Python
Datagrunt 4.0: A Rust-Accelerated CSV Core and the AI Feature Removal
Datagrunt 4.0 ships a bundled Rust engine for CSV delimiter and dialect inference (~5x faster), removes the deprecated AI/LLM features, and adds document-level batch PDF processing. Plus the same-day 4.0.1 packaging and security patch.
June 13, 2026
Datagrunt Updates: Context Managers, Class-Level Normalization, and Hardening (v3.2.0 - v3.4.1)
A consolidated review of Datagrunt releases from 3.2.0 through 3.4.1, highlighting class-level column normalization, context manager support, AI feature deprecation, and key performance and correctness fixes.
June 13, 2026
Datagrunt 3.1.40: A Hardening and Performance Release
Datagrunt 3.1.40 lands 33 merged pull requests focused on correctness, robustness, security, and speed. Highlights include eliminating silent CSV data loss, fail-fast input validation, ~500x faster repeated DuckDB queries, bounded-memory sampling, thread-safe PDF execution, a contained image-dedupe security fix, and a test suite that grew from 281 to 408 tests.
June 11, 2026
Datagrunt 3.1.10: Defaulting to Sequential PDF Execution for Guard-Free Scripts
Datagrunt 3.1.10 changes the default worker count for PDF readers and writers to 1. This avoids process-spawning overhead for small documents, prevents distributed container conflicts, and removes the need for __main__ guards in simple scripts.
June 9, 2026
Datagrunt 3.1.9: Fixing DuckDB CSV Quote Auto-Detection for Large Datasets
Datagrunt 3.1.9 fixes a bug in the DuckDB query engine where CSV files containing quoted fields with commas later in the file would fail to parse correctly.
June 9, 2026
Datagrunt 3.1.8: Pre-Parsed JSON & Dict Support for PDF Exports
Datagrunt 3.1.8 introduces direct JSON and dict inputs for PDFWriter and PDFReader, cutting PDF-to-Markdown export times by 50% for multi-format pipelines.
June 9, 2026
Exporting PDFs to Clean Markdown: Bridging Structured PDFs and LLMs
Why we built Datagrunt's PDF-to-Markdown export capability, the value of formatted markdown for RAG and LLMs, and how it handles tables, list structures, and images under the hood.
June 7, 2026
Optimizing PDFium: Resolving Multiprocessing Overhead and Layout Parser Recursion
How we optimized datagrunt's new default PDF engine (PDFium) to run 200x faster on single-page documents and fixed a critical infinite recursion bug in layout-aware column splitting.
June 7, 2026
Datagrunt 3.1.7: Resilient CSV Ingestion with Lenient Loading
Datagrunt 3.1.7 delivers robust CSV parsing for real-world datasets. Featuring strict fail-loudly parsing by default, a new lenient=True parameter with warnings for mismatched columns, native detection for legacy carriage return (\r) newlines, and efficiency optimizations for gigabyte-scale files.
June 8, 2026
Datagrunt 3.1.0: A Permissively-Licensed PDF Engine by Default
Datagrunt 3.1.0 makes PDFium the default PDF engine — permissively licensed (BSD-3 / Apache-2.0), fast, and a drop-in for the existing unified output. PyMuPDF remains available, and a new lean mode is 20-80x faster when you don't need tables.
June 6, 2026