overthinkos/ollama
ollama/skills/ollama/SKILL.md
Standalone Ollama LLM inference server with CUDA GPU support. Runs as a supervisord service on port 11434 with persistent model storage. MUST be invoked before building, deploying, configuring, or troubleshooting the ollama box.
development
Updated Jun 11, 2026
$ install --global
skillsauthnpx skillsauth add overthinkos/overthink-plugins ollamaInstall 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: Jun 11, 2026, 7:21 AM115.0s1 file scanned
SKILL.md
- name:
- ollama
- description:
- |
ollama
GPU-accelerated Ollama LLM inference server.
Box Properties
| Property | Value | |----------|-------| | Base | nvidia | | Candies | agent-forwarding, ollama | | Platforms | linux/amd64 | | Ports | 11434 | | Registry | ghcr.io/overthinkos |
Full Candy Stack
fedora→nvidia(CUDA base)pixi→python→supervisord(transitive)ollama— LLM server, models volume
GPU is box-level — the ollama candy is GPU-agnostic
The ollama candy installs the distro-agnostic Ollama binary (a tarball
extracted to /usr) and a supervisord ollama serve service. The binary
auto-detects the GPU at runtime and falls back to CPU inference when none is
present, so the candy carries no cuda dependency — it only require:s
supervisord. GPU support is a composition choice made at the box level:
- This
ollamabox keeps GPU acceleration viabase: nvidia(thenvidiabox composes thecudacandy, inherited through the base chain). - CPU-only consumers compose the
ollamacandy on a non-NVIDIA base and get CPU inference for free — e.g./charly-openclaw:openclaw-desktop(cachyos base, nocuda).
Ports
| Port | Service | Protocol | |------|---------|----------| | 11434 | Ollama API | HTTP |
Volumes
| Name | Path | Purpose | |------|------|---------| | models | ~/.ollama | Model storage |
Quick Start
charly box build ollama
charly config ollama
charly start ollama
charly shell ollama -c "ollama pull llama3"
charly shell ollama -c "ollama run llama3 'Hello'"
Host Alias
charly alias install ollama
# Now: ollama pull llama3 (runs inside the container)
Service Environment
When deployed via charly config ollama, this box automatically provides OLLAMA_HOST=http://charly-ollama:11434 to all other deployed containers via the env_provide mechanism. Use --update-all to propagate to already-deployed services:
charly config ollama --update-all
This means containers like jupyter-ml-notebook automatically discover the Ollama endpoint without manual OLLAMA_HOST configuration.
Key Candies
/charly-ollama:ollama— Ollama binary, supervisord service, model volume/charly-distros:cuda— GPU support (via nvidia base)
Related Boxes
/charly-distros:nvidia— parent (GPU without Ollama)- CachyOS variant —
cachyos.ollamais the CachyOS GPU sibling (built on thecachyos.nvidiaGPU base) in theoverthinkos/cachyossubmodule. See/charly-distros:cachyos. /charly-openclaw:openclaw-desktop— composes theollamacandy CPU-only (cachyos base, nocuda) alongside a streaming desktop + the openclaw gateway + the nested charly toolchain/charly-jupyter:jupyter-ml-notebook— Jupyter with Ollama integration notebooks (receivesOLLAMA_HOSTautomatically via env_provide when ollama is deployed)/charly-openwebui:openwebui— Open WebUI (receivesOLLAMA_HOSTvia env_provide, auto-configures asOLLAMA_BASE_URL)
Related Candies
/charly-ollama:ollama— the Ollama binary candy/charly-jupyter:notebook-ollama— 6 Jupyter notebooks demonstrating Ollama APIs (requests, OpenAI, ollama lib, Anthropic, HuggingFace, GPU)
Verification
After charly start:
charly status ollama— container runningcharly service status ollama— all services RUNNINGcurl -s http://localhost:11434/api/tags— Ollama API responds
When to Use This Skill
MUST be invoked when the task involves the ollama box, LLM model serving, or the standalone Ollama deployment. Invoke this skill BEFORE reading source code or launching Explore agents.
Related
/charly-image:image— image family umbrella (box:entries incharly.yml, build/validate/inspect/list)/charly-build:build—build.ymlvocabulary (distros, builders, init-systems)
Related Skills
overthinkos/charly
tools
VerifiedTrustedCommunity
OpenCharly CLI (charly) binary installed into container/VM images for in-container use. Use when working with charly binary deployment inside containers, native D-Bus support, or the full charly toolchain (charly binary + virtualization + gocryptfs + socat).
SKILL.mdUpdated Jun 10, 2026
overthinkos/charly-cachyos
development
VerifiedTrustedCommunity
Operator CachyOS workstation profile — a kind:local template + target:local deploy that installs the full dev stack (30 candies) onto a CachyOS host via ShellExecutor. Lives in the overthinkos/cachyos submodule. MUST be invoked before editing or applying the charly-cachyos workstation profile.
SKILL.mdUpdated Jun 10, 2026
overthinkos/charly-fedora
tools
VerifiedTrustedCommunity
Fedora box with the full charly toolchain using shared candies. Rootless-first — runs as uid=1000 with passwordless sudo (no root, no cap_add: ALL). Same candy list as charly-arch. Includes NVIDIA GPU runtime. MUST be invoked before building, deploying, configuring, or troubleshooting the charly-fedora box.
SKILL.mdUpdated Jun 10, 2026
overthinkos/charly-arch
tools
VerifiedTrustedCommunity
Arch Linux box with the full charly toolchain. Rootless-first — runs as uid=1000 with passwordless sudo (no root, no cap_add: ALL). Composes /charly-coder:charly-mcp so the box is reachable as an MCP gateway on port 18765. NVIDIA GPU runtime composed in. MUST be invoked before building, deploying, configuring, or troubleshooting the charly-arch box.
SKILL.mdUpdated Jun 10, 2026