diff --git a/.gitignore b/.gitignore
index 85ce3fa3..4ce172a4 100644
--- a/.gitignore
+++ b/.gitignore
@@ -58,8 +58,6 @@ coverage.xml
 # Django stuff:
 *.log
 local_settings.py
-db.sqlite3
-db.sqlite3-journal
 
 # Flask stuff:
 instance/
@@ -141,7 +139,6 @@ migration/content/**/*.md
 .obsidian
 
 *.zip
-*.sqlite3
 *.rdb
 .DS_Store
 /dump
diff --git a/Dockerfile b/Dockerfile
index 371ce1bf..49405943 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -5,5 +5,5 @@ ADD nginx.conf.sigil ./
 RUN /usr/local/bin/python -m pip install --upgrade pip
 WORKDIR /usr/src/app
 COPY requirements.txt ./
-RUN set -ex && pip install -r requirements.txt
+RUN pip install -r requirements.txt
 COPY . .
diff --git a/ai/preprocess.py b/ai/preprocess.py
new file mode 100644
index 00000000..afd8dbd8
--- /dev/null
+++ b/ai/preprocess.py
@@ -0,0 +1,75 @@
+import re
+import nltk
+from bs4 import BeautifulSoup
+from nltk.corpus import stopwords
+from pymystem3 import Mystem
+from string import punctuation
+from transformers import BertTokenizer
+
+nltk.download("stopwords")
+
+
+def get_clear_text(text):
+    soup = BeautifulSoup(text, 'html.parser')
+
+    # extract the plain text from the HTML document without tags
+    clear_text = ''
+    for tag in soup.find_all():
+        clear_text += tag.string or ''
+
+    clear_text = re.sub(pattern='[\u202F\u00A0\n]+', repl=' ', string=clear_text)
+
+    # only words
+    clear_text = re.sub(pattern='[^A-ZА-ЯЁ -]', repl='', string=clear_text, flags=re.IGNORECASE)
+
+    clear_text = re.sub(pattern='\s+', repl=' ', string=clear_text)
+
+    clear_text = clear_text.lower()
+
+    mystem = Mystem()
+    russian_stopwords = stopwords.words("russian")
+
+    tokens = mystem.lemmatize(clear_text)
+    tokens = [token for token in tokens if token not in russian_stopwords \
+              and token != " " \
+              and token.strip() not in punctuation]
+
+    clear_text = " ".join(tokens)
+
+    return clear_text
+
+
+# if __name__ == '__main__':
+#
+#     # initialize the tokenizer with the pre-trained BERT model and vocabulary
+#     tokenizer = BertTokenizer.from_pretrained('bert-base-multilingual-cased')
+#
+#     # split each text into smaller segments of maximum length 512
+#     max_length = 512
+#     segmented_texts = []
+#     for text in [clear_text1, clear_text2]:
+#         segmented_text = []
+#         for i in range(0, len(text), max_length):
+#             segment = text[i:i+max_length]
+#             segmented_text.append(segment)
+#         segmented_texts.append(segmented_text)
+#
+#     # tokenize each segment using the BERT tokenizer
+#     tokenized_texts = []
+#     for segmented_text in segmented_texts:
+#         tokenized_text = []
+#         for segment in segmented_text:
+#             segment_tokens = tokenizer.tokenize(segment)
+#             segment_tokens = ['[CLS]'] + segment_tokens + ['[SEP]']
+#             tokenized_text.append(segment_tokens)
+#         tokenized_texts.append(tokenized_text)
+#
+#     input_ids = []
+#     for tokenized_text in tokenized_texts:
+#         input_id = []
+#         for segment_tokens in tokenized_text:
+#             segment_id = tokenizer.convert_tokens_to_ids(segment_tokens)
+#             input_id.append(segment_id)
+#         input_ids.append(input_id)
+#
+#     print(input_ids)
diff --git a/base/orm.py b/base/orm.py
index 3001f40b..8d2e65ad 100644
--- a/base/orm.py
+++ b/base/orm.py
@@ -7,12 +7,9 @@ from sqlalchemy.sql.schema import Table
 
 from settings import DB_URL
 
-if DB_URL.startswith("sqlite"):
-    engine = create_engine(DB_URL)
-else:
-    engine = create_engine(
-        DB_URL, echo=False, pool_size=10, max_overflow=20
-    )
+engine = create_engine(
+    DB_URL, echo=False, pool_size=10, max_overflow=20
+)
 
 T = TypeVar("T")
 
diff --git a/requirements.txt b/requirements.txt
index 9863d82f..a1a718ed 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -7,7 +7,7 @@ pyjwt>=2.6.0
 starlette~=0.23.1
 sqlalchemy>=1.4.41
 graphql-core>=3.0.3
-gql
+gql~=3.4.0
 uvicorn>=0.18.3
 pydantic>=1.10.2
 passlib~=1.7.4
@@ -29,3 +29,6 @@ lxml
 sentry-sdk>=1.14.0
 # sse_starlette
 graphql-ws
+nltk~=3.8.1
+pymystem3~=0.2.0
+transformers~=4.28.1
diff --git a/settings.py b/settings.py
index 0bfa9785..54897dfa 100644
--- a/settings.py
+++ b/settings.py
@@ -4,7 +4,7 @@ PORT = 8080
 
 DB_URL = (
     environ.get("DATABASE_URL") or environ.get("DB_URL") or
-    "postgresql://postgres@localhost:5432/discoursio" or "sqlite:///db.sqlite3"
+    "postgresql://postgres@localhost:5432/discoursio"
 )
 JWT_ALGORITHM = "HS256"
 JWT_SECRET_KEY = environ.get("JWT_SECRET_KEY") or "8f1bd7696ffb482d8486dfbc6e7d16dd-secret-key"
diff --git a/test.py b/test.py
new file mode 100644
index 00000000..627cc372
--- /dev/null
+++ b/test.py
@@ -0,0 +1,11 @@
+from sqlalchemy import select
+from ai.preprocess import get_clear_text
+from base.orm import local_session
+from orm import Shout
+
+if __name__ == "__main__":
+    with local_session() as session:
+        q = select(Shout)
+        for [shout] in session.execute(q):
+            clear_shout_body = get_clear_text(shout.body)
+            print(clear_shout_body)