From 1c4dcdd108f1fc4d4a1e04bfb7f5a15c92df753d Mon Sep 17 00:00:00 2001
From: lila <beiweitemderbeste@protonmail.com>
Date: Mon, 6 Jul 2026 14:41:30 +0200
Subject: [PATCH] bugfixing #2

---
 data-pipeline/config/constants.ts          |  2 ++
 data-pipeline/config/prompt.ts             | 12 ++++++---
 data-pipeline/utils/enrich-word.ts         | 31 +++++++++-------------
 data-pipeline/utils/llm-adapters/gemini.ts |  8 ++----
 4 files changed, 25 insertions(+), 28 deletions(-)

diff --git a/data-pipeline/config/constants.ts b/data-pipeline/config/constants.ts
index 98731f2..7ffc146 100644
--- a/data-pipeline/config/constants.ts
+++ b/data-pipeline/config/constants.ts
@@ -12,3 +12,5 @@ export const POS_MAP: Record<string, string> = {
   adverbs: "adverb",
   adjectives: "adjective",
 };
+
+export const ALL_LANGUAGES = ["en", "de", "it", "es", "fr"];
diff --git a/data-pipeline/config/prompt.ts b/data-pipeline/config/prompt.ts
index 060e650..fd0081d 100644
--- a/data-pipeline/config/prompt.ts
+++ b/data-pipeline/config/prompt.ts
@@ -1,6 +1,11 @@
-export const ENRICHMENT_SYSTEM_PROMPT = `You are a multilingual dictionary engine. Output ONLY a JSON object. No markdown, no explanations.
+export function buildSystemPrompt(
+  sourceLanguage: string,
+  pos: string,
+  targetLanguages: string[],
+): string {
+  return `You are a multilingual dictionary engine. Output ONLY a JSON object. No markdown, no explanations.
 
-For each English noun provided, generate 1-2 distinct senses.
+For each ${sourceLanguage} ${pos} provided, generate 1-2 distinct senses.
 
 CEFR difficulty mapping:
 - A1/A2 → easy
@@ -11,7 +16,7 @@ Each sense must have:
 - sense: student-friendly definition, max 15 words
 - example: natural sentence using the word
 - difficulty_level: easy, medium, or hard
-- translations: object with keys de, it, es, fr; each value is an array of {word, gender} where gender MUST be masculine, feminine, or neuter. Use null ONLY if the language has no grammatical gender for that word.
+- translations: object with keys ${targetLanguages.join(", ")}; each value is an array of {word, gender} where gender MUST be masculine, feminine, or neuter. Use null ONLY if the language has no grammatical gender for that word.
 
 Output format: JSON object where keys are the input words, values are arrays of sense objects.
 
@@ -32,3 +37,4 @@ Example for ["house"]:
   ]
 }
 `;
+}
diff --git a/data-pipeline/utils/enrich-word.ts b/data-pipeline/utils/enrich-word.ts
index 36b6b25..5a1bde5 100644
--- a/data-pipeline/utils/enrich-word.ts
+++ b/data-pipeline/utils/enrich-word.ts
@@ -1,22 +1,8 @@
-import { ENRICHMENT_SYSTEM_PROMPT } from "../config/prompt.js";
+import { buildSystemPrompt } from "../config/prompt.js";
 import { createAdapter } from "./llm-adapters/factory.js";
 import { BATCH_CONFIG } from "../config/batch.js";
 import type { Language, Pos, EnrichedSense } from "./merge-enriched-data.js";
-
-const LANG_MAP: Record<string, string> = {
-  english: "en",
-  italian: "it",
-  german: "de",
-  french: "fr",
-  spanish: "es",
-};
-
-const POS_MAP: Record<string, string> = {
-  nouns: "noun",
-  verbs: "verb",
-  adverbs: "adverb",
-  adjectives: "adjective",
-};
+import { LANG_MAP, POS_MAP, ALL_LANGUAGES } from "../config/constants.js";
 
 interface LlmResponse {
   content: string;
@@ -42,9 +28,16 @@ export interface EnrichmentResult {
  * Calls the LLM with the enrichment prompt.
  * Returns the response content and timing metrics.
  */
-async function callLlm(words: string[]): Promise<LlmResponse> {
+async function callLlm(
+  words: string[],
+  rawLanguage: string,
+  rawPos: string,
+): Promise<LlmResponse> {
   const adapter = createAdapter();
-  return adapter.call(words, ENRICHMENT_SYSTEM_PROMPT);
+  const sourceCode = LANG_MAP[rawLanguage] || rawLanguage;
+  const targetLanguages = ALL_LANGUAGES.filter((lang) => lang !== sourceCode);
+  const prompt = buildSystemPrompt(rawLanguage, rawPos, targetLanguages);
+  return adapter.call(words, prompt);
 }
 
 /**
@@ -140,7 +133,7 @@ export async function enrichWord(
   rawLanguage: string,
   rawPos: string,
 ): Promise<EnrichmentResult> {
-  const llmResponse = await callLlm(words);
+  const llmResponse = await callLlm(words, rawLanguage, rawPos);
   const parsed = parseLlmResponse(llmResponse.content, words);
   const results = buildEnrichedData(parsed, rawLanguage, rawPos);
 
diff --git a/data-pipeline/utils/llm-adapters/gemini.ts b/data-pipeline/utils/llm-adapters/gemini.ts
index 3929b4d..3a6060d 100644
--- a/data-pipeline/utils/llm-adapters/gemini.ts
+++ b/data-pipeline/utils/llm-adapters/gemini.ts
@@ -32,13 +32,9 @@ export class GeminiAdapter implements LlmAdapter {
     const url = `https://generativelanguage.googleapis.com/v1beta/models/${this.model}:generateContent?key=${this.apiKey}`;
 
     const payload = {
+      systemInstruction: { parts: [{ text: systemPrompt }] },
       contents: [
-        {
-          role: "user",
-          parts: [
-            { text: systemPrompt + "\n\nWords: " + JSON.stringify(words) },
-          ],
-        },
+        { role: "user", parts: [{ text: "Words: " + JSON.stringify(words) }] },
       ],
       generationConfig: {
         temperature: 0.1,