Knowledge base — /learn with full-text search

The knowledge base at /learn is ImtehanHub's library of long-form study content — chapter explainers, formula sheets, exam-style worked examples, and concept guides. Every entry is public, searchable, and bilingual where applicable. Think of it as the "encyclopedia" sibling of the blog.

TL;DR — Open /learn, search or browse by class / subject / type. Each entry opens at /learn/:id/:slug. Entries are richer and more permanent than blog posts; they are indexed and refreshed periodically.

How it differs from the blog

	Blog	Knowledge base
Cadence	Reverse-chronological news	Permanent reference
Length	500-1500 words	800-3000 words
Editing	Rarely edited after publish	Refreshed when curriculum changes
Surface	/blog, /blog/:slug	/learn, /learn/:id/:slug
Tags	Free tags	Tied to chapter / subject / topic
Search	Tag + title	Full-text
RSS	Yes (/feed.xml)	Combined into /feed.xml

A blog post on "Class 10 Physics revision" is news ("here's how to plan your week"). A knowledge-base entry on "Newton's laws of motion explained" is reference (everything you need to know about that concept, kept current).

Where knowledge-base data lives

aiKnowledgeBase/{id} in Firestore (the prefix aiKnowledgeBase is historical — the collection holds all knowledge-base entries, AI-assisted or not):

{
  id: 'auto',
  slug: 'newtons-laws-explained',
  title: 'Newton\'s laws of motion explained',
  urduTitle: 'نیوٹن کے قوانینِ حرکت',
  excerpt: 'A complete student-friendly explainer...',
  body: { /* TipTap JSON content */ },
  bodyEn: { /* English version */ },
  bodyUr: { /* Urdu version */ },
  type: 'concept-guide',       // 'concept-guide' | 'formula-sheet' | 'worked-example' | 'chapter-explainer'
  classId: 'c9',
  subjectId: 'c9-physics',
  chapterIds: ['c9-physics-kinematics'],
  tags: ['newton', 'mechanics', 'force'],
  authorId: 'admin-uid',
  status: 'published',
  publishedAt: '2026-04-15T...',
  updatedAt: '2026-05-01T...',
  searchBlob: 'newtons laws motion force mass acceleration...',  // pre-computed lowercase blob for fast substring search
}

The searchBlob field is the secret to fast full-text search — see below.

Browsing

/learn shows entries grouped by:

• Class — pick Class 5, 6, ..., 2nd Year FA/FSc.
• Subject — drilling further.
• Type — concept guide, formula sheet, worked example, chapter explainer.

Each entry card shows title (English + Urdu), excerpt, type badge, and last-updated date.

Full-text search

The search bar at the top of /learn does substring matching against the pre-computed searchBlob field. Matching:

• Lowercase-folded.
• Common-synonym substituted ("equals" ↔ "=", "kinematics" ↔ "motion equations").
• Result is sorted by relevance (number of match hits + recency boost).

Performance: with ~500 entries indexed, search is 5-10 ms per keystroke because the blob is pre-computed at write time, not at query time. The query is a simple array-contains on the blob (split into chunks).

This is the same fuzzy-search pattern used on the Sitemap page — see the developer notes in Batch 6 for the implementation.

Reading an entry

Each entry opens at /learn/:id/:slug. The reading view:

• Renders TipTap content with prose-friendly typography.
• Shows the type badge, class, subject, and relevant chapters.
• Has a "Take a test on this chapter" CTA at the bottom (jumps to the configure-test page).
• Cross-links to related entries.

JSON-LD Article is emitted for SEO. BreadcrumbList includes Home › Knowledge › Class 9 › Physics › Newton's laws.

Bilingual entries

Entries can have both English and Urdu bodies. The reader's language toggle picks which to render. Switching languages mid-entry re-renders without losing scroll position.

If only one language exists, the toggle is disabled with a "Translation pending" message.

Who writes knowledge-base entries

• The team — most entries are admin-authored.
• Promoted community members — chapter explanations from the community module that cross the eligibility threshold and pass admin review can be promoted into the knowledge base as authored entries.
• AI-assisted seeding — admins can use BYOK to draft an entry from chapter content. The draft always goes through admin review before publish.

Editing and freshness

Knowledge-base entries are refreshed when:

• A board changes the syllabus (Federal Board annual update).
• A textbook is updated to a new edition (page references shift).
• Better explanations or examples emerge.

The updatedAt field tracks the last refresh and shows in the reading view ("Last updated: 1 May 2026").

Common questions

Can I bookmark a knowledge-base entry?

Bookmarks are per-question, not per-knowledge-entry. Use the Save for later button on entry cards (added 2026-04 onwards) — saved entries appear in your Profile.

Why is the search not perfect?

Substring + synonym substitution is fast but not semantic. A query like "force" finds entries with "force" in the blob; it does not find entries about "Newton's laws" unless they also mention force. For semantic search, the AI knowledge-base API (BYOK-powered, V2) is the more flexible path; see Module 23 in the project's roadmap.

Are entries indexed by AI search engines?

Yes. The <noscript> body is rendered server-side via the postbuild SEO pipeline. GPTBot / ClaudeBot / PerplexityBot can read the full content without JavaScript. See the SEO docs (lands in Batch 6).

What's the difference between a chapter explanation in the community module and a knowledge-base entry?

Both are long-form study aids. Community chapter explanations live in the community surface and are voted by community members. Knowledge-base entries are admin-published reference content — typically more carefully edited and more permanent. The two surfaces cross-link.

Next

• Blog — news-format content.
• Sitemap page — discover everything else.
• Search and filters — search across the whole site.