mkdocs.yml

site_name: AI Factory Operations Lab
site_description: A hands-on, mostly-free course in AI/HPC GPU infrastructure operations - Kubernetes GPU scheduling, queueing (KAI), GPU sharing (HAMi), Slurm, observability, inference, and cluster lifecycle.
site_author: Lovedeep Singh
site_url: https://ld-singh.github.io/ai-factory-ops-lab/

repo_url: https://github.com/ld-singh/ai-factory-ops-lab
repo_name: ld-singh/ai-factory-ops-lab

# Content is synced from portfolio-lab/ etc. into docs/ by scripts/sync-docs.sh.
docs_dir: docs
site_dir: site

theme:
  name: material
  custom_dir: overrides
  icon:
    logo: material/expansion-card-variant
    repo: fontawesome/brands/github
  palette:
    - media: "(prefers-color-scheme: light)"
      scheme: default
      primary: green
      accent: teal
      toggle:
        icon: material/weather-night
        name: Switch to dark mode
    - media: "(prefers-color-scheme: dark)"
      scheme: slate
      primary: green
      accent: teal
      toggle:
        icon: material/weather-sunny
        name: Switch to light mode
  features:
    - navigation.tabs
    - navigation.sections
    # navigation.indexes is intentionally OFF: it promotes a section's first page to be
    # the section's landing page, which silently absorbed "Lesson 1" into the "Lessons"
    # group (it stopped appearing as its own item). Without it, every lesson - including
    # the "Overview" pages of 1C and 4 - shows as an explicit, predictable nav item.
    - navigation.top
    - navigation.footer
    - toc.follow
    - search.suggest
    - search.highlight
    - content.code.copy
    - content.tooltips

markdown_extensions:
  - admonition
  - attr_list
  - md_in_html
  - tables
  - footnotes
  - toc:
      permalink: true
      # GitHub-compatible heading slugs so manual #anchors match in both the repo and site.
      slugify: !!python/object/apply:pymdownx.slugs.slugify
        kwds:
          case: lower
  - pymdownx.details
  - pymdownx.highlight:
      anchor_linenums: true
  - pymdownx.inlinehilite
  - pymdownx.snippets
  - pymdownx.tabbed:
      alternate_style: true
  - pymdownx.emoji:
      emoji_index: !!python/name:material.extensions.emoji.twemoji
      emoji_generator: !!python/name:material.extensions.emoji.to_svg
  - pymdownx.superfences:
      custom_fences:
        - name: mermaid
          class: mermaid
          format: !!python/name:pymdownx.superfences.fence_code_format

extra:
  social:
    - icon: fontawesome/brands/linkedin
      link: https://www.linkedin.com/in/lovedeep-singh-cloud-infra/
      name: Lovedeep Singh on LinkedIn
    - icon: fontawesome/brands/github
      link: https://github.com/ld-singh
      name: ld-singh on GitHub

extra_css:
  - stylesheets/extra.css

copyright: >
  Built by <a href="https://www.linkedin.com/in/lovedeep-singh-cloud-infra/">Lovedeep Singh</a>.
  Third-party tools belong to their respective projects.

# Pages are referenced at their synced paths under docs/.
nav:
  - Home: index.md
  - Overview & setup: course.md
  - Lessons:
      - "1 - Kubernetes GPU scheduling": portfolio-lab/01-k8s-gpu-platform/README.md
      - "1B - Queue scheduling (KAI)": portfolio-lab/01-k8s-gpu-platform/kai-scheduler/README.md
      - "1C - GPU sharing (HAMi, sim)":
          - Overview: portfolio-lab/01-k8s-gpu-platform/hami/README.md
          - "Scheduling sim (no GPU)": portfolio-lab/01-k8s-gpu-platform/hami/hami-scheduling-sim/README.md
      - "1D - GPU fleet scale simulation (Volcano)": portfolio-lab/01-k8s-gpu-platform/volcano-scale-sim/README.md
      - "2 - Slurm workload management": portfolio-lab/02-slurm-gpu-platform/README.md
      - "3 - GPU observability": portfolio-lab/03-observability/README.md
      - "4 - Inference serving": portfolio-lab/04-inference-serving/README.md
      - "5 - Cluster lifecycle (BCM-style)": portfolio-lab/05-bcm-style-cluster-lifecycle/README.md
      - "6 - Real GPU (one-rental capstone)":
          - Overview: portfolio-lab/real-gpu-session/README.md
          - "Setup scripts (bare GPU VM → k3s)": portfolio-lab/real-gpu-session/scripts/README.md
          - "Part A - GPU runtime path": portfolio-lab/01-k8s-gpu-platform/gpu-operator-real/README.md
          - "Part B - HAMi isolation (real GPU)": portfolio-lab/01-k8s-gpu-platform/hami/hami-isolation-realgpu/README.md
          - "Part C - HAMi + GPU Operator coexistence": portfolio-lab/01-k8s-gpu-platform/hami/hami-gpu-operator-coexistence/README.md
          - "Part D - Inference benchmark (real GPU)": portfolio-lab/04-inference-serving/inference-realgpu/README.md
          - "Part E - Slurm real GRES": portfolio-lab/02-slurm-gpu-platform/slurm-realgpu/README.md
      - "7 - Security (planned)": portfolio-lab/07-security/README.md
  - Deep dives:
      - kind (local cluster): portfolio-lab/01-k8s-gpu-platform/kind/README.md
      - KWOK (fake nodes): portfolio-lab/01-k8s-gpu-platform/kwok/README.md
      - fake-gpu-operator: portfolio-lab/01-k8s-gpu-platform/fake-gpu-operator/README.md
      - Control plane app: control-plane/README.md
  - Runbooks:
      - Overview: runbooks/README.md
      - Device plugin not advertising GPUs: runbooks/device-plugin-not-advertising-gpus.md
      - HAMi troubleshooting (GPU sharing): runbooks/hami-troubleshooting.md
      - k3s default runtime / containerd config: runbooks/k3s-default-runtime-containerd-config.md
      - GPU node not ready: runbooks/gpu-node-not-ready.md
      - GPU operator driver pod failing: runbooks/gpu-operator-driver-pod-failing.md
      - CUDA_VISIBLE_DEVICES debugging: runbooks/cuda-visible-devices-debugging.md
      - GPU memory pressure: runbooks/gpu-memory-pressure.md
      - GPU capacity planning: runbooks/gpu-capacity-planning.md
      - DCGM exporter no metrics: runbooks/dcgm-exporter-no-metrics.md
      - KAI queue starvation: runbooks/kai-scheduler-queue-starvation.md
      - Slurm job pending (GRES): runbooks/slurm-job-pending-reason-gres.md
      - Slurm node drained: runbooks/slurm-node-drained.md
  - Diagrams:
      - Architecture: diagrams/ai-factory-operations-architecture.md
      - GPU path to a pod: diagrams/gpu-path-to-pod.md
      - Fake vs real GPU validation: diagrams/fake-vs-real-gpu-validation.md
      - Slurm GPU job lifecycle: diagrams/slurm-gpu-job-lifecycle.md
      - BCM-style cluster lifecycle: diagrams/bcm-style-cluster-lifecycle.md
  - Lab notebook:
      - Overview: portfolio-lab/06-validation-reports/README.md
      - Fake vs real limitations: portfolio-lab/06-validation-reports/fake-vs-real-limitations.md
      - Local simulation report: portfolio-lab/06-validation-reports/local-simulation-report.md
      - GPU scale simulation (Volcano): portfolio-lab/06-validation-reports/gpu-scale-sim-validation.md
      - Slurm GRES validation: portfolio-lab/06-validation-reports/slurm-gres-validation.md
      - Real GPU validation: portfolio-lab/06-validation-reports/real-gpu-validation-report.md
      - HAMi isolation validation: portfolio-lab/06-validation-reports/hami-isolation-validation.md
      - HAMi + GPU Operator coexistence: portfolio-lab/06-validation-reports/hami-gpu-operator-coexistence-validation.md
      - Inference benchmark: portfolio-lab/06-validation-reports/inference-benchmark-report.md
  - About: about.md