Tune gunicorn lifecycle for streaming requests

Dockerfile

@@ -49,5 +49,5 @@ USER nobody
 # Expose port
 EXPOSE 8000
 
-# Use environment variable for workers count and optimize for database connections
-CMD ["sh", "-c", "gunicorn app.main:app -k uvicorn.workers.UvicornWorker --workers ${WORKERS:-5} --bind 0.0.0.0:8000 --timeout 120 --max-requests 1000 --max-requests-jitter 100"]
+# Use Gunicorn config so worker lifecycle settings are explicit and env-overridable.
+CMD ["gunicorn", "app.main:app", "--config", "gunicorn_conf.py"]

gunicorn_conf.py

@@ -0,0 +1,21 @@
+import os
+
+
+def _int_env(name: str, default: int) -> int:
+    value = os.getenv(name)
+    if value is None or value == "":
+        return default
+    return int(value)
+
+
+bind = f"0.0.0.0:{os.getenv('PORT', '8000')}"
+worker_class = "uvicorn.workers.UvicornWorker"
+workers = _int_env("WORKERS", 5)
+
+# Long provider streams should not be interrupted by the default 30s graceful
+# shutdown window, and workers should not recycle while streaming requests run.
+timeout = _int_env("GUNICORN_TIMEOUT", 300)
+graceful_timeout = _int_env("GUNICORN_GRACEFUL_TIMEOUT", 300)
+keepalive = _int_env("GUNICORN_KEEPALIVE", 75)
+max_requests = _int_env("GUNICORN_MAX_REQUESTS", 0)
+max_requests_jitter = _int_env("GUNICORN_MAX_REQUESTS_JITTER", 0)