KWIPKwipily

Historical release · 2026-10-14 · Build 17

Kwipily 1.2.0

More ways to prepare, explore and save data

Bring tables, documents and images into your workflow, shape data with focused tools, and inspect or export the results.

These notes describe the original paid release and its limits. The current free native app requires no KWIP account or activation.

What changed

  • Table Input now reads CSV, Parquet, JSON and JSON Lines, Excel, Markdown, XML and HTML tables, with explicit parsing options and read-only database table access.
  • New File Input reads local file metadata, document text and image pixels, with explicit folder, page and frame choices.
  • Polars-backed preparation adds bounded columnar processing. Choose a working sample for exploration; View and Table Output request complete upstream data.
  • Explore captured tables in View with paging, search and sorting, including typed date/time values and verified image cells.
  • Use focused Columns, Filter, Formula, Partition and Preprocess editors, and combine tables with Join, Union and Summarize.
  • New Sort, Unique, Reshape Table and Split Text tools support ordered keys, complementary duplicate outputs, long/wide layouts and delimited or fixed-width extraction.
  • Formula adds guided date/time parsing, formatting, timezone conversion, extraction and differences, with explicit precision and daylight-saving policies.
  • Table Output saves complete selected data as CSV/TSV, Parquet, JSON/NDJSON or Excel, with atomic replacement, precision controls and optional copied-image bundles.
  • Model Output now offers optional shape and type checks, plus coordinate or matrix inspection of actually captured tensor values.
  • Configure data access and batches directly in Train. Set starting values and trainability on individual learned layers, with clearer grouped training and experiment controls.
  • Read generated Code for supported preparation, model and training scopes, inspect verified recorded-run code, and export supported current-workflow packages.
  • Results has clearer summaries, artifact lists and activity tables. Refreshed Learn browsing includes editable practices for the new tools.
  • Fixes workflow-settings navigation and repeated welcome-window updates, and strengthens experiment History verification and failed-export cleanup.
  • Adds the client support needed for future signed delta updates; this release continues to use full installers.

Compatibility at release

  • Apple Silicon Macs running macOS 14 or later.
  • All v1.x updates are included with a v1 purchase. Future major upgrades are separate.
  • First paid activation requires internet access. An issued paid v1 license then works permanently offline.
  • Existing workflows remain readable; ambiguous legacy recipe configurations require explicit repair.

Known issues and limits

  • Readers, table preparation, image bundles, exports and training retain format-specific size and memory limits. A working sample does not establish complete-source results.
  • File Input does not include OCR, audio/video decoding or cloud file sources.
  • Code packages cover supported workflow scopes and require compatible Python dependencies and explicit source or destination bindings.
  • Captured results describe their recorded run; viewing History does not revalidate current source files.
  • Broader hardware and minimum-macOS coverage remain unverified.
Historical build reference

Source: b9d1dd0374bdd1e911d8fe2cb23fb23a6ecf4468

Installer SHA-256: 011319f0446394bb67f41a6555a3bb465c7c7c7d31641f6c9c0763dce92de2ed