@phren/cli/dist/core/search.js

Knowledge layer for AI agents - CLI, MCP server, and data layer

import { STOP_WORDS } from "../utils.js";
import { queryDocRows } from "../shared/index.js";
/**
 * Keyword overlap fallback for when FTS5 returns no results.
 * Scans all docs (optionally filtered by project/type), scores each by
 * how many query terms appear in its content, and returns top matches.
 *
 * Shared between the MCP search tool and CLI `phren search`.
 */
export function keywordFallbackSearch(db, query, opts) {
    let fallbackSql = "SELECT project, filename, type, content, path FROM docs";
    const fallbackParams = [];
    const clauses = [];
    if (opts.project) {
        clauses.push("project = ?");
        fallbackParams.push(opts.project);
    }
    if (opts.type) {
        clauses.push("type = ?");
        fallbackParams.push(opts.type);
    }
    if (clauses.length)
        fallbackSql += " WHERE " + clauses.join(" AND ");
    const allRows = queryDocRows(db, fallbackSql, fallbackParams);
    if (!allRows)
        return null;
    const terms = query
        .toLowerCase()
        .replace(/[^\w\s-]/g, " ")
        .split(/\s+/)
        .filter(w => w.length > 1 && !STOP_WORDS.has(w));
    if (terms.length === 0)
        return null;
    const scored = allRows
        .map((row) => {
        const content = row.content.toLowerCase();
        let score = 0;
        for (const term of terms) {
            if (content.includes(term))
                score++;
        }
        return { row, score };
    })
        .filter(r => r.score > 0)
        .sort((a, b) => b.score - a.score)
        .slice(0, opts.limit);
    if (scored.length === 0)
        return null;
    return scored.map(s => s.row);
}