tact-dictionary 0.2.0

Build, validate, and query .tactdict language packs

Tact Dictionary

Tact Dictionary builds, validates, and queries .tactdict language packs for text-entry software.

The tact-dictionary tool and related Python package can prepare source data, build binary packs, validate pack files, and inspect completions and variants. The project was built with the Tact Android keyboard in mind, but the keyboard is an adopter of the format rather than the only intended consumer. Current packs contain a normalized lexicon graph, word records, prefix suggestion caches, and variant groups for long-press word alternatives.

The tool and Python package are licensed under the MIT License. Generated language packs are data artifacts with source-derived licensing and attribution recorded in the sidecar manifest. The code license does not relicense frequency lists, morphology data, blocklists, or other input datasets.

Status

This project is pre-alpha. The tool can produce and validate an English pack, and the Python API can inspect completions and variants.

Implemented pack sections:

• STRINGS: shared UTF-8 string pool
• SYMBOLS: normalized input symbols
• WORDS: word records and source masks
• VARIANTS: variant groups such as run, runs, running, runner
• SUGGESTS: prefix suggestion lists attached to graph states
• GRAPH: lexicon word graph

Not implemented yet: context models, full typo-correction graph data, blocklist sections, user overlays, and runtime memory-map readers.

Development

uv sync
uv run tact-dictionary --version
uv run ruff format --check src tests
uv run ruff check
uv run pytest

Build Python distribution artifacts with:

uv build

Build A Pack

Download the current English source files:

• Leipzig Corpora Collection English news corpus: https://downloads.wortschatz-leipzig.de/corpora/eng_news_2025_100K.tar.gz
• MorphyNet English inflection data: https://raw.githubusercontent.com/kbatsuren/MorphyNet/main/eng/eng.inflectional.v1.tsv
• MorphyNet English derivation data: https://raw.githubusercontent.com/kbatsuren/MorphyNet/main/eng/eng.derivational.v1.tsv

Convert external source files into tact-dictionary build inputs:

uv run tact-dictionary import-language-data \
  --output-dir build/en_US \
  --leipzig-corpus sources/eng_news_2025_100K.tar.gz \
  --morphynet-inflection sources/eng.inflectional.v1.tsv \
  --morphynet-derivation sources/eng.derivational.v1.tsv

Build and validate the pack:

uv run tact-dictionary build --config build/en_US/en_US.tactbuild
uv run tact-dictionary validate build/en_US/generated/en_US.tactdict

The build writes:

• build/en_US/generated/en_US.tactdict
• build/en_US/generated/en_US.manifest.json
• build/en_US/generated/en_US.symbols.json

See Building Language Packs for input file formats and build configuration details.

Query A Pack

Python consumers can inspect a built pack without shelling out:

from tact_dictionary.query import DictionaryLookup

lookup = DictionaryLookup.from_file("en_US.tactdict")
for completion in lookup.find_completions("ru"):
    print(completion.surface, completion.kind, completion.completion_cost)

for variant in lookup.find_variants("run"):
    print(variant.surface, variant.relation_to_primary)

The same lookup is available from the CLI:

tact-dictionary --file en_US.tactdict find-completions --limit 12 ru
tact-dictionary --file en_US.tactdict find-variants run

Documentation

• Building Language Packs
• Pack Format
• Release Checklist