Skip to content

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

fix: tighten missing-page existence matching

nash_su11c755302fa5
diff --git a/src/lib/lint.test.ts b/src/lib/lint.test.ts
index f7512703..139af7be 100644
--- a/src/lib/lint.test.ts
+++ b/src/lib/lint.test.ts
@@ -137,7 +137,7 @@ describe("runSemanticLint — missing-page false positives (#537)", () => {
"维特根斯坦在概念页中被引用但未建立独立实体页。",
"PAGES: 概念页.md",
"---END LINT---",
- "---LINT: missing-page | warning | 阿德勒实体页缺失---", // name + suffix, page exists
+ "---LINT: missing-page | warning | 缺失页面:阿德勒---", // standard prefix, page exists
"阿德勒被引用但缺少页面。",
"PAGES: 概念页.md",
"---END LINT---",
@@ -158,6 +158,32 @@ describe("runSemanticLint — missing-page false positives (#537)", () => {
expect(missingPages).toHaveLength(1)
expect(missingPages[0].page).toBe("尼采")
})
+
+ it("does not suppress an unrelated finding that merely contains an existing short title", async () => {
+ const pages = [
+ makeFileNode("AI.md", "---\ntitle: AI\n---\n# AI"),
+ makeFileNode("overview.md", "References [[FAIR data governance]]."),
+ ]
+ mockListDirectory.mockResolvedValue(pages.map((p) => p.node))
+ mockReadFile.mockImplementation(async (path) => {
+ const match = pages.find((p) => p.node.path === path)
+ return match?.content ?? ""
+ })
+ mockStreamChat.mockImplementation(async (_c, _m, cb) => {
+ cb.onToken([
+ "---LINT: missing-page | warning | FAIR data governance---",
+ "The concept has no dedicated page.",
+ "PAGES: overview.md",
+ "---END LINT---",
+ ].join("\n"))
+ cb.onDone()
+ })
+
+ const results = await runSemanticLint("/project", fakeLlmConfig())
+
+ expect(results).toHaveLength(1)
+ expect(results[0].page).toBe("FAIR data governance")
+ })
})
describe("runSemanticLint — activity & early returns", () => {
diff --git a/src/lib/lint.ts b/src/lib/lint.ts
index ee772a67..a88fb947 100644
--- a/src/lib/lint.ts
+++ b/src/lib/lint.ts
@@ -5,6 +5,7 @@ import type { FileNode } from "@/types/wiki"
import { useActivityStore } from "@/stores/activity-store"
import { getFileName, getRelativePath, normalizePath } from "@/lib/path-utils"
import { buildLanguageDirective } from "@/lib/output-language"
+import { normalizeReviewTitle } from "@/lib/review-utils"
export interface LintResult {
type: "orphan" | "broken-link" | "no-outlinks" | "semantic"
@@ -67,16 +68,15 @@ function normalizeLinkTarget(target: string): string {
* and compatibility forms so CJK / full-width variants compare equal.
*/
function normalizeForExistence(s: string): string {
- return s.normalize("NFKC").trim().toLowerCase()
+ return normalizeReviewTitle(s).normalize("NFKC").trim().toLowerCase()
}
/**
* Decide whether an LLM `missing-page` finding actually refers to a page that
* already exists. The LLM does not reliably cross-reference the file list, so it
- * flags entities whose page is already present — especially in non-English wikis
- * where its free-form titles ("维特根斯坦", "阿德勒实体页缺失", "缺失页面:X") don't
- * match a fixed prefix. Drop the finding when a known page name matches exactly
- * or appears within the finding title (guarded by length to avoid over-matching).
+ * flags entities whose page is already present. Only exact normalized names are
+ * accepted here. Substring matching is unsafe because short, valid page titles
+ * can also be ordinary words inside an unrelated missing-page finding.
*/
function missingPageAlreadyExists(
llmTitle: string,
@@ -84,11 +84,7 @@ function missingPageAlreadyExists(
): boolean {
const norm = normalizeForExistence(llmTitle)
if (!norm) return false
- if (existingPageNames.has(norm)) return true
- for (const name of existingPageNames) {
- if (name.length >= 2 && norm.includes(name)) return true
- }
- return false
+ return existingPageNames.has(norm)
}
function extractTitle(content: string, fallbackPath: string): string {
@@ -406,6 +402,7 @@ export async function runSemanticLint(
"- stale: information that appears outdated or superseded",
"- missing-page: an important concept is heavily referenced but has no dedicated page",
"- suggestion: a question or source worth adding to the wiki",
+ "For missing-page findings, Short title must be only the exact missing concept or entity name, without explanatory prefixes or suffixes.",
"",
"Severities:",
"- warning: should be addressed",