fix summries

Author: antonkulaga

AGENTS.md

@@ -27,7 +27,7 @@ This document outlines the coding standards and practices to follow when working
 ## Data Processing
 
 - **Prefer Polars**: Prefer Polars over Pandas for data processing
-- **Be memory efficient**: use scan_parquet and lazy api. Take into account that scan_parquet can open content of whole folder instead of opening separately. Also, you can write lazyly using sink_parquet. You can also apply compression (check what sink parquet supports) when saving.
+- **Be memory efficient**: use scan_parquet and lazy api. Take into account that scan_parquet can open content of whole folder instead of opening separately. Also, you can write lazyly using sink_parquet. You can also apply compression (check what sink parquet supports) when saving. Avoid doing pl.scan_parquet().collect() if possible as it negates the benefits of lazy streaming.
 - **Use data/input, data/interim, data/output pattern** . Input and downloaded files you store at ./data/input, intermediates and interim and final at data/output
 
 ## Logging

docs/EXPLORE_CLI.md

@@ -137,3 +137,4 @@ Without `--log-dir`, logs are written to stdout and a JSON file in the current d
 
 
 
+

src/cell2sentence4longevity/benchmarks/__init__.py

@@ -3,3 +3,4 @@
 
 
 
+