kb-search

/kb-search -- Knowledge Base Search

You search across id8Labs knowledge bases for articles matching a query. Faster than /kb-query -- this finds what exists, not synthesizes answers.

Arguments: $ARGUMENTS

---

Route by Arguments

Parse $ARGUMENTS:

Input	Action
<term>	Search all KBs for articles matching the term
<term> --kb <name>	Search a specific KB only
<term> --type <concepts|sources|connections>	Filter by article type
(empty)	Show KB stats (article counts per KB)

---

Setup

1. Read ~/Development/id8/knowledge/manifest.json
2. For each target KB, read wiki/_index.md

---

Search Process

Step 1: Load Indexes

For each target KB:

• Read {kb}/wiki/_index.md
• Parse all article entries (the [[wikilinks]] and their descriptions)
• Build a flat list of: {kb_name, article_path, display_name, description, type}

Step 2: Fuzzy Match

Search the flat list against the query term:

• Match against display name, description, and article type
• Case-insensitive substring matching
• Also search frontmatter tags if articles have them (read the first 10 lines of each matching candidate)
• Rank by relevance: exact title match > title contains > description contains > tag match

Step 3: Deep Search (if few results)

If fewer than 3 matches from index search:

• Use Grep to search inside wiki/**/*.md for the term
• Report files that contain the term with a snippet of context
• This catches content that isn't surfaced in the index descriptions

Step 4: Report

Display results grouped by KB:

## Search: "{term}"

### research (N matches)
- [[concepts/constraint-as-consciousness|Constraint as Consciousness]] -- F-005, write-protection as precondition
- [[registries/findings|Findings Registry]] -- contains 2 mentions

### design (N matches)
- [[concepts/shadow-as-border|Shadow-as-Border]] -- Vercel's 0px 0px 0px 1px pattern
- [[sources/vercel|Vercel]] -- shadow-as-border technique

### memory (N matches)
- (none)

Total: N articles across M knowledge bases