@cachly-dev/mcp-server 0.9.17 → 0.9.19

package/dist/search.d.ts

@@ -0,0 +1,101 @@
+import { Redis } from 'ioredis';
+/**
+ * Stopwords — filtered out during tokenization.
+ * Covers: English, German, French, Spanish, Italian, Portuguese,
+ *         Chinese (Simplified + Traditional), Japanese, Korean.
+ * Keeps the index small and scores meaningful.
+ */
+declare const STOPWORDS: Set<string>;
+/**
+ * Convert a katakana string to Hepburn romaji.
+ * Handles digraphs, geminate consonants (ッ), and long vowel marks (ー).
+ */
+declare function katakanaToRomaji(kata: string): string;
+/**
+ * Arabic light stemmer — strips definite article and common prefix particles.
+ * Handles: ال (al-), و (wa-), ب (bi-), ل (li-), ف (fa-), ك (ka-).
+ * Runs iteratively (max 3 passes) so compound prefixes like فال- are fully
+ * resolved: فالخطأ → الخطأ → خطأ.
+ * Only strips when the result is still ≥ 3 chars to avoid over-stemming.
+ */
+declare function arabicLightStem(word: string): string;
+declare const CROSS_LINGUAL_MAP: Map<string, string[]>;
+/**
+ * Expand a token to its cross-lingual synonyms.
+ * Returns synonyms to be added to the token stream (pre-tokenized).
+ * CJK synonyms are NOT pre-bigram'd here — caller handles that.
+ */
+declare function expandCrossLingual(token: string): string[];
+/**
+ * Tokenize text into meaningful keywords.
+ * Handles accented Latin, RTL (Arabic, Hebrew), and CJK (Chinese, Japanese, Korean).
+ *
+ * Features:
+ *  • CJK → character bigrams
+ *  • Katakana → additionally emits Hepburn romaji tokens (enables romaji queries)
+ *  • Arabic  → word tokenization + iterative prefix stemming (ال/و/ب/ل/ف/ك)
+ *  • Hebrew  → word tokenization + iterative prefix stemming (ה/ו/ב/ל/מ/כ)
+ *  • Farsi   → word tokenization + suffix/prefix stemming (ها/های plural, می prefix)
+ *  • Hindi   → Devanagari word tokenization + light suffix stemming
+ *  • All scripts → cross-lingual synonym expansion (EN↔JA↔ZH↔KO↔AR↔HE↔FA↔HI↔DE↔FR↔ES↔IT↔PT)
+ */
+declare function tokenize(text: string): string[];
+/**
+ * Split a multi-topic query into sub-queries.
+ * Detects numbered lists, semicolons, "and also", line breaks, etc.
+ *
+ * Example: "deploy API fix routing and also check auth" →
+ *   ["deploy API fix routing", "check auth"]
+ *
+ * Example: "1. deploy 2. fix routing 3. auth" →
+ *   ["deploy", "fix routing", "auth"]
+ */
+declare function splitMultiQuery(query: string): string[];
+/**
+ * Zero-results query log — in-memory FIFO ring (max 500 entries).
+ * Exposed via /health for observability. Never persisted.
+ */
+interface ZeroResultEntry {
+    query: string;
+    ts: number;
+}
+declare const ZERO_RESULTS_LOG: ZeroResultEntry[];
+declare let zeroResultsTotal: number;
+declare function logZeroResult(query: string): void;
+/** Global index vocabulary — rebuilt on each keywordSearch call from the doc set. */
+declare let _indexVocab: Set<string>;
+interface KeywordMatch {
+    key: string;
+    content: string;
+    score: number;
+    matchedWords: string[];
+    subQuery?: string;
+}
+/**
+ * Levenshtein distance — edit distance between two strings.
+ * Used for typo-tolerant fuzzy matching (distance ≤ 2 = match).
+ * O(n*m) but strings are short (tokens), so this is fast.
+ */
+declare function levenshtein(a: string, b: string): number;
+/** Extract a timestamp from JSON content (looks for "ts" or "created" fields). */
+declare function extractTimestamp(content: string): number | undefined;
+/** Recency multiplier: 1.0 for now, 0.5 after half-life days, exponential decay. */
+declare function recencyBoost(timestampMs: number | undefined): number;
+/**
+ * BM25+ with Bigram Proximity, Recency Boost, and Fuzzy Matching.
+ *
+ * Architecture:
+ * ┌─────────────────────────────────────────────────────────────────┐
+ * │  Query: "deploy API fix routing; check auth"                    │
+ * │    ↓ splitMultiQuery                                            │
+ * │  Sub-queries: ["deploy API fix routing", "check auth"]          │
+ * │    ↓ for each sub-query                                         │
+ * │  Tokenize → BM25+ score per doc → Bigram boost → Fuzzy match   │
+ * │    ↓ merge & deduplicate                                        │
+ * │  Recency boost → Sort → Top-K                                  │
+ * └─────────────────────────────────────────────────────────────────┘
+ */
+declare function keywordSearch(redis: Redis, patterns: string[], query: string, topK?: number): Promise<KeywordMatch[]>;
+export { tokenize, splitMultiQuery, levenshtein, recencyBoost, extractTimestamp, STOPWORDS, katakanaToRomaji, arabicLightStem, expandCrossLingual, CROSS_LINGUAL_MAP };
+export type { KeywordMatch };
+export { keywordSearch, ZERO_RESULTS_LOG, logZeroResult, _indexVocab as indexVocab, zeroResultsTotal };