Installation

Installation

pyaegean requires Python ≥ 3.10. The core has zero hard third-party dependencies — the wheel ships code and JSON only, so import aegean is instant. Everything heavier is an optional extra: pandas (DataFrame interop), the provider AI SDKs, and the Greek NLP backends are all imported lazily inside their adapters and pulled in only when you ask for them.

From PyPI

pip install pyaegean            # core: Linear A + Greek, zero hard deps

Optional extras

Extra	Pulls in	For
pyaegean[data]	pandas	DataFrame interop (to_dataframe)
pyaegean[neural]	onnxruntime, tokenizers, numpy	the neural Greek pipeline (use_neural_pipeline()) and lemmatizer (use_neural_lemmatizer())
pyaegean[anthropic]	anthropic	Anthropic (default) AI provider
pyaegean[openai]	openai	OpenAI provider
pyaegean[grok]	openai	xAI Grok (OpenAI-API-compatible)
pyaegean[gemini]	google-genai	Google Gemini provider
pyaegean[ai]	all of the above providers	the full AI layer
pyaegean[epidoc]	lxml	EpiDoc I/O
pyaegean[geo]	geopandas, shapely	geographic analysis
pyaegean[parquet]	pyarrow	Parquet export (io.to_parquet)
pyaegean[all]	ai, epidoc, geo, data	everything except neural and parquet

pip install "pyaegean[ai]"
pip install "pyaegean[neural]"
pip install "pyaegean[all]"

Verify

import aegean
print(aegean.__version__)
print(aegean.registered_scripts())       # ['cypriot', 'cyprominoan', 'greek', 'lineara', 'linearb']
print(len(aegean.load("lineara")))        # 1721
print(len(aegean.load("greek")))          # 5  (bundled sample corpus)

Offline & data

The compact text corpora (Linear A inscriptions/signs, Greek seeds) ship inside the wheel and work fully offline. Large assets are not bundled — they are fetched on demand into a user cache on first use: the ~116 MB Linear A facsimile imagery, plus the opt-in Greek backends' data — the Perseus AGDT treebank (~75 MB, greek.use_treebank()) and the full Perseus LSJ (~270 MB, greek.use_lsj()). The pure-Python backends each cache a small trained model: greek.use_parser(), greek.use_tagger(), and greek.use_lemmatizer(). The neural lemmatizer (greek.use_neural_lemmatizer(), the [neural] extra) fetches a ~232 MB int8 ONNX GreTa model; the neural joint pipeline (greek.use_neural_pipeline(), same extra) fetches a ~518 MB fp32 ONNX model bundle. All remain offline after the first fetch. See Data & Provenance.

From source

See Development.