Quickstart

This guide helps you install CLTK and run working examples using each backend.

Requirements

• Python 3.13+

Install

• Base library:
  • pip: pip install cltk
• Optional extras (choose any):
  • Stanza: pip install "cltk[stanza]"
  • OpenAI: pip install "cltk[openai]"
  • Mistral: pip install "cltk[mistral]"
  • Ollama (local or cloud): pip install "cltk[ollama]"
• You can combine extras, e.g. pip install "cltk[openai,stanza,ollama,mistral]".

Environment Variables

Remote LLM require an API key for access. To set an environment variable, do one of the following:

• Set environment variables directly through the shell: export OPENAI_API_KEY='<YOUR-SECRET-KEY>').
• Setting the variable inside your script: os.environ["OPENAI_API_KEY"] = "<YOUR-SECRET-KEY>"
• Important: Do not do this if you need to share your code publicly.
• Putting them in a .env file located in the directory from which you run your code:

OPENAI_API_KEY=<YOUR-SECRET-KEU>

Minimal Examples

Stanza

• When NLP() is called without specifying the backend parameter, it defaults to executing Stanza-specific Pipelines.
• Install cltk[stanza].
• Stanza only supports the labeling of morphology, dependency syntax, and lemmatization.
• Language models (which will be installed by as needed).
• See Languages for languages supported by Stanza models.

from cltk import NLP
nlp = NLP("lati1261", backend="stanza", suppress_banner=True)
doc = nlp.analyze("Gallia est omnis divisa in partes tres.")
for w in doc.words[:10]:
    print(w.string, getattr(w.upos, "tag", None), w.lemma)

OpenAI

• Install cltk[openai].
• Requires OPENAI_API_KEY.
• Defaults to model gpt-5-mini if not specified.

import os
os.environ["OPENAI_API_KEY"] = "..."  # or set in your shell/.env

from cltk import NLP
nlp = NLP("lati1261", backend="openai", suppress_banner=True)
doc = nlp.analyze("Gallia est omnis divisa in partes tres.")
print("FORM\tLEMMA\tUPOS\tFEATS")
for w in doc.words:
    upos = getattr(w.upos, "tag", "_")
    feats = "_"
    if getattr(w, "features", None) and getattr(w.features, "features", None):
        items = []
        for f in w.features.features:
            if getattr(f, "key", None) and getattr(f, "value", None):
                items.append(f"{f.key}={f.value}")
        feats = "|".join(items) if items else "_"
    print(f"{w.string}\t{w.lemma}\t{upos}\t{feats}")

Ollama

Ollama Local

• Install cltk[ollama].
• Install the Ollama server and run on http://127.0.0.1:11434).
• Defaults to model llama3.1:8b if not specified. You can pass any available model string.
• Note: You may override the Ollama host, port, and model. See Advanced Configuration

from cltk import NLP
nlp = NLP("lati1261", backend="ollama", suppress_banner=True)
doc = nlp.analyze("Gallia est omnis divisa in partes tres.")
print(len(doc.words), "tokens")

Ollama Cloud

• Install cltk[ollama].
• Set OLLAMA_CLOUD_API_KEY and use backend ollama-cloud.

import os
os.environ["OLLAMA_CLOUD_API_KEY"] = "oc-..."  # or set in your shell/.env

from cltk import NLP
nlp = NLP("lati1261", backend="ollama-cloud", suppress_banner=True)
doc = nlp.analyze("Gallia est omnis divisa in partes tres.")
print("usage:", doc.genai_use)

Choosing Language Identifiers

• NLP(language_code=...) accepts Glottolog IDs (e.g. "lati1261"), ISO codes (e.g. "lat"), or exact language names (e.g. "Latin"). Internally, CLTK resolves to a Glottolog ID. See [languages.md] for more.