LLM Wiki is a cross-platform desktop application that turns your documents into an organized, interlinked knowledge base — automatically. Instead of traditional RAG (retrieve-and-answer from scratch every time), the LLM incrementally builds and maintains a persistent wiki from your sources。
TypeScript
import { describe, it, expect } from "vitest"
import { parseFrontmatter } from "./frontmatter"
describe("parseFrontmatter", () => {
it("returns null + full body when content has no frontmatter", () => {
const r = parseFrontmatter("# Hello\nbody text")
expect(r.frontmatter).toBeNull()
expect(r.body).toBe("# Hello\nbody text")
expect(r.rawBlock).toBe("")
})
it("strips a basic frontmatter block and returns the body", () => {
const content = `---\ntype: entity\ntitle: "Foo"\n---\n\n# Body`
const r = parseFrontmatter(content)
expect(r.frontmatter).toEqual({ type: "entity", title: "Foo" })
expect(r.body).toBe("# Body")
})
it("rawBlock + body equals the original content for well-formed input", () => {
const cases = [
`---\ntype: x\n---\nbody`,
`---\ntype: x\n---\n\nbody with leading blank line`,
`# No frontmatter at all`,
`---\ntype: x\ntitle: "Foo"\ntags: [a, b]\n---\n\n# Heading`,
]
for (const content of cases) {
const r = parseFrontmatter(content)
expect(r.rawBlock + r.body).toBe(content)
}
})
it("documents that recovered fallback frontmatter may be lossy", () => {
const content = "```yaml\n---\ntitle: Fixed\n---\n```\n\n# Body"
const r = parseFrontmatter(content)
expect(r.frontmatter).toEqual({ title: "Fixed" })
expect(r.rawBlock).toBe("---\ntitle: Fixed\n---\n")
expect(r.rawBlock + r.body).not.toBe(content)
expect(r.body).toBe("# Body")
})
it("parses inline arrays", () => {
const r = parseFrontmatter(`---\ntags: [foo, bar, baz]\n---\nx`)
expect(r.frontmatter).toEqual({ tags: ["foo", "bar", "baz"] })
})
it("parses block-style arrays (the YAML `- item` syntax)", () => {
const content =
`---\nrelated:\n - foo.md\n - bar.md\n - baz.md\n---\nbody`
const r = parseFrontmatter(content)
expect(r.frontmatter).toEqual({ related: ["foo.md", "bar.md", "baz.md"] })
expect(r.body).toBe("body")
})
it("doesn't interpret YYYY-MM-DD scalars as Date objects (JSON_SCHEMA)", () => {
const r = parseFrontmatter(`---\ncreated: 2026-04-09\n---\nbody`)
expect(r.frontmatter).toEqual({ created: "2026-04-09" })
})
it("handles CJK/Unicode tags without quoting", () => {
const r = parseFrontmatter(
`---\ntags: [污水处理, 功能微生物, NOB优势菌]\n---\nbody`,
)
expect(r.frontmatter).toEqual({
tags: ["污水处理", "功能微生物", "NOB优势菌"],
})
})
it("strips quotes from string scalars", () => {
const r = parseFrontmatter(`---\ntitle: "Hello World"\n---\nx`)
expect(r.frontmatter).toEqual({ title: "Hello World" })
})
it("handles a frontmatter-only file with no body", () => {
const r = parseFrontmatter(`---\ntype: entity\n---\n`)
expect(r.frontmatter).toEqual({ type: "entity" })
expect(r.body).toBe("")
})
it("handles CRLF line endings", () => {
const content = "---\r\ntype: query\r\n---\r\n\r\nbody"
const r = parseFrontmatter(content)
expect(r.frontmatter).toEqual({ type: "query" })
expect(r.body).toBe("body")
})
it("returns null frontmatter when YAML is malformed but still strips the fence block", () => {
const content = `---\nthis is not: : valid\n malformed\n---\nbody`
const r = parseFrontmatter(content)
// YAML parser may or may not error on this; either way we
// shouldn't crash. If it errors, we return body without the
// fence block. If it parses, we return whatever it produced.
expect(typeof r.body).toBe("string")
expect(r.body).toBe("body")
})
it("matches the real Accumulibacter frontmatter (block array of paths)", () => {
const content =
`---\ntype: entity\ntitle: "Accumulibacter"\ncreated: 2026-04-09\nupdated: 2026-04-09\ntags: [microorganism, PAO, DPAO, phosphorus-removal, Betaproteobacteria]\nrelated:\n - wiki/entities/dpao.md\n - wiki/entities/paos.md\n - wiki/concepts/denitrifying-phosphorus-removal.md\nsources: ["research-foo-2026-04-09.md"]\n---\n\n# Body`
const r = parseFrontmatter(content)
expect(r.frontmatter?.type).toBe("entity")
expect(r.frontmatter?.title).toBe("Accumulibacter")
expect(r.frontmatter?.related).toEqual([
"wiki/entities/dpao.md",
"wiki/entities/paos.md",
"wiki/concepts/denitrifying-phosphorus-removal.md",
])
expect(r.frontmatter?.sources).toEqual(["research-foo-2026-04-09.md"])
expect(r.body).toBe("# Body")
})
it("matches the real Nitrospira frontmatter (all inline arrays, CJK tags)", () => {
const content =
`---\ntype: entity\ntitle: Nitrospira\ncreated: 2026-04-07\nupdated: 2026-04-07\ntags: [污水处理, 功能微生物, 亚硝酸盐氧化细菌, NOB优势菌]\nrelated: [nitrite-oxidizing-bacteria, nitrifying-bacteria, nh3-n]\nsources: ["research--2026-04-07.md"]\n---\n`
const r = parseFrontmatter(content)
expect(r.frontmatter?.type).toBe("entity")
expect(r.frontmatter?.title).toBe("Nitrospira")
expect(r.frontmatter?.tags).toEqual([
"污水处理", "功能微生物", "亚硝酸盐氧化细菌", "NOB优势菌",
])
expect(r.frontmatter?.related).toEqual([
"nitrite-oxidizing-bacteria",
"nitrifying-bacteria",
"nh3-n",
])
expect(r.frontmatter?.sources).toEqual(["research--2026-04-07.md"])
})
it("locates a frontmatter block wrapped in ```yaml … ``` and strips the closing fence", () => {
const content =
"```yaml\n---\ntype: entity\ntitle: \"Accumulibacter\"\n---\n```\n\n# Accumulibacter\n\nbody"
const r = parseFrontmatter(content)
expect(r.frontmatter?.type).toBe("entity")
expect(r.frontmatter?.title).toBe("Accumulibacter")
// Both the opening AND closing code-fence lines are gone so
// the body the renderer sees is balanced markdown. Without
// stripping, ReactMarkdown reads the orphan ``` as a
// never-closing code block and renders every heading /
// list / table below as raw source.
expect(r.body).not.toContain("```")
expect(r.body.trimStart().startsWith("# Accumulibacter")).toBe(true)
})
it("strips the closing fence even when the closing ``` is on the line directly after ---", () => {
// Real-user file pattern (硝化菌.md): the ``` closing fence
// lands immediately after the closing ---, with no blank
// line between them.
const content = [
"```yaml",
"---",
"type: entity",
'title: "硝化菌"',
"tags: [微生物]",
"---",
"```",
"",
"# 硝化菌",
"",
"## 基本信息",
"",
"- **类型**:自养型好氧细菌",
].join("\n")
const r = parseFrontmatter(content)
expect(r.frontmatter?.type).toBe("entity")
expect(r.frontmatter?.title).toBe("硝化菌")
expect(r.body).not.toContain("```")
expect(r.body).toContain("# 硝化菌")
expect(r.body).toContain("## 基本信息")
})
it("locates a frontmatter block prefixed by a `frontmatter:` key (LLM corruption)", () => {
const content =
"frontmatter:\n---\ntype: entity\ntitle: LSTM\n---\n\n# LSTM\n\nbody"
const r = parseFrontmatter(content)
expect(r.frontmatter?.type).toBe("entity")
expect(r.frontmatter?.title).toBe("LSTM")
})
it("does not pick up a `---` horizontal rule deep in the body as frontmatter", () => {
// The fallback-search window is small enough to skip horizontal
// rules used as section dividers below paragraphs of body text.
const content =
"# Heading\n\nfirst paragraph\n\nsecond paragraph\n\n---\n\nthird paragraph after rule"
const r = parseFrontmatter(content)
expect(r.frontmatter).toBeNull()
expect(r.body).toBe(content)
})
it("repairs `key: [[a]], [[b]]` (LLM-emitted invalid wikilink list) via retry", () => {
const content =
"---\ntype: entity\ntitle: LTTC\nrelated: [[riyuu-jiaoben]], [[zi-yuan-bu]], [[wai-wu]]\n---\nbody"
const r = parseFrontmatter(content)
expect(r.frontmatter?.type).toBe("entity")
expect(r.frontmatter?.related).toEqual([
"[[riyuu-jiaoben]]",
"[[zi-yuan-bu]]",
"[[wai-wu]]",
])
})
it("doesn't mangle a legitimate nested-array YAML value during repair", () => {
// Repair only fires when the FIRST yaml.load throws. Legitimate
// nested arrays parse on the first pass and are left alone.
const content = `---\nmatrix: [[1, 2], [3, 4]]\n---\nbody`
const r = parseFrontmatter(content)
// Nested arrays come through normalize as a JSON-stringified
// string per element, which is acceptable surface behavior — the
// important part is we didn't crash or corrupt the structure.
expect(r.frontmatter).not.toBeNull()
expect(r.body).toBe("body")
})
it("matches the real BOD entity frontmatter (quoted wikilink items in block array)", () => {
const content =
`---\ntype: entity\ntitle: BOD(生化需氧量)\ncreated: 2026-04-07\nupdated: 2026-04-07\ntags: [水质指标, 环境监测, 污水处理, 核心参数]\nrelated:\n - "[[nh3-n]]"\n - "[[soft-sensor-watertreatment]]"\n - "[[ai-effluent-water-quality-prediction]]"\n - "[[digital-twin-wastewater]]"\nsources: ["research-ai-2026-04-07.md"]\n---\n\n# Body`
const r = parseFrontmatter(content)
expect(r.frontmatter?.title).toBe("BOD(生化需氧量)")
expect(r.frontmatter?.related).toEqual([
"[[nh3-n]]",
"[[soft-sensor-watertreatment]]",
"[[ai-effluent-water-quality-prediction]]",
"[[digital-twin-wastewater]]",
])
expect(r.body).toBe("# Body")
})
})