our System” doc
  • Include escalation paths — add team names and Slack channels so Foggy can suggest who to contact
  • Disable rather than delete — if an entry becomes outdated, toggle it off. Re-enable it when it’s relevant again.

​
Next steps

Automations

Set up scheduled and alert-triggered investigations that run automatically.

Slack Bot

Investigate incidents and get automation results directly in Slack.

How Investigations Work

Chain-of-thought, follow-ups, @mentions, and model selection.

Connect data sources

Add Grafana, Kubernetes, and more to power investigations.