Skip to content

Releases: mythopoeic/sheet-compressor

Release list

v0.1.1 — docs: encoding + LLM prompt examples

v0.1.1 — docs: encoding + LLM prompt examples Latest
Latest

Choose a tag to compare

View all tags
@mythopoeic mythopoeic released this 18 Jun 14:05
v0.1.1
cbb56e1

Docs-only patch. No code changes — all encodings and APIs are identical to v0.1.0.

What's new

  • READMEs now show the same sheet in all three encodings (structural-anchor, inverted index, format aggregation) with real output strings.
  • New "Reading the output with an LLM" sections: how to pair the bundled reader/task prompt templates with a compressed string to call a model (runnable Claude example; any chat model works).

Install (unchanged commands; now resolve to 0.1.1)

  • npm: npm install sheet-compressor
  • PyPI: pip install sheet-compressor
  • NuGet: dotnet add package SheetCompressor (+ SheetCompressor.Xlsx)
  • Go: go get github.com/mythopoeic/sheet-compressor/packages/go/sheetcompressor@v0.1.1
Assets 2
Loading

v0.1.0 — first public release

v0.1.0 — first public release

Choose a tag to compare

View all tags
@mythopoeic mythopoeic released this 18 Jun 04:46
v0.1.0
a64c916

First public release of sheet-compressor — an independent, multi-language implementation of the SheetCompressor encoding from the SpreadsheetLLM paper (Dong et al., Microsoft, 2024). It turns a spreadsheet into a compact, LLM-friendly text representation — three interchangeable encodings plus token estimates — and makes no LLM calls itself.

Independent, community implementation. Not affiliated with or endorsed by Microsoft.

Install

  • npm (TypeScript / Node ≥ 20.11) — npm install sheet-compressor
  • PyPI (Python ≥ 3.9) — pip install sheet-compressor
  • Go — go get github.com/mythopoeic/sheet-compressor/packages/go/sheetcompressor@v0.1.0
  • C# (NuGet) — not yet published; build from source (packages/csharp)
  • VBA / Office Script — copy from source (packages/vba, packages/officescript)

What's in 0.1.0

  • Three encodings — structural-anchor skeleton, inverted index, format aggregation — each as string + JSON + token estimate.
  • Portable CHART(...) chart descriptors and a raw-token baseline.
  • Six implementations (TypeScript, Python, C#, Go, VBA, Office Script) verified byte-for-byte against one shared golden corpus.
  • Per-encoding reader prompt templates shipped with each package.

npm and PyPI are live; the Go module is importable at the tag above. See the README and spec.

Loading