termlint 0.1.0a4

Terminology linter for a given subject area text

termlint

Terminology linter for projects. termlint extracts term candidates from text and checks them against your glossary or ontology.

Alpha Status

termlint is currently alpha.

Implemented now:

• rule-based extraction (RuleExtractor / spaCy)
• C-Value extraction (CValueExtractor)
• verification: exact, fuzzy
• JSON reports: verification, ontology_update, quality_gate, extraction
• glossary tooling: glossary from-report, glossary merge

Current support is intentionally narrow:

• Python >=3.10
• spaCy-based extraction is currently supported on Python <3.14
• officially tested with English and Russian spaCy models
• other spaCy models may work, but should be treated as experimental in this alpha stage

Quick Start

1. Install:

# Recommended for CLI usage
pipx install "termlint[base]"

# Alternative: install into a project environment
pip install --pre "termlint[base]"

# Install a spaCy model into the same environment
python -m spacy download en_core_web_sm

For pipx, install the model inside the pipx environment:

pipx runpip termlint install en-core-web-sm
# or for Russian
pipx runpip termlint install ru-core-news-sm

2. Create a glossary (glossary.json):

[
  { "id": "ml:001", "label": "machine learning", "synonyms": ["ML"] },
  { "id": "ml:002", "label": "artificial intelligence", "synonyms": ["AI"] }
]

3. Create an input file (input.txt):

Artificial intelligence and machine learning are used in data analytics.

4. Run verification:

termlint verify input.txt --source glossary.json --verifier fuzzy --threshold 85

This command works without any local pyproject.toml. In the no-config case, termlint uses built-in defaults for the full verification pipeline and English spaCy model settings.

Example output:

Files     ... 100%
✅ input.txt ... 100%
📊 Coverage: 33.3% (2/6)
⚠️  Quality Gate would FAIL in CI mode

Generated reports:

• reports/verification.json
• reports/ontology_update.json
• reports/quality_gate.json

Exit behavior:

• verify exits 0 on a successful run by default, even if the quality gate would fail in CI mode
• verify --fail-on-quality-gate exits 1 when quality gates fail

Configuration

Project configuration lives in pyproject.toml under [tool.termlint].

Example:

[tool.termlint.logging]
level = "WARNING"
log_file = "reports/termlint.log"

[tool.termlint.extraction]
extractors = ["rule", "cvalue"]
rules = { model = "en_core_web_sm", auto_download_model = false }
cvalue = { threshold = 0.25, min_freq = 1, min_length = 2, max_length = 4, use_ling_filter = true, model = "en_core_web_sm", auto_download_model = false }

Notes:

• use termlint -v or termlint -vv for more verbose logs
• keep auto_download_model = false in CI or reproducible environments
• glossary sources are JSON arrays of entities with required fields id and label

Minimal valid glossary:

[
  {
    "id": "ml:001",
    "label": "machine learning"
  }
]

Config lookup order:

1. --config <PATH>
2. nearest pyproject.toml with [tool.termlint]
3. user config:
   • $XDG_CONFIG_HOME/termlint/config.toml
   • ~/.config/termlint/config.toml
   • %APPDATA%/termlint/config.toml
   • ~/.termlint/config.toml
4. built-in defaults

User config files may use either [tool.termlint] or [termlint].

More Docs

• Usage examples: samples/USAGE.md
• Architecture: docs/ARCHITECTURE.md
• Release history: CHANGELOG.md

License

This project is licensed under the MIT License. See LICENSE.