Try this

sql/combine.sql

@@ -6,5 +6,5 @@ FROM read_parquet('input/*.parquet', union_by_name = True, filename = true);
 COPY (
     select *
     from temp_table
-    order by (path)
+    order by (project_name)
 ) TO 'output.parquet' (FORMAT PARQUET, COMPRESSION zstd);

sql/unique_python_files.prql

@@ -3,4 +3,9 @@ prql target:sql.duckdb
 from (read_parquet $1)
 filter s"suffix({path}, '.py')"
 filter skip_reason == ""
-select {s"distinct on({hash}) {hash}", path}
+select {hash, repository, path, uploaded_on}
+group {hash} (
+  sort uploaded_on
+  take 1
+)
+sort path

sql/unique_python_files_combine.prql

@@ -1,5 +1,9 @@
 prql target:sql.duckdb
 
 from (read_parquet $1)
-select {s"distinct on({hash}) {hash}", path}
-
+select {hash, repository, path, uploaded_on}
+group {hash} (
+  sort uploaded_on
+  take 1
+)
+sort path