fix: Implement garbage collection in run_conversion_job to prevent memory accumulation and add resource limits in docker-compose for better performance

abogen/web/conversion_runner.py

@@ -8,6 +8,7 @@ import subprocess
 import sys
 import tempfile
 import traceback
+import gc
 from datetime import datetime
 from collections import defaultdict
 from contextlib import ExitStack
@@ -1949,6 +1950,17 @@ def run_conversion_job(job: Job) -> None:
         if subtitle_writer:
             subtitle_writer.close()
 
+        # Explicitly release the pipeline and force garbage collection to prevent
+        # memory accumulation in the worker process, which can lead to host lockups.
+        pipeline = None
+        gc.collect()
+        try:
+            import torch
+            if torch.cuda.is_available():
+                torch.cuda.empty_cache()
+        except ImportError:
+            pass
+
         if (
             audio_output_path
             and job.output_format.lower() == "m4b"

docker-compose.yaml

@@ -33,6 +33,9 @@ services:
     # below if you are deploying to a CPU-only host.
     deploy:
       resources:
+        limits:
+          cpus: '4.0'
+          memory: 8G
         reservations:
           devices:
             - capabilities: [gpu]