Quick navigation: If you're new, start with Quickstart → One Pager → Triage Methodology.
- Quickstart Guide - First investigation in 5 minutes
- One Pager - Leadership overview (why this exists)
- Environment Variables - Configuration reference
- Sample Reports & Screenshots - See what Tarka produces without running the full stack
- Deployment Guide - Kubernetes deployment
- Operations - Webhook setup, smoke testing, troubleshooting
- Testing - Running unit and integration tests
- UI Styles - Console UI styling guidelines
- Architecture Overview - System design and key abstractions
- Investigation Pipeline - How investigations work (SSOT, 11 stages)
- Diagnostic Modules - Universal failure mode detection
- Playbook System - Alert routing and evidence collection
- Extending Playbooks - How to add new alert types
- Triage Methodology - Quality philosophy (honesty over guessing)
These define the contract for deterministic triage reports:
- Triage Methodology - Core philosophy
- Base Contract - label + why + next; blocked scenarios A-D
- Family Specs - Alert-family-specific enrichment rules
- Scoring Overview - Impact, confidence, noise axes
- Scoring Contract - 0-100 scales and classification rules
- Golden Corpus - Real-world calibration examples
- Tool-Using Chat - Case investigation assistant (PromQL, K8s, logs, memory)
- Chat Tools Reference - Full list of agent tools, parameters, and policy gates
- Action Proposals - Policy-gated remediation suggestions
- Case Memory - Learn from past incidents (PostgreSQL + embeddings)
- Completed Initiatives - Major features shipped
- Planned Enhancements - Future work and backlog
Path 1: Decision Maker (10 min)
Path 2: On-Call Engineer (20 min)
Path 3: Developer Adding Features (45 min)
Path 4: Operator Deploying to Production (30 min)