pyproject.toml

@@ -11,7 +11,6 @@ dependencies = [
     "langchain>=0.3.27",
     "langchain-openai>=0.3.35",
     "langchain-qdrant>=1.1.0",
-    "langgraph-checkpoint-postgres>=3.0.0",
     "psycopg[binary]>=3.2.12",
     "pydantic>=2.0",
     "pypdf>=6.1.2",

vacancies/conf/settings.py

@@ -14,8 +14,6 @@ import os
 from pathlib import Path
 
 import sentry_sdk
-import clickhouse_connect
-from qdrant_client import QdrantClient
 
 # Build paths inside the project like this: BASE_DIR / 'subdir'.
 BASE_DIR = Path(__file__).resolve().parent.parent.parent
@@ -92,8 +90,6 @@ DATABASES = {
     },
 }
 
-DB_URI = f"postgres://{DATABASES['default']['USER']}:{DATABASES['default']['PASSWORD']}@{DATABASES['default']['HOST']}:{DATABASES['default']['PORT']}/{DATABASES['default']['NAME']}?sslmode=disable"
-
 
 # Password validation
 # https://docs.djangoproject.com/en/5.2/ref/settings/#auth-password-validators
@@ -164,7 +160,3 @@ LOGGING = {
         },
     },
 }
-
-CLICKHOUSE_CLIENT = clickhouse_connect.create_client(host="127.0.0.1", port=18123)
-
-QDARNT_CLIENT = QdrantClient(url="http://localhost:6333")

vacancies/main/bot.py

@@ -7,9 +7,8 @@ from pypdf import PdfReader
 from vacancies.main.models import Customer, CustomerCV
 from langchain.agents import create_agent
 from langchain_openai import ChatOpenAI
-from langgraph.checkpoint.postgres.aio import AsyncPostgresSaver
+from langgraph.checkpoint.memory import InMemorySaver
 from vacancies.main.vector_store import add_vectors, extract_features
-from vacancies.conf.settings import DB_URI
 
 SYSTEM_PROMPT = """
 Ты — карьерный копилот для ИТ. Ты можешь отвечать на любые вопросы по тематике карьеры.
@@ -26,6 +25,14 @@ async def get_user_resume(user_id: int):
     return customer_cv.content if customer_cv else ""
 
 
+agent = create_agent(
+    model=ChatOpenAI(model_name="gpt-5-mini", reasoning_effort="minimal"),
+    tools=[get_user_resume],
+    system_prompt=SYSTEM_PROMPT,
+    checkpointer=InMemorySaver(),
+)
+
+
 async def start(update: Update, context: ContextTypes.DEFAULT_TYPE):
     await Customer.objects.aget_or_create(
         telegram_id=update.effective_user.id,
@@ -39,22 +46,14 @@ async def start(update: Update, context: ContextTypes.DEFAULT_TYPE):
 
 
 async def prompt(update: Update, context: ContextTypes.DEFAULT_TYPE):
-    async with AsyncPostgresSaver.from_conn_string(DB_URI) as checkpointer:
+    message = await context.bot.send_message(update.effective_chat.id, "📝 Обрабатываю твой запрос. Пожалуйста, подождите...")
-        agent = create_agent(
-            model=ChatOpenAI(model_name="gpt-5-mini", reasoning_effort="minimal"),
-            tools=[get_user_resume],
-            system_prompt=SYSTEM_PROMPT,
-            checkpointer=checkpointer,
-        )
 
-        message = await context.bot.send_message(update.effective_chat.id, "📝 Обрабатываю твой запрос. Пожалуйста, подождите...")
+    response = await agent.ainvoke(
+        input={"messages": [{"role": "user", "content": f'user_id = {update.effective_user.id}\n{update.message.text}'}]},
+        config={"configurable": {"thread_id": update.effective_user.id}},
+    )
 
-        response = await agent.ainvoke(
+    await context.bot.editMessageText(response['messages'][-1].content, update.effective_chat.id, message.id)
-            input={"messages": [{"role": "user", "content": f'user_id = {update.effective_user.id}\n{update.message.text}'}]},
-            config={"configurable": {"thread_id": update.effective_user.id}},
-        )
-
-        await context.bot.editMessageText(response['messages'][-1].content, update.effective_chat.id, message.id)
 
 
 async def error_handler(update: object, context: ContextTypes.DEFAULT_TYPE) -> None:

vacancies/main/management/commands/collect_vacancies_from_telegram_messages.py

@@ -1,7 +1,8 @@
 from django.core.management import BaseCommand
+import clickhouse_connect
 from vacancies.main.vector_store import add_vectors, extract_features, client as qdrant
-from vacancies.conf.settings import CLICKHOUSE_CLIENT
 
+clickhouse_client = clickhouse_connect.create_client(host="127.0.0.1", port=18123)
 
 query = """
 SELECT id, chat_username, telegram_id, message, timestamp
@@ -39,7 +40,7 @@ class Command(BaseCommand):
             next_page_offset = response[1]
         exist_points_set = tuple(set(exist_points_ids))
 
-        result_rows = CLICKHOUSE_CLIENT.query(query, parameters={"exist_points": exist_points_set}).result_rows
+        result_rows = clickhouse_client.query(query, parameters={"exist_points": exist_points_set}).result_rows
         result_rows_len = len(result_rows)
         for index, row in enumerate(result_rows):
             (id, chat_username, telegram_id, message, timestamp) = row

vacancies/main/management/commands/runbot.py

@@ -1,19 +1,9 @@
-import sys
-import asyncio
 from django.core.management import BaseCommand
 from vacancies.main.bot import application
-from langgraph.checkpoint.postgres import PostgresSaver
-from vacancies.conf.settings import DB_URI
 
 
 class Command(BaseCommand):
     help = "Run bot"
 
     def handle(self, *args, **options):
-        with PostgresSaver.from_conn_string(DB_URI) as checkpointer:
-            checkpointer.setup()
-
-        if sys.platform == "win32":
-                asyncio.set_event_loop_policy(asyncio.WindowsSelectorEventLoopPolicy())
-
         application.run_polling()

vacancies/main/vector_store.py

@@ -1,10 +1,11 @@
 from qdrant_client import models
 from langchain_openai import OpenAIEmbeddings
 from langchain_openai import ChatOpenAI
+from qdrant_client import QdrantClient
 from qdrant_client.models import Filter
 from vacancies.main.models import VacancyFeatures
-from vacancies.conf.settings import QDARNT_CLIENT
 
+client = QdrantClient(url="http://localhost:6333")
 
 FEATURE_NAMES = [
     "job_title", "employment_type", "work_format", "experience", "position_level", "industry", "tech_stack",
@@ -31,13 +32,13 @@ vectors_config = {
     name: models.VectorParams(size=3072, distance=models.Distance.COSINE) for name in FEATURE_NAMES
 }
 
-if not QDARNT_CLIENT.collection_exists("vacancies"):
+if not client.collection_exists("vacancies"):
-    QDARNT_CLIENT.create_collection(
+    client.create_collection(
         collection_name="vacancies",
         vectors_config=vectors_config
     )
-if not QDARNT_CLIENT.collection_exists("cvs"):
+if not client.collection_exists("cvs"):
-    QDARNT_CLIENT.create_collection(
+    client.create_collection(
         collection_name="cvs",
         vectors_config=vectors_config
     )
@@ -70,7 +71,7 @@ def add_vectors(collection_name: str, _id: int, features: dict, payload: dict):
     max_similarities = {}
     for name, vec in vectors.items():
         if any(v != 0 for v in vec):
-            results = QDARNT_CLIENT.query_points(
+            results = client.query_points(
                 collection_name="vacancies",
                 query=vec,
                 using=name,
@@ -92,7 +93,7 @@ def add_vectors(collection_name: str, _id: int, features: dict, payload: dict):
     if scored and scored[0]["score"] > 33:  # threshold
         return
 
-    QDARNT_CLIENT.upsert(
+    client.upsert(
         collection_name=collection_name,
         points=[
             models.PointStruct(
@@ -105,7 +106,7 @@ def add_vectors(collection_name: str, _id: int, features: dict, payload: dict):
 
 
 def search_similarities(query_filter: Filter, cv_id: int):
-    cv = QDARNT_CLIENT.retrieve(
+    cv = client.retrieve(
         collection_name="cvs",
         ids=[cv_id],
         with_vectors=True,
@@ -115,7 +116,7 @@ def search_similarities(query_filter: Filter, cv_id: int):
     vacancies_content = {}
     for name, vec in cv.vector.items():
         if any(v != 0 for v in vec):
-            results = QDARNT_CLIENT.query_points(
+            results = client.query_points(
                 collection_name="vacancies",
                 query=vec,
                 using=name,