🤖 I have created a release beep boop

2.5.0 (2025-05-23)

⚠ BREAKING CHANGES

• the updated ai.map() parameter list is not backward-compatible

Features

• Add bpd.options.reset() method (#1743) (36c359d)
• Add DataFrame.round method (#1742) (3ea6043)
• Add deferred data uploading (#1720) (1f6442e)
• Add deprecation warning to Gemini-1.5-X, text-embedding-004, and remove remove legacy models in notebooks and docs (#1723) (80aad9a)
• Add structured output for ai map, ai filter and ai join (#1746) (133ac6b)
• Add support for df.loc[list, column(s)] (#1761) (768a757)
• Include bq schema and query string in dry run results (#1752) (bb51147)
• Support unique() for Index (#1750) (27fac78)
• Support astype conversions to and from JSON dtypes (#1716) (8ef4de1)
• Support dtype parameter in read_csv for bigquery engine (#1749) (50dca4c)
• Use read api for some peek ops (#1731) (108f4d2)

Bug Fixes

• Fix clip int series with float bounds (#1739) (d451aef)
• Prevent creating unnecessary client objects in multithreaded environments (#1757) (1cf9f5e)
• Reduce bigquery table modification via DML for to_gbq (#1737) (545cdca)
• Stop ignoring arguments to MatrixFactorization.score(X, y) (#1726) (55c07e9)
• Support JSON and STRUCT for bbq.sql_scalar (#1754) (190390b)
• Support str.replace re.compile with flags (#1736) (f8d2cd2)

Performance Improvements

• Faster local data comparison using idenitity (#1738) (2858b1e)

Dependencies

• Avoid gcsfs==2025.5.0 (#1762) (68d5e2c)

Documentation

• Add llm output_schema notebook (#1732) (b2261cc)
• Updated multimodal notebook instructions (#1745) (1df8ca6)
• Use partial ordering mode in the quickstart sample (#1734) (476b7dd)

This PR was generated with Release Please. See documentation.