DCAI 300-640 Hands-On Companion

Hands-on code, configurations, and visual labs for Cisco 300-640 DCAI v1.0 (Implementing Cisco Data Center AI Infrastructure).

Live site: https://frankellydeleon.github.io/dcai-code-study-sheet/ Companion theory sheet: https://frankellydeleon.github.io/dcai-study-sheet/ Official exam blueprint (PDF, 2025-11-05): https://learningcontent.cisco.com/documents/marketing/exam-topics/300-640-DCAI-v1.0_2025Nov5.pdf

Not affiliated with, endorsed by, or reviewed by Cisco Systems, Inc. Cisco, UCS, Intersight, Nexus, and APIC are trademarks of Cisco. This is an independently authored study aid; always verify current commands, API schemas, and product capabilities against official Cisco documentation before using anything here in production.

Last reviewed: 2026-09-14.

What this is

A single-page, objective-aligned practical companion to the DCAI 300-640 v1.0 blueprint. Every numbered domain and named sub-objective gets:

• A short theory recap
• An inline, original SVG diagram (topology, data flow, policy relationship, lifecycle, or decision tree) with a caption and a text alternative
• One or more annotated, runnable-where-possible code examples (Python, Bash, YAML, JSON, NX-OS, Intersight/UCS API patterns, or PromQL)
• Expected output, verification steps, and common failure modes for each example
• A language badge and a risk badge so you know what's safe to run locally right now versus what needs real hardware or a maintenance window
• A link back to the matching objective on the theory study sheet

The coverage matrix on the page itself maps every one of the 18 blueprint sub-objectives (across the four weighted domains) to its examples, diagram, and languages, and the left-hand navigation, search box, and language/risk filters let you jump straight to what you need.

Every objective is reachable by two equivalent deep-link forms: the descriptive #obj-1-1-style anchor used by this page's own navigation, and a short, reference-compatible #s11..#s44 alias (domain digit + sub- objective digit, no separator) matching the fragment convention used by the companion theory study sheet. Each objective's "Back to theory" link also points at the matching #sNN fragment on the theory site.

This project takes information-architecture inspiration (single scrolling page, objective-aligned snippets) from community DCAI/300-640-style study sheets, but all wording, code, diagrams, and visual design here are original. It contains no proprietary Cisco courseware, exam dumps, or copied content.

Site structure

index.html              -- the entire site (generated, see below)
assets/css/style.css    -- original stylesheet (light/dark theme, print, a11y)
assets/js/app.js        -- search/filter, copy buttons, theme toggle, nav drawer
assets/img/             -- favicon (SVG), social-preview source (SVG) and its
                            committed PNG render (assets/img/social-card.png)
scripts/build_site.py   -- dev-time generator: content data + HTML templates
scripts/validate.mjs    -- dependency-free structural validator (Node stdlib only)
scripts/check_examples.py -- dependency-free syntax checks for runnable examples

index.html is a plain static file with no framework and no build step required to serve or view it -- GitHub Pages (or any static file host) serves it as-is. scripts/build_site.py is a development convenience that regenerates index.html from structured content data; it is never executed by visitors or by GitHub Pages.

Example inventory

30 runnable/documented code examples across 18 objectives:

Language	Count	Notes
Python	14	Fully runnable with the standard library only (no pip installs required)
Bash	7	Three run locally as-is (troubleshooting flow, NVMe/multipath listing pattern, layered health-check script); four require specific hardware (NVIDIA GPU host for topology/Xid/dcgmi/nvidia-bug-report.sh, multi-GPU host for NCCL-tests, RDMA NICs for perftest) to reproduce real output
JSON	3	Intersight/UCS and APIC/ACI policy payload patterns, plus a conceptual AI Canvas investigation transcript -- syntactically valid, illustrative of documented product behavior
YAML	3	Lifecycle manifest, Kubernetes GPU pod spec, Prometheus scrape config
NX-OS	2	Documented QoS/DCB and ECMP show/config command patterns -- require a real or virtual Nexus switch to execute
PromQL	1	Illustrative alerting expression -- depends on which exporters are deployed

Every example carries one of four risk badges:

• Safe · Local -- runs immediately with no external dependencies, hardware, or credentials.
• Illustrative · Pseudocode -- demonstrates a pattern or control flow; placeholders must be replaced with real logic/credentials before use.
• Platform-dependent -- syntactically valid but requires specific hardware (GPUs, a Cisco switch/UCS domain/Intersight account) to produce real output.
• Production-impacting -- changes device/fabric state; requires a maintenance window and a rollback plan.

Safety conventions

• No real secrets, API keys, or credentials anywhere in this repository -- examples read credentials from environment variables (e.g., INTERSIGHT_API_KEY_ID).
• Example IP addresses use RFC 5737 TEST-NET documentation ranges (e.g., 203.0.113.0/24), never real infrastructure addresses.
• Cisco NX-OS, Intersight, and APIC/ACI commands and object/class names (fvTenant, power.Policy, priority-flow-control, etc.) are drawn from Cisco's own published configuration and API guides (linked as citations on each objective) and labeled Illustrative or Platform-dependent where they cannot be verified without live hardware -- no invented commands or endpoints.
• No runtime CDN dependencies: all CSS, JavaScript, fonts, and diagrams (inline SVG) are self-hosted in this repository.

Validation

Both validation scripts use only built-in language features (Node's standard library and Python's standard library) -- no npm install or pip install is required to run them.

# Regenerate index.html from the content data (only needed after editing content)
python3 scripts/build_site.py

# Structural validation: required anchors, no duplicate IDs, internal links
# resolve, every example has badges/copy button/expected output/verification/
# failure modes, every objective has a captioned+described diagram, the
# coverage matrix covers all 18 sub-objectives, required metadata (including
# an absolute, PNG social-preview image) is present, and there are no
# external CDN references.
node scripts/validate.mjs

# Syntax/sample checks for locally checkable examples (Python via `compile()`,
# Bash via `bash -n`, JSON via `json.loads`, YAML via PyYAML if installed --
# skipped gracefully, not failed, if PyYAML is absent). NX-OS/PromQL examples
# are intentionally skipped as documented CLI/query patterns, not local
# scripts. A curated set of deterministic, side-effect-free Python examples
# is additionally *executed*, with real stdout diffed against the documented
# "Expected output" text, so a stale hand-typed expectation can't recur.
python3 scripts/check_examples.py

Both scripts exit non-zero on failure and are safe to wire into CI.

License

MIT.