implementing phase 3 pipeline: gemini structured output, validation, sqlite staging

Prompt is now a template (fixes the hardcoded en/es leftovers in rules
2, 3, 15, 16, 26, 31). pipeline.ts replaces the pseudocode: wordlist
normalization, skip-already-staged idempotency, batches of 20 against
gemini-3.6-flash with responseSchema, raw responses persisted per batch,
per-entry validation with rejection log, one transaction per word into
db/staging.db. Flags: --langs --pos --max-batches --delay-ms --dry-run.

Smoke run: 40/40 words staged (de+es, one batch each), 0 rejections.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

data-pipeline/promptTemplate.ts

@@ -0,0 +1,56 @@
+import { readFileSync } from "node:fs";
+import path from "node:path";
+import type { SupportedLanguageCode, SupportedPos } from "@lila/shared";
+
+export const LANGUAGE_NAMES: Record<SupportedLanguageCode, string> = {
+  en: "English",
+  it: "Italian",
+  de: "German",
+  fr: "French",
+  es: "Spanish",
+};
+
+export type PromptParams = {
+  sourceLanguage: SupportedLanguageCode;
+  pos: SupportedPos;
+  targetLanguages: readonly SupportedLanguageCode[];
+  words: readonly string[];
+};
+
+export const loadPromptTemplate = (): string =>
+  readFileSync(path.join(import.meta.dirname, "prompt"), "utf-8");
+
+export const renderPrompt = (
+  template: string,
+  params: PromptParams,
+): string => {
+  const { sourceLanguage, pos, targetLanguages, words } = params;
+  if (words.length === 0) {
+    throw new Error("renderPrompt: empty word batch");
+  }
+  if (
+    targetLanguages.length === 0 ||
+    targetLanguages.includes(sourceLanguage)
+  ) {
+    throw new Error(
+      `renderPrompt: target languages must be non-empty and exclude the source language (got ${targetLanguages.join(", ")})`,
+    );
+  }
+
+  const rendered = template
+    .replaceAll("{{SOURCE_LANGUAGE_NAME}}", LANGUAGE_NAMES[sourceLanguage])
+    .replaceAll("{{SOURCE_LANGUAGE_CODE}}", sourceLanguage)
+    .replaceAll("{{POS}}", pos)
+    .replaceAll("{{TARGET_LANGUAGE_CODES}}", targetLanguages.join(", "))
+    .replaceAll(
+      "{{TARGET_LANGUAGE_UNION}}",
+      targetLanguages.map((code) => `"${code}"`).join(" | "),
+    )
+    .replaceAll("{{INPUT_WORDS}}", words.join("\n"));
+
+  const leftover = rendered.match(/\{\{[A-Z_]+\}\}/);
+  if (leftover) {
+    throw new Error(`renderPrompt: unreplaced placeholder ${leftover[0]}`);
+  }
+  return rendered;
+};