@tikoci/rosetta 0.7.6 → 0.7.7

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tikoci/rosetta",
-  "version": "0.7.6",
+  "version": "0.7.7",
   "description": "RouterOS documentation as SQLite FTS5 — RAG search + command glossary via MCP",
   "type": "module",
   "license": "MIT",

package/src/classify.ts

@@ -14,7 +14,49 @@
  */
 
 import { canonicalize } from "./canonicalize.ts";
-import { KNOWN_TOPICS } from "./query.ts";
+
+/**
+ * Known RouterOS topics — extracted from changelog categories, top-level
+ * command tree names, and common subsystem tokens. Used by the classifier
+ * to recognize domain-specific terms before FTS search.
+ *
+ * Source: `SELECT DISTINCT category FROM changelogs` (153 values) plus
+ * top-level commands (app, interface, ip, system) and common synonyms.
+ *
+ * Defined here (not in query.ts) so classify.ts stays pure (no DB imports)
+ * and classify.test.ts cannot transitively load db.ts before test setup.
+ */
+export const KNOWN_TOPICS = new Set([
+  // --- From changelog categories (high-frequency subsystems) ---
+  "api", "backup", "bgp", "bluetooth", "bonding", "bridge", "capsman",
+  "certificate", "chr", "cloud", "conntrack", "console", "container",
+  "defconf", "discovery", "disk", "dns", "dot1x", "dude",
+  "ethernet", "export", "fetch", "filesystem", "firewall",
+  "gps", "graphing", "hardware", "health", "hotspot",
+  "ike1", "ike2", "interface", "ipsec", "ipv6",
+  "l2tp", "l3hw", "ldp", "led", "log", "lora", "lte",
+  "macsec", "mlag", "modem", "mpls", "mqtt",
+  "netinstall", "netwatch", "ntp",
+  "ospf", "ovpn",
+  "poe", "ppp", "pppoe", "pptp", "ptp",
+  "queue", "quickset",
+  "radius", "resource", "rip", "romon", "route", "routing",
+  "serial", "sfp", "smb", "sms", "sniffer", "snmp", "socks", "ssh", "ssl", "sstp",
+  "switch", "system",
+  "traceroute", "tunnels", "upgrade", "upnp", "ups", "usb", "user",
+  "vlan", "vpls", "vrf", "vrrp", "vxlan",
+  "webfig", "wifi", "wifiwave2", "winbox", "wireguard", "wireless",
+  "zerotier",
+  // --- From changelog (DHCP variants, routing sub-protocols) ---
+  "dhcp", "dhcpv4", "dhcpv6", "rpki", "pimsm",
+  // --- Top-level command paths ---
+  "app", "ip",
+  // --- Common subsystem shorthand / synonyms ---
+  "nat", "mangle", "raw", "filter",
+  "bgp-vpn", "user-manager", "traffic-flow", "traffic-generator",
+  "route-filter", "routing-filter", "mac-telnet",
+  "w60g", "tr069",
+]);
 
 export type CommandFragment = {
   /** `key=value` pairs extracted from the input (e.g. `chain=forward`). */

package/src/extract-dude.test.ts

@@ -1,7 +1,16 @@
+// Force in-memory DB BEFORE importing extract-dude.ts (which transitively imports
+// db.ts). Without this, db.ts evaluates against the project's real ros-help.db
+// and any later test file (e.g. query.test.ts) that calls DELETE in beforeAll
+// will wipe the CI-built database — exactly the bug that shipped 3-page DBs in
+// release v0.7.6. See BACKLOG.md "Test DB-leak guards".
+process.env.DB_PATH = ":memory:";
+
 import { describe, expect, test } from "bun:test";
 import { readFileSync } from "node:fs";
 import { join } from "node:path";
-import { parseDudePage } from "./extract-dude.ts";
+
+// Dynamic import so the DB_PATH assignment above wins over module hoisting.
+const { parseDudePage } = await import("./extract-dude.ts");
 
 const probesHtml = readFileSync(join(import.meta.dirname, "..", "dude", "pages", "Probes.html"), "utf-8");
 

package/src/extract-html.ts

@@ -558,6 +558,16 @@ function main() {
   console.log(`  Total sections:  ${totalSections} (across ${pagesWithSections} pages)`);
   console.log(`  FTS index rows:  ${ftsCount}`);
 
+  // Hard fail if no pages were extracted — silent zero-page runs were the
+  // root cause of release v0.7.6 shipping a near-empty DB. CI must fail loud.
+  if (extracted === 0) {
+    console.error(
+      `\n::error::extract-html: 0 pages extracted from ${process.argv[2] ?? "(default dir)"}. ` +
+        `Check the HTML export download/unzip step.`,
+    );
+    process.exit(1);
+  }
+
 // Quick search test
   const testResults = db
     .prepare(

package/src/query.test.ts

@@ -13,7 +13,19 @@ import { beforeAll, describe, expect, test } from "bun:test";
 process.env.DB_PATH = ":memory:";
 
 // Dynamic imports so the env-var assignment above is visible to db.ts
-const { db, initDb, getDbStats, checkSchemaVersion, SCHEMA_VERSION } = await import("./db.ts");
+const { db, initDb, getDbStats, checkSchemaVersion, SCHEMA_VERSION, DB_PATH } = await import("./db.ts");
+
+// Hard guard: if some other test file imported db.ts before us with a real
+// path, the singleton will be pointing at the project's ros-help.db and the
+// DELETEs in beforeAll() below would wipe it. Fail fast with a clear message
+// so this can never silently ship an empty DB again (release v0.7.6 regression).
+if (DB_PATH !== ":memory:") {
+  throw new Error(
+    `query.test.ts: DB singleton is at "${DB_PATH}" — expected ":memory:". ` +
+      `Another test file imported db.ts before this one without setting DB_PATH=:memory:. ` +
+      `Refusing to run because beforeAll() would DELETE FROM real tables.`,
+  );
+}
 const {
   extractTerms,
   buildFtsQuery,

package/src/query.ts

@@ -90,46 +90,9 @@ const COMPOUND_TERMS: [string, string][] = [
   ["address", "list"],
 ];
 
-/**
- * Known RouterOS topics — extracted from changelog categories, top-level
- * command tree names, and common subsystem tokens. Used by the classifier
- * (future) and glossary lookup to recognize domain-specific terms before
- * FTS search. Maintained as a static Set for O(1) lookup.
- *
- * Source: `SELECT DISTINCT category FROM changelogs` (153 values) plus
- * top-level commands (app, interface, ip, system) and common synonyms.
- */
-export const KNOWN_TOPICS = new Set([
-  // --- From changelog categories (high-frequency subsystems) ---
-  "api", "backup", "bgp", "bluetooth", "bonding", "bridge", "capsman",
-  "certificate", "chr", "cloud", "conntrack", "console", "container",
-  "defconf", "discovery", "disk", "dns", "dot1x", "dude",
-  "ethernet", "export", "fetch", "filesystem", "firewall",
-  "gps", "graphing", "hardware", "health", "hotspot",
-  "ike1", "ike2", "interface", "ipsec", "ipv6",
-  "l2tp", "l3hw", "ldp", "led", "log", "lora", "lte",
-  "macsec", "mlag", "modem", "mpls", "mqtt",
-  "netinstall", "netwatch", "ntp",
-  "ospf", "ovpn",
-  "poe", "ppp", "pppoe", "pptp", "ptp",
-  "queue", "quickset",
-  "radius", "resource", "rip", "romon", "route", "routing",
-  "serial", "sfp", "smb", "sms", "sniffer", "snmp", "socks", "ssh", "ssl", "sstp",
-  "switch", "system",
-  "traceroute", "tunnels", "upgrade", "upnp", "ups", "usb", "user",
-  "vlan", "vpls", "vrf", "vrrp", "vxlan",
-  "webfig", "wifi", "wifiwave2", "winbox", "wireguard", "wireless",
-  "zerotier",
-  // --- From changelog (DHCP variants, routing sub-protocols) ---
-  "dhcp", "dhcpv4", "dhcpv6", "rpki", "pimsm",
-  // --- Top-level command paths ---
-  "app", "ip",
-  // --- Common subsystem shorthand / synonyms ---
-  "nat", "mangle", "raw", "filter",
-  "bgp-vpn", "user-manager", "traffic-flow", "traffic-generator",
-  "route-filter", "routing-filter", "mac-telnet",
-  "w60g", "tr069",
-]);
+// KNOWN_TOPICS is defined in classify.ts (pure module, no DB) and re-exported
+// from here for backward compatibility. See classify.ts for the canonical definition.
+export { KNOWN_TOPICS } from "./classify.ts";
 
 function getSpecialSearchHint(question: string): string | undefined {
   const normalized = question.toLowerCase();

package/src/release.test.ts

@@ -387,6 +387,26 @@ describe("release.yml", () => {
     expect(src).toContain("npm publish");
     expect(src).toContain("NPM_TOKEN");
   });
+
+  test("validates DB content before publishing (regression: v0.7.6 shipped 3 pages)", () => {
+    const src = readText(".github/workflows/release.yml");
+    // Hard guard step that hard-fails the workflow if the built DB is degenerate.
+    expect(src).toContain("Validate DB has expected content");
+    // Must run BEFORE the artifact build, container push, GH Release, and npm publish.
+    const validateIdx = src.indexOf("Validate DB has expected content");
+    const buildIdx = src.indexOf("Build release artifacts");
+    const releaseIdx = src.indexOf("gh release create");
+    const npmIdx = src.indexOf("npm publish");
+    expect(validateIdx).toBeGreaterThan(0);
+    expect(validateIdx).toBeLessThan(buildIdx);
+    expect(validateIdx).toBeLessThan(releaseIdx);
+    expect(validateIdx).toBeLessThan(npmIdx);
+    // Must check minimum thresholds for the four critical tables.
+    expect(src).toMatch(/PAGES.*-lt 200/);
+    expect(src).toMatch(/COMMANDS.*-lt 1000/);
+    expect(src).toMatch(/DEVICES.*-lt 100/);
+    expect(src).toMatch(/PROPERTIES.*-lt 1000/);
+  });
 });
 
 // ---------------------------------------------------------------------------