overthinkos/jupyter-ml
ov-images/skills/jupyter-ml/SKILL.md
Full CUDA ML JupyterLab image with real-time collaboration and CRDT MCP server. Base: nvidia. Port 8888. GPU-accelerated ML training + collaborative notebooks. MUST be invoked before building, deploying, or troubleshooting the jupyter-ml image.
tools
Updated Apr 24, 2026
$ install --global
skillsauthnpx skillsauth add overthinkos/overthink-plugins jupyter-mlInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 24, 2026, 2:25 PM142.6s1 file scanned
SKILL.md
- name:
- jupyter-ml
- description:
- |
- Base:
- nvidia. Port 8888. GPU-accelerated ML training + collaborative notebooks.
jupyter-ml -- GPU ML Jupyter with CRDT MCP
Image Definition
jupyter-ml:
base: nvidia
layers:
- agent-forwarding
- jupyter-ml
- notebook-templates
- dbus
- ov
ports:
- "8888:8888"
platforms:
- linux/amd64
Layer Composition
The jupyter-ml layer is a Tier 2 environment-owning meta-layer that composes:
llama-cpp— llama.cpp prebuilt binaries + GGUF toolsunsloth— vLLM 0.19 cu130 inference engine + fine-tuning (unsloth + unsloth-zoo) + vLLM torch.compile patchjupyter-mcp— CRDT MCP extension (fastmcp + jupyter_mcp package)
Additional layers from the image:
agent-forwarding— SSH/GPG agent forwardingnotebook-templates— Starter notebooks (data layer, seeds ~/workspace)dbus— D-Bus session busov— Overthink CLI
Base Chain
quay.io/fedora/fedora:43
└── fedora
└── fedora-nonfree (RPM Fusion repos)
└── nvidia (NVIDIA drivers + container toolkit)
└── jupyter-ml
Ports
| Port | Service |
|------|---------|
| 8888 | JupyterLab + MCP endpoint at /mcp |
Volumes
| Name | Path | Purpose | |------|------|---------| | workspace | ~/workspace | Persistent notebook storage | | models | ~/.cache/huggingface | HuggingFace model cache (from unsloth sub-layer) |
Service Environment Integration
This image receives env_provides variables from infrastructure layers when they are deployed:
| Variable | Injected by | Value |
|----------|------------|-------|
| OLLAMA_HOST | /ov-layers:ollama | http://ov-ollama:11434 |
| PGHOST | /ov-layers:postgresql | ov-postgresql |
| PGPORT | /ov-layers:postgresql | 5432 |
| REDIS_URL | /ov-layers:redis or /ov-layers:valkey | redis://ov-<image>:6379 |
These variables are injected automatically into the container environment at ov config time when the corresponding service is deployed. No manual -e flags needed.
Key Capabilities
- JupyterLab with real-time collaboration (jupyter-collaboration, Y-CRDT)
- CRDT MCP Server at
/mcp— 13 tools for programmatic notebook access - PyTorch >= 2.10.0 with CUDA 13.0
- vLLM 0.19 inference engine
- Unsloth fine-tuning (LoRA, QLoRA)
- LangChain RAG stack (chromadb, faiss-cpu)
- Evaluation tools (evidently, sacrebleu, rouge-score)
- Notebook templates seeded into ~/workspace
Quick Start
ov image build jupyter-ml
ov config jupyter-ml
ov start jupyter-ml
ov status jupyter-ml
ov logs jupyter-ml -f
# JupyterLab: http://localhost:8888
# MCP endpoint: http://localhost:8888/mcp
Verify
ov shell jupyter-ml -c "pixi run verify-pytorch"
ov shell jupyter-ml -c "pixi run verify-vllm"
ov shell jupyter-ml -c "pixi run verify-unsloth"
ov shell jupyter-ml -c "pixi run verify-mcp"
ov shell jupyter-ml -c "pixi run verify-collaboration"
Comparison with Other Jupyter Images
| | jupyter | jupyter-ml | |---|---|---| | Base | fedora | nvidia | | CUDA | No | Yes | | Arch | amd64 + arm64 | amd64 | | MCP | CRDT (13 tools) | CRDT (13 tools) | | Collaboration | Yes | Yes | | ML Stack | No | Full | | Volume | workspace | workspace + models | | Notebook dir | ~/workspace | ~/workspace |
Related Images
/ov-images:jupyter-ml-notebook— Same stack with fine-tuning notebooks/ov-images:jupyter— Lightweight variant (no CUDA, multi-arch)/ov-images:python-ml— ML base without Jupyter
MCP testing: inherits 3 deploy-scope mcp: checks from the jupyter-ml layer (ping, list-tools, call list_notebooks). Run ov test jupyter-ml --filter mcp or probe ad-hoc with ov test mcp list-tools jupyter-ml. See /ov:mcp.
When to Use This Skill
MUST be invoked before building, deploying, configuring, or troubleshooting the jupyter-ml image.
Related
/ov:image— image family umbrella (image:entries inoverthink.yml, build/validate/inspect/list)/ov:build—build.ymlvocabulary (distros, builders, init-systems)
Related Skills
overthinkos/plugin
tools
VerifiedTrustedCommunity
Use when authoring or modifying a charly PLUGIN — a candy with a `plugin:` block that contributes Providers (verbs/kinds/deploy-targets/steps/builders/commands), its own CUE schema, builtin (compiled-in) or external (out-of-tree git repo). Covers the unified Provider model, the per-plugin CUE-schema contract (single source → Go params for dev + schema-over-Describe RPC for runtime), the SDK, and the loader.
SKILL.mdUpdated Jun 22, 2026
overthinkos/cue
tools
VerifiedTrustedCommunity
The CUE data-validation / configuration CLI (cue), pinned to v0.16.1. Use when working with the cue candy, installing the cue binary into a box or onto a target:local dev host, or running the offline schema-vendoring pipeline that feeds charly's egress validation.
SKILL.mdUpdated Jun 16, 2026
overthinkos/egress
tools
VerifiedTrustedCommunity
CUE EGRESS validation — validating (and, where it adds value, generating) the config files charly WRITES to a system BEFORE the bytes hit disk. MUST be invoked before working on charly/egress.go, the vendored schemas under candy/plugin-egress/egress-schemas/vendor/, the ValidateEgress / registerVendoredEgressKind path, the offline `task cue:vendor` pipeline, or adding an egress schema for any written artifact (cloud-init, k8s manifests, traefik routes, runtime config, install ledger, systemd/quadlet units, ssh_config, libvirt XML).
SKILL.mdUpdated Jun 16, 2026
overthinkos/check-k8s
tools
VerifiedTrustedCommunity
Kubernetes cluster-probe declarative check verb — the `kube:` check verb (nodes, pods, ingress, storage class, addon health, apply/delete, and arbitrary resource GETs) served out-of-process by the candy/plugin-kube plugin (vendored client-go; no external kubectl required).
SKILL.mdUpdated Jun 14, 2026