Reference
Every component, what provides it, and whether it stays running.
Ask your agent: "install Qdrant and Langfuse". Core components are always present and set up by the installer. Optional ones install with what they need and use no CPU or memory while switched off. This page is the full list, with what each provides.
Optional components are installed from Thinkube Control under Optional Components. Each one has a toggle that scales it to zero.
Core
The first column is the name the dashboard shows. The second is the component behind it, the name to search for when something goes wrong.
| Shown as | Component | What it does |
|---|---|---|
Thinkube Identity |
Keycloak |
One login across every service |
Thinkube Registry |
Harbor |
Container images, built here and mirrored from upstream |
Thinkube Git |
Gitea |
Your repositories, in-cluster |
Argo Workflows |
Argo Workflows |
Builds images from your repositories and runs your applications' pipelines |
ArgoCD |
ArgoCD |
Deploys what is in Git |
PostgreSQL |
PostgreSQL |
Relational storage for the platform and your apps |
Thinkube Storage |
SeaweedFS |
S3-compatible; models and artifacts |
JuiceFS |
JuiceFS |
A POSIX filesystem on SeaweedFS, shared by every area; its metadata is in PostgreSQL |
Thinkube Packages |
DevPI |
Private Python packages |
Thinkube Notebooks |
JupyterHub |
Notebooks and environments |
Thinkube Experiments |
MLflow |
Runs, registered models, the model registry |
GPU Operator |
NVIDIA GPU Operator |
Device access; four time-slice slots per GPU by default |
Thinkube IDE |
code-server |
VS Code in the browser; Claude Code is installed from Anthropic and used with your own account |
Control panel |
Thinkube Control |
The web UI and the MCP endpoint |
Cluster |
kubeadm, Cilium |
Unmodified upstream Kubernetes; Cilium as network and load balancer |
Optional
| Component | Category | Needs | What it does |
|---|---|---|---|
vLLM |
ai |
— |
Inference backend for the LLM Gateway |
TensorRT-LLM |
ai |
— |
NVIDIA-optimised inference backend |
Ollama |
ai |
Thinkube Registry |
Inference backend for quantised GGUF models |
text-embeddings |
ai |
— |
Embedding backend (TEI) |
Langflow |
ai |
Thinkube Identity, PostgreSQL |
Visual builder for LLM applications |
Langfuse |
ai |
PostgreSQL, Thinkube Identity, ClickHouse, Valkey |
Tracing and observability for LLM applications |
Chroma |
ai |
Thinkube Registry |
Vector store |
Qdrant |
ai |
— |
Vector store |
Weaviate |
ai |
Thinkube Registry |
Vector store |
Argilla |
ai |
Thinkube Registry, Thinkube Identity, OpenSearch, Valkey |
Data labelling and curation |
CVAT |
ai |
— |
Image and video annotation |
ClickHouse |
data |
— |
Column-oriented analytics database |
OpenSearch |
data |
— |
Search and analytics |
Valkey |
data |
— |
Cache and key-value store (Redis-compatible) |
pgAdmin |
data |
PostgreSQL, Thinkube Identity |
Web UI for PostgreSQL |
Knative |
infrastructure |
— |
Scale-to-zero serving |
NATS |
infrastructure |
— |
Messaging |
Prometheus |
monitoring |
— |
Metrics |
Perses |
monitoring |
Thinkube Identity, Prometheus |
Dashboards |