[Docs Accuracy] Pipeline parallel nodes with merge-branch patterns in Tech-Hub pg - #775

Walkthrough

The documentation clarifies routing inputs, router behavior, node data flow, and fallback tags. It adds a Tech Hub page for merging parallel branches and links it from related documentation and site navigation. It also updates tags formatting and changelog references.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Merge Risk: 🔵 Low · up to 42e80

This documentation update clarifies pipeline behavior and adds merge guidance. One example may repeatedly defer processing for empty branch output, and a minor changelog wording issue remains; these are bounded documentation risks.

Suggested reviewers: snopoke

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

_{Comment @coderabbitai help to get the list of available commands.}

Docs review

Structurally this is a good change — the concepts/ → tech-hub/ split is the right call under this repo's page-type contract, and the residual parallel.md still reads as a coherent concept page.

Link and anchor integrity — verified

strict: true means broken refs fail the build, so I checked each new/changed link by hand against actual heading slugs. All resolve:

• merging_parallel_branches.md → parallel.md#which-input-a-node-receives and #uneven-branches — both headings survive the trim (parallel.md:60,73).
• changelog.md → merging_parallel_branches.md#merging-branches-that-are-optional — resolves.
• routers/index.md:37 → #the-default-output — heading exists at line 12.
• nodes.md → ../../concepts/participant_data.md — resolves (verbose vs ../participant_data.md, but matches the pre-existing pattern already on that page, so not a new inconsistency).
• Searched docs/ for the removed #optional-parallel-branches anchor — zero stale references. The move was clean.

⚠️ Caveat: I could not run uv run zensical build --clean or the markdownlint hook in this environment (both blocked by sandbox permissions), so treat the above as manual verification, not a green build. Please confirm CI is clean before merging. Manual check against the active .markdownlint.yaml ruleset (MD013/MD024/MD038/MD040/MD051/MD059) found no violations in the changed lines.

Confirmed: the two code bugs are real

Both were already flagged, so not re-reporting — but I verified the mechanism against src/python_node/__init__.py:147, whose docstring states wait_for_next_input() "will abort the current execution." It raises rather than returning, which is what makes bug #1 bite: on NodeD's first run b is set and c is None, so b is None and c is None is False, nothing aborts, and the function falls straight through to return f"{b}\nNone" — a literal "None" in the output. or is the correct fix.

Findings (2 inline comments, both non-blocking)

1. docs/tech-hub/python_node.md:68 — the new tip box sits ~7 lines below the rendered wait_for_next_input docstring, which contains the same not a and not b truthiness bug. Fixing only the new page leaves this one page rendering both the corrected and uncorrected pattern. Recommend fixing src/python_node/__init__.py:157 in this PR too.
2. docs/how-to/routers/index.md:37 — the :default example is a three-part tag, but the convention block four lines above states a two-part format. Suggested wording to make the optional segment explicit.

Content accuracy

The Extract Structured Data ("output replaces the input") and Update Participant Data ("passthrough") rewrites are consistent with each other, and I found no other page still asserting the old "Extract Structured Data is a passthrough" claim — so the correction doesn't leave a contradiction anywhere. The router_nodes.md Conversation Context rewrite matches the LLM Router / Static Router subsections later on that same page. I can't verify any of this against the OCS source from here, so the behavioural claims still want a developer's eye — as your PR description asks for.

Two notes, no action needed

• Dropping the grid cards block for a plain "See also" list is more consistent with repo convention, not less — grid cards is otherwise used only on landing/index pages, never on a regular concept page.
• The changelog.md mid-bullet line break is a lazy continuation; it renders as a single paragraph, so it's diff noise rather than a rendering change.

Nit, take or leave: ## Related pages on the new page vs ## Related Pages in template_and_email_nodes.md:95. Tech-hub is already mixed on this, so low priority.