Added ordered how_item steps for 19 older converters and fixed PDF template pollution plus near-duplicate URL differentiation. | Dev Logs

Date: 2026-09-06 01:10 Summary: Added ordered how_item steps for 19 older converters and fixed PDF template pollution plus near-duplicate URL differentiation. Visibility: people

[question] 处理问题：“仍未做（需你定夺）

~20 个老转换器缺有序 how_item_*（量大，当前 lint 不拦） PDF 模板污染与近重复 URL（属 PDF 簇，不在本轮）”

[try to solve]

Context

After the text-tools capability/SEO pass, two deferred items remained: (1) about nineteen older converter/SEO tools had how_body but no ordered how_item_1…n, so How sections rendered as prose only; (2) PDF cluster leftovers — write-PDF / add-text still carried extract/OCR template copy, and extract-text vs pdf-to-markdown plus edit-pdf vs turn-editable were near-duplicate in capability and messaging.

Process

1. Confirmed English how_item_1…4 already existed for the 19 slugs and that pages used howItemCount: 4 (except markdown-to-html, which still used a hand-written How block without <ol>).
2. Wired markdown-to-html How to four ordered list items from tool_markdown_how_item_*.
3. Completed ten-locale how_item_* for all 19 tools (native actionable steps; inventory check: 19 × 10 locales × 4 keys present).
4. Differentiated pdf-to-markdown code: each page becomes ## Page N plus text, joined with ---, while extract-text stays plain .txt.
5. Rewrote write-pdf and add-text masters (and locales) to match real behavior: write creates from textarea only; add-text stamps on every page including scans; removed false OCR/notext framing and unused notext page binding.
6. Tightened FAQ/rules for extract vs markdown and edit vs turn (primary download .pdf vs .txt).
7. Ran merge:tools, spot lint:tool-page, and npm run build:site (205 tools × 10 langs prerender).

Root cause / analysis

Older tools predated the ordered How convention; lint did not require how_item_*. PDF pages shared a multipurpose template, so create/stamp tools inherited extract/OCR strings. extract vs markdown and edit vs turn shared the same pdf.js/pdf-lib path with only thin output/CTA differences — better to differentiate copy and light Markdown structure than to merge URLs without redirects.

Solution

• Ordered How on the 19 converters/SEO tools across ten locales; markdown-to-html list wired in the page.
• write-pdf / add-text pollution cleaned; add-text no longer binds unused err_notext.
• pdf-to-markdown outputs ## Page N drafts; IG/FAQ state the .txt vs .md and edit vs turn download differences.
• Site rebuild refreshed prerendered tool HTML (e.g. base64 How <ol> visible under /_pages/).

Notes / boundaries

• Did not 301-merge PDF near-duplicates; kept separate URLs with clearer product differentiation.
• Dead template keys (choose_file on write-pdf, unused err_* ) may still exist as non-rendered strings; visible empty/how/rules/faq were corrected.
• Temp inject/fix scripts under scripts/tmp/ were removed after use; audit-prompt-i18n.mjs may remain from earlier work.

[actions]

• Updated src/pages/markdownToHtmlPage.ts, src/pages/pdfToMarkdownPage.ts, src/pages/addTextToPdfFilePage.ts
• Updated i18n shards for 19 how_item tools + write-pdf / add-text / pdf-to-markdown / extract-text / edit-pdf / turn-editable (ten locales where needed)
• Ran npm run merge:tools and npm run build:site