Best viewed on desktop. Scroll horizontally to see the full diagram.
GKG Infrastructure
Click any component for details. Data flows left to right.
GitLab / GCP
Siphon CDC (Go)
NATS
GKG Services (Rust)
ClickHouse
DAP
External Agents
GitLab Platform
CDC Pipeline
GKG — GKE Cluster (GCP)
ClickHouse Cloud
Consumers
🗄
PostgreSQL
GitLab main DB
WAL Enabled
🗄
PostgreSQL
CI database
WAL Enabled
📦
Gitaly
Git object storage
gRPC
🔐
GitLab Rails
Auth + permissions
Ruby on Rails
💧
Siphon
CDC connector
Go / LabKit
📨
NATS
Message broker
JetStream
⚙️
ETL Consumer
SDLC transformer
Rust
🔍
Code Indexer
Tree-sitter analysis
Rust
🕐
Dispatcher
Namespace cron
Rust / ~1min
🧩
Query Engine
DSL → SQL compiler
Rust / Security
🌐
Webserver
REST + gRPC API
Rust / Axum
❤️
Health Check
Status reporting
Rust / Prometheus
⚡
ClickHouse
Graph database
nodes / edges
SDLC base tables
code tables
materialized views
Columnar / Managed
🔑
GCP Secret Manager
Credentials + certs
External Secrets Op.
📋
Cloud Logging
→ Elastic / OpenSearch
OTel / LabKit-rs
🤖
Duo Agent Platform (DAP)
Zero-rated — GKG is COGS
query_graph / get_graph_schema
🔀
AI Gateway (AIGW)
Request routing + JWT
Proxy / Tool routing
⚡
MCP Clients
Claude Code, Codex, etc.
Billed via GitLab Credits
💻
glab CLI
Terminal / headless access
Billed via GitLab Credits
WAL CDC events SDLC push events fetch files write nodes/edges schedules SQL AuthZ check / permissions MCP tools (zero-rated) REST API (billed Credits) inject secrets
~5s lag
~3s lag
<1s query