Use llm for vacancies post processing

vacancies/main/bot.py

@@ -62,6 +62,8 @@ async def error_handler(update: object, context: ContextTypes.DEFAULT_TYPE) -> N
 
 
 async def handle_document(update: Update, context: ContextTypes.DEFAULT_TYPE):
+    message = await context.bot.send_message(update.effective_chat.id, "📝 Обрабатываю твой запрос. Пожалуйста, подождите...")
+
     if not update.message.document:
         await context.bot.send_message(chat_id=update.effective_chat.id, text="Не удалось прочитать информацию из файла! Попробуйте другой формат.")
         return
@@ -84,7 +86,7 @@ async def handle_document(update: Update, context: ContextTypes.DEFAULT_TYPE):
         {'content': customer_cv.content, 'features_json': features.model_dump()},
     )
 
-    await context.bot.send_message(chat_id=update.effective_chat.id, text="Отлично! Запомнил Ваше резюме.")
+    await context.bot.editMessageText("Отлично! Запомнил Ваше резюме.", update.effective_chat.id, message.id)
 
 
 application = ApplicationBuilder().token(os.environ["BOT_TOKEN"]).build()

vacancies/main/management/commands/generate_recommended_vacancies.py

@@ -20,7 +20,11 @@ class Command(BaseCommand):
             ).values_list('vacancy_id', flat=True)
 
             query_filter = Filter(must_not=[HasIdCondition(has_id=recommended_vacancy_ids)])
-            search_result_id, vacancy_content, link = search_similarities(query_filter, customer_cv.id)
+            result = search_similarities(query_filter, customer_cv.id)
+            if not result:
+                continue
+
+            search_result_id, vacancy_content, link = result
 
             recommendation = RecommendedVacancy.objects.create(
                 customer=customer_cv.customer,

vacancies/main/vector_store.py

@@ -106,7 +106,7 @@ def add_vectors(collection_name: str, _id: int, features: dict, payload: dict):
     )
 
 
-def search_similarities(query_filter: Filter, cv_id: int) -> list[dict]:
+def search_similarities(query_filter: Filter, cv_id: int):
     vectors = client.retrieve(
         collection_name="cvs",
         ids=[cv_id],
@@ -141,10 +141,19 @@ def search_similarities(query_filter: Filter, cv_id: int) -> list[dict]:
         total = sum(feature_sims[feature] * weights.get(feature, 1) for feature in feature_sims)
         scored.append({"id": vid, "score": total, "content": vacancies_content[vid]["content"], "link": vacancies_content[vid]["link"]})
 
-    scored.sort(key=lambda x: x["score"], reverse=True)
-    for i in range(20):
-        print(f"{scored[i]['content']} {scored[i]['score']}")
-    return scored[0]["id"], scored[0]["content"], scored[0]["link"]
+    prompt = f"""
+    Среди вакансий ниже выбери одну наиболее релевантную и выведи ее индекс.
+    Если среди вакансий нет подходящих, то верни -1.
+    В ответе выведи только число.
+    {scored}
+    """
+    openai_client = ChatOpenAI(model_name="gpt-5-mini", reasoning_effort="minimal", temperature=0, seed=42, top_p=1)
+    response = openai_client.invoke(prompt)
+    index = int(response.content)
+    if index == -1:
+        return None
+
+    return scored[index]["id"], scored[index]["content"], scored[index]["link"]
 
 
 def extract_features(content: str) -> VacancyFeatures: