Merge pull request #97 from DefangLabs/linda-cpu-reduction

Author: nullfunc

app/Dockerfile

@@ -49,6 +49,4 @@ EXPOSE 5050
 ENV FLASK_APP=app.py
 
 # Run the application using uWSGI
-# CMD ["uwsgi", "--lazy-apps", "--http", "0.0.0.0:5050", "--wsgi-file", "app.py", "--callable", "app", "--processes", "4"]
-# longer timeout settings to allow for slower rag and llm responses
-CMD ["uwsgi", "--lazy-apps", "--http", "0.0.0.0:5050", "--wsgi-file", "app.py", "--callable", "app", "--processes", "4", "--harakiri", "300", "--socket-timeout", "300", "--http-timeout", "300", "--http-keepalive", "--http-auto-chunked"]
+CMD ["uwsgi", "--lazy-apps", "--http", "0.0.0.0:5050", "--wsgi-file", "app.py", "--callable", "app", "--processes", "2"]

app/test_rag_system.py

@@ -1,6 +1,5 @@
 import unittest
 
-from torch import cosine_similarity
 from rag_system import RAGSystem
 
 class TestRAGSystem(unittest.TestCase):

compose.yaml

@@ -30,7 +30,7 @@ services:
       resources:
         reservations:
           memory: 4G
-          cpus: 2.0
+          cpus: 1.0
     healthcheck:
       test: ["CMD", "curl", "-f", "http://localhost:5050/"]
       interval: 30s