architecture-diagram

Interactive Architecture Diagrams

Build single-file, drop-in HTML pages that let workshop attendees, clients, or new team members click through how a system works — step by step, with animated data packets flowing between nodes, payload details on the side panel, and toggleable modes (offline/online, dev/prod, v1/v2). Battle-tested on the Eskadra Bielik RAG workshop; reuse the design system everywhere else.

This skill produces diagrams in a dark, didactic-first aesthetic — nodes are bounded, wires are gentle quadratic curves, packets glow only on the active step. No "AI slop" rainbow gradients. The CSS tokens are exposed in assets/css-tokens.css if you want to rebrand.

---

When to reach for this skill

Situation	Use this skill?
"Show me how the auth flow works" (interactive, for a workshop)	Yes
"Design the RAG pipeline before we build it" (planning new service)	Yes
"Map our microservices and how they communicate"	Yes
"Visualize the CI/CD pipeline for onboarding docs"	Yes
"Show the data flow through our ETL pipeline"	Yes
"Draw an agentic multi-agent system with tool calls"	Yes
"Build me an architecture mockup we can iterate on with the team"	Yes
"Draw a sequence diagram for the PR" (static, goes in markdown)	No — a Mermaid sequence diagram inline is simpler
"Add a diagram to a slide deck or PDF" (static, not interactive)	No — this produces interactive HTML, not images or slides
"I just need a static boxes-and-arrows topology" (no flows/steps)	No — a Mermaid flowchart is enough

The differentiator is interactivity + sequenced data flow. If the value is "click through it and watch what happens", this skill applies. If the output needs to be a static image, a slide, or inline markdown, this is the wrong tool.

---

The mental model

Every diagram has four things:

1. Nodes — services, datastores, users, queues, external systems. Each has a role (color) and metadata (tech stack, port, deployment target).
2. Flows — named scenarios the user can pick (e.g. RAG Query, Direct Query, Ingest, Auth). Each flow is an ordered list of steps.
3. Steps — {from, to, route, payload, desc, chips}. Each step lights up one wire and one target node.
4. Modes — orthogonal toggle (offline/online, dev/prod, v1/v2). Modes can:
   • hide/show entire nodes (e.g. Seed only exists offline)
   • rename a node (Qdrant → BigQuery)
   • swap payload bodies, ports, auth headers, latency chips

Modes are NOT alternative flows. Flows describe scenarios ("user asks a question"); modes describe deployment shape ("on Docker" vs "on Cloud Run").

---

Workflow when invoked

Step 1 — Capture intent

If the user has uploaded a SOLUTION.md, README, OpenAPI spec, or any architecture description, read it first before asking questions. Extract:

• Service names + tech stack + ports
• Deployment modes (if any)
• Named endpoints/operations and what they do (these become flows)
• For each operation, the chain of internal service calls (these become steps)

Then ask ONLY the gaps. Don't ask things already in the doc.

If there's no doc, ask the user briefly:

• "What system are we drawing?"
• "What are the main flows you want to show? (e.g. login, search, checkout)"
• "Any mode toggles? (offline/online, dev/prod, v1/v2)"
• "Who's the audience? Workshop attendees? Engineers? Clients?"

Step 2 — Plan the topology on paper first

Before writing code, sketch on paper (literally in your head or a scratch file) WHICH services exist, WHERE they sit relative to each other, and WHICH flows connect WHICH nodes. Avoid wire crossings — this is the single biggest readability win. See references/flow-design-patterns.md for layout heuristics.

Quick rules:

• Orchestrator/API gateway in the middle, dependencies fanning out
• User on the far left (entry point), datastores on the right (exit), services in between
• Read-only services above, write-heavy services below (intuitive: data flows down into them)
• Optional/one-shot services (seed jobs, cron) tucked in a corner

Step 3 — Build from the template

Copy assets/template.html to the working directory. The template is a fully working drop-in: dark theme, side panel, player controls (play/pause/step), keyboard shortcuts, animated SVG packets, mode toggle. You only edit two regions:

1. The nodes block in HTML — one .node div per service with data-id, data-role, style="left/top", label, tech, port.
2. The flows object in JS — one entry per scenario, each with an ordered steps[] array.

Don't reinvent the CSS, the player, the SVG wire renderer, or the side panel — they all work.

Step 4 — Validate before showing the user

After generating, do this self-check:

• Open the file mentally: does every flow's step sequence make narrative sense? (Step 1 sets up, last step delivers a result to the user.)
• Do wires cross unnecessarily? If yes, reposition nodes (try moving the late-arriving dependency away from the early ones).
• Are there flows where a single node is the to of two consecutive steps? That's fine if it's a loop, but check the description explains why.
• Mode swap: pick the most different node between modes (usually the database) and verify ALL its mode-dependent fields are filled in (label, tech, port, payload, chips).
• Did you generate the companion architecture.md? (Required — see Step 6.)

If puppeteer is available in the environment, take a screenshot to spot-check layout. Code in references/screenshot.js.

Step 5 — Output the files

Always two files:

• architecture.html — the interactive diagram
• architecture.md — text description (components, flows step-by-step, mode differences, scenarios for the workshop)

Save to the current working directory. Platform-specific overrides:

• Claude.ai — use /mnt/user-data/outputs/ instead
• Claude Code / Gemini CLI / OpenCode / Copilot CLI — save to the current working directory (default)

Both should be self-sufficient — someone can read the markdown without opening the HTML, and vice versa.

Step 6 — Present the output

Tell the user where the files were saved and how to open them (open architecture.html on macOS, xdg-open architecture.html on Linux, or just open in browser). If present_files is available (Claude.ai), call it to display the HTML first, then the markdown.

---

File reference

File	When to read
assets/template.html	Always. This is the starting point you copy and modify.
assets/css-tokens.css	If user asks to restyle (e.g. light theme, different brand colors). Standalone tokens you can splice into existing CSS.
references/flow-design-patterns.md	When planning a new system from scratch. Heuristics for node placement, naming flows, sequencing steps, choosing modes.
references/component-library.md	When building the nodes list. Copy-paste node snippets for User/API/DB/Queue/Cache/LLM/Cron/External/etc. with proper roles and icons.
references/screenshot.js	When you want to spot-check the result before showing the user. Run with node if puppeteer-core is installed.
examples/rag-eskadra-bielik.json	Reference example: a real RAG system with 5 flows and offline/online modes. Read for inspiration on payload formatting and chip choices.
examples/auth-oauth2.json	Reference example: an OAuth2 + session flow with dev/prod modes. Shows redirect flows and external IdP nodes.
examples/event-driven.json	Reference example: event-driven microservices with Kafka, showing async fan-out flows.
examples/cicd-pipeline.json	Reference example: CI/CD pipeline with GitHub Actions, staging/prod modes, rollback flow.
examples/agentic-tool-calling.json	Reference example: agentic AI system with tool calling, single-agent vs multi-agent modes.

---

Quick reference — node roles and colors

Hardcoded in CSS variables in template.html. Pick one per node:

Role	Color	Use for
user	mint #46f9b8	Browser, mobile client, end user, CLI user
orch	sky #6aa9ff	API gateway, orchestrator, BFF, controller
compute	magenta #ff6b8a	LLM, ML model, expensive compute
embed	amber #ffc14d	Embedding service, transformer, preprocessor
vector	violet #b388ff	Database (any), vector store, cache, blob storage
seed	orange #ff9457	Cron, one-shot job, migration, seed, scheduled task

If you need more roles, add CSS variables. Keep the count ≤ 6 — past that, the legend becomes a colorblind nightmare.

If your system has e.g. two distinct datastores (Postgres + Redis), still use vector for both but distinguish via the icon and label, not by inventing a 7th color.

---

Visual states — what the viewer sees on each step

This is the most important didactic mechanic in the diagram. Understand it before modifying the JS or CSS, or you'll break the "click a flow tab, see the whole path" intuition.

When a user picks a flow tab and lands on step N, every node and wire is classified into one of these states:

Element	State	Looks like	When
Node	active	Full ring + glow + step-number badge	Target of current step (e.g. LLM when step says "Orch → LLM")
Node	active-from	Soft ring, no badge	Source of current step
Node	participant	Near-full opacity, slightly desaturated	Appears anywhere else in this flow, just not in this step
Node	dimmed	25% opacity, washed out	Not in this flow at all (e.g. seed during a login flow)
Node	hidden	Invisible	Mode-excluded entirely (e.g. seed node in online mode)
Wire	active	Bright role color, glow, packet animating	The wire of the current step
Wire	preview	Role color at 50% opacity, dashed	Belongs to this flow but not the active step — shows the full path ahead
Wire	muted	~8% opacity	Doesn't belong to this flow

Why participant matters. Without it, only the currently-active node has full color; every other node (including the "source" of the current step and any nodes used in later steps) looks dead. The viewer can't tell that the LLM is going to be used in step 6 just by looking at the canvas — they have to click through all 5 prior steps to discover it.

With participant, the very first click on a flow tab shows the viewer: "here are the players, here are the wires connecting them, and here's where the first step happens". They can scan the whole story before pressing play. This is the entire point of the diagram.

Implementation gotcha. Both the JS (applyStep) and the CSS (.node.participant, .wire.preview) need to be consistent. If you add new node states, update both. Test by walking through every step of every flow and verifying the source node has the right state — common bug is leaving the source dimmed (looks dead) instead of active-from or at least participant.

---

Interaction model

The diagram supports five ways to interact, in increasing order of viewer expertise:

Interaction	What happens	Who uses it
Click flow tab	Jump to step 1 of that flow, side panel updates	Everyone
Click Next/Prev (or →/←, Space)	Advance/rewind one step	Workshop attendees walking through
Click a node	Jump to the first step where this node appears	Curious viewers exploring
Drag a node	Reposition the node on the canvas; wires redraw live	Layout tweakers / presenters
Fullscreen (F)	Expand canvas to fill the screen; Escape to exit	Workshop presenters

Node click vs drag. The template uses a 5px movement threshold to distinguish a click (jump to step) from a drag (reposition). Pointer events handle both mouse and touch. Dragged positions snap to a 32px grid matching the canvas background grid for clean alignment.

Node click is the easiest to miss when modifying the skill. Three things must be in place:

1. cursor: pointer on .node (visual affordance — "this is clickable")
2. :hover state that reveals the role color (confirms interactivity)
3. The JS handler that does findIndex on the current flow's steps and jumps there

If a node isn't part of the current flow (.dimmed state), clicking it shakes the node instead of doing nothing. Silent ignore is bad UX — the viewer thinks the page is broken.

Keyboard accessibility: nodes have tabIndex=0, Enter/Space triggers the same action as click.

Layout persistence. When a viewer drags nodes, their positions are saved to localStorage (keyed by page title). A "reset layout" button (↻) appears in the control bar after any drag. Press R or click the button to restore original positions.

Progress bar. A thin progress bar below the player controls shows how far into the current flow the viewer is. Updates on every step change.

Don't break this when adding features. It's the only way for a viewer to think "wait, where is the LLM used?" and find out instantly. Without it, they'd have to step through all 8 steps just to discover one node's role.

---

Quick reference — payload format

Payloads in the side panel are monospace + minimal syntax highlight. Keep them realistic but trimmed — show the shape, omit boring fields:

{
  // good
  payloadOffline:
`POST /ask
{
  "query": "Ile kosztuje parking?",
  "limit": 3,
  "temperature": 0.7
}`,

  // bad — too verbose
  payloadOffline:
`POST /ask HTTP/1.1
Host: localhost:3000
Content-Type: application/json
User-Agent: Mozilla/5.0...
Accept: */*
Connection: keep-alive

{ ...full 50-line body... }`
}

Use 4-space indent, no tabs. Use // comments for inline annotations.

If a step is the same in both modes, use payload (single key). If they differ, use payloadOffline + payloadOnline (or whatever the mode names are).

---

Quick reference — chips (metadata badges)

Chips below the payload are short metadata labels. Two semantic types auto-style themselves:

• Latency (matches /ms|s$|min/) — gets a stopwatch icon, e.g. "~150ms", "5-30s", "timeout 120s"
• Size/count (matches /dim|×|loaded|rules|count/) — gets a database icon, e.g. "768 dim", "38 loaded", "×N"
• Other — neutral, e.g. "COSINE", "baseline", "idempotent", "UUID5"

Maximum 3 chips per step. They're hints, not specifications.

---

Naming conventions

Thing	Convention	Example
Flow key in JS	snake_case	ask, ask_direct, ingest_text
Flow display name	"Title Case (METHOD /route)"	"RAG Query (POST /ask)"
Node data-id	snake_case, short	orch, vector, embed
Mode key	snake_case, descriptive	offline/online, dev/prod, v1/v2
Step title	Polish or English, depending on workshop language. Consistent within one file.	"Pytanie z UI" or "Question from UI"

---

Common pitfalls

</script> in payloads breaks everything. If a payload template literal contains the string </script> (e.g. showing an HTML snippet), the browser's HTML parser treats it as the closing tag for the main <script> block — killing all JS. Escape it as <\/script> or </script> inside template literals. This also applies to any </style> or similar closing tags in payload strings.

Source node looks dead. If after refactoring the source node of the current step appears dimmed instead of active-from, you regressed the participant logic. See the "Visual states" section. Test by clicking each flow tab in turn and verifying every node that EVER appears in the flow stays visible at near-full opacity.

Wire crossings. Caused by bad node placement, not bad code. If two steps cross, swap node positions. The template auto-curves wires in alternating directions per step index, which helps but doesn't fix bad layout.

Too many flows. More than 6 flows clutters the tab bar. If you have 8+, group them: one diagram per group (e.g. "read flows" vs "write flows" as two separate HTML files).

Modes that aren't really modes. If the toggle changes the entire topology (different services, different protocols), it's not a mode — it's a second system. Build two HTML files.

Static screenshots beat live diagrams sometimes. If the diagram will end up in a PDF report, a slide deck, or a printed handout, this skill is the wrong choice — produce a static diagram instead. This skill is for interactive consumption in a browser.

Forgetting the .md. The HTML is the showpiece, but the .md is what people read on GitHub, paste into Slack, and grep for endpoint names. Always generate both.

Polish vs English. Match the workshop language. Bilingual settings typically use Polish for internal workshops and English for client deliverables. Ask if unsure.

---

Extending the template

If a user asks for something the template doesn't support, here's the difficulty ranking:

Request	Difficulty	Where to edit
New node	trivial	Add .node div, position via left/top %
New flow	trivial	Add entry to flows object
New mode (3rd one)	small	Add button to .modepick, extend state.mode switch, add payloadX fields to steps
Different brand colors	small	Edit CSS variables in :root (see assets/css-tokens.css)
Light theme	built-in	Click the moon/sun button or press T. To make light the default, add class="light" to <body>.
Drag-and-drop nodes	built-in	Already supported — drag any node; positions snap to 32px grid and persist in localStorage.
Fullscreen mode	built-in	Press F or click the expand icon. Press Escape to exit.
Layout reset	built-in	Press R or click the ↻ button (appears after any drag).
Progress bar	built-in	Thin bar below player shows step position in current flow.
Parallel steps (two simultaneous wires)	medium	Group steps as {parallel: [step, step]}; player advances both
Branching flows (if-then-else)	hard	Currently linear. Either fork into separate flows or add a branch field with UI for picking the branch.
Saving state in URL	small	Hash-based: #mode=online&flow=ask&step=3 — parse in boot, update on changes

Don't promise features the template doesn't support without flagging the work involved.

---

Iteration tips

When the user gives feedback like "the diagram is too cramped" or "step 5 is confusing":

• "Too cramped" → spread nodes wider; you have 100% width to work with. Don't go below 12% horizontal gap between adjacent nodes.
• "Confusing step" → the step description is doing too much. Split into two steps OR rewrite to focus on ONE handoff (X sends Y to Z, that's it).
• "It doesn't match how it really works" → ask for the actual code path. Don't guess; the diagram has zero value if it lies about reality.
• "Make it pop more" → resist. Didactic clarity > visual punch. Maybe one accent element (a glow on the result node when flow completes) but stop there.
• "Add a legend" → already there. If they want it bigger, increase the .legend font-size and gap.