💎 ruby-spacy

Overview

ruby-spacy is a wrapper module for using spaCy from the Ruby programming language via PyCall. This module aims to make it easy and natural for Ruby programmers to use spaCy. This module covers the areas of spaCy functionality for using many varieties of its language models, not for building ones.

	Functionality
✅	Tokenization, lemmatization, sentence segmentation
✅	Part-of-speech tagging and dependency parsing
✅	Named entity recognition
✅	Syntactic dependency visualization
✅	Access to pre-trained word vectors
✅	LLM integration: OpenAI, Anthropic (Claude), and local models

Current Version: 0.6.0

• Ruby 3.2 to 4.0 supported (PyCall 1.5.3 or later required)
• spaCy 3.8 supported
• Multi-provider LLM API: OpenAI, Anthropic (Claude), and local models via Ollama or any OpenAI-compatible server
• Structured outputs (JSON Schema) support
• Block-based LLM API with linguistic analysis

Installation of Prerequisites

IMPORTANT: Make sure that the enable-shared option is enabled in your Python installation. You can use pyenv to install any version of Python you like. spaCy 3.8 supports Python 3.10 and later (wheels are provided for Python 3.10–3.14), so we recommend using one of those versions. Install Python 3.13, for instance, using pyenv with enable-shared as follows:

$ env CONFIGURE_OPTS="--enable-shared" pyenv install 3.13

Remember to make it accessible from your working directory. It is recommended that you set global to the version of python you just installed.

$ pyenv global 3.13

Then, install spaCy. If you use pip, the following command will do:

$ pip install spacy

Install trained language models. For a starter, en_core_web_sm will be the most useful to conduct basic text processing in English. However, if you want to use advanced features of spaCy, such as named entity recognition or document similarity calculation, you should also install a larger model like en_core_web_lg.

$ python -m spacy download en_core_web_sm
$ python -m spacy download en_core_web_lg

See Spacy: Models & Languages for other models in various languages. To install models for the Japanese language, for instance, you can do it as follows:

$ python -m spacy download ja_core_news_sm
$ python -m spacy download ja_core_news_lg

Installation of ruby-spacy

Add this line to your application's Gemfile:

gem'ruby-spacy'

And then execute:

$ bundle install

Or install it yourself as:

$ gem install ruby-spacy

Usage

See Examples below.

Examples

Many of the following examples are Python-to-Ruby translations of code snippets in spaCy 101. For more examples, look inside the examples directory.

Tokenization

→ spaCy: Tokenization

Ruby code:

require"ruby-spacy"require"terminal-table"nlp=Spacy::Language.new("en_core_web_sm")doc=nlp.read("Apple is looking at buying U.K. startup for $1 billion")row=[]doc.eachdo |token|
row << token.textendheadings=[1,2,3,4,5,6,7,8,9,10]table=Terminal::Table.newrows: [row],headings: headingsputstable

Output:

1	2	3	4	5	6	7	8	9	10	11
Apple	is	looking	at	buying	U.K.	startup	for	$	1	billion

Part-of-speech and Dependency

→ spaCy: Part-of-speech tags and dependencies

Ruby code:

require"ruby-spacy"require"terminal-table"nlp=Spacy::Language.new("en_core_web_sm")doc=nlp.read("Apple is looking at buying U.K. startup for $1 billion")headings=["text","lemma","pos","tag","dep"]rows=[]doc.eachdo |token|
rows << [token.text,token.lemma,token.pos,token.tag,token.dep]endtable=Terminal::Table.newrows: rows,headings: headingsputstable

Output:

text	lemma	pos	tag	dep
Apple	Apple	PROPN	NNP	nsubj
is	be	AUX	VBZ	aux
looking	look	VERB	VBG	ROOT
at	at	ADP	IN	prep
buying	buy	VERB	VBG	pcomp
U.K.	U.K.	PROPN	NNP	dobj
startup	startup	NOUN	NN	advcl
for	for	ADP	IN	prep
$	$	SYM	$	quantmod
1	1	NUM	CD	compound
billion	billion	NUM	CD	pobj

Part-of-speech and Dependency (Japanese)

Ruby code:

require"ruby-spacy"require"terminal-table"nlp=Spacy::Language.new("ja_core_news_lg")doc=nlp.read("任天堂は1983年にファミコンを14,800円で発売した。")headings=["text","lemma","pos","tag","dep"]rows=[]doc.eachdo |token|
rows << [token.text,token.lemma,token.pos,token.tag,token.dep]endtable=Terminal::Table.newrows: rows,headings: headingsputstable

Output:

text	lemma	pos	tag	dep
任天堂	任天堂	PROPN	名詞-固有名詞-一般	nsubj
は	は	ADP	助詞-係助詞	case
1983	1983	NUM	名詞-数詞	nummod
年	年	NOUN	名詞-普通名詞-助数詞可能	obl
に	に	ADP	助詞-格助詞	case
ファミコン	ファミコン	NOUN	名詞-普通名詞-一般	obj
を	を	ADP	助詞-格助詞	case
14,800	14,800	NUM	名詞-数詞	fixed
円	円	NOUN	名詞-普通名詞-助数詞可能	obl
で	で	ADP	助詞-格助詞	case
発売	発売	VERB	名詞-普通名詞-サ変可能	ROOT
し	する	AUX	動詞-非自立可能	aux
た	た	AUX	助動詞	aux
。	。	PUNCT	補助記号-句点	punct

Morphology

→ POS and morphology tags

Ruby code:

require"ruby-spacy"require"terminal-table"nlp=Spacy::Language.new("en_core_web_sm")doc=nlp.read("Apple is looking at buying U.K. startup for $1 billion")headings=["text","shape","is_alpha","is_stop","morphology"]rows=[]doc.eachdo |token|
morph=token.morphology.mapdo |k,v|
"#{k} = #{v}"end.join("\n")rows << [token.text,token.shape,token.is_alpha,token.is_stop,morph]endtable=Terminal::Table.newrows: rows,headings: headingsputstable

Output:

text	shape	is_alpha	is_stop	morphology
Apple	Xxxxx	true	false	NounType = Prop Number = Sing
is	xx	true	true	Mood = Ind Number = Sing Person = 3 Tense = Pres VerbForm = Fin
looking	xxxx	true	false	Aspect = Prog Tense = Pres VerbForm = Part
at	xx	true	true
buying	xxxx	true	false	Aspect = Prog Tense = Pres VerbForm = Part
U.K.	X.X.	false	false	NounType = Prop Number = Sing
startup	xxxx	true	false	Number = Sing
for	xxx	true	true
$	$	false	false
1	d	false	false	NumType = Card
billion	xxxx	true	false	NumType = Card

Visualizing Dependency

→ spaCy: Visualizers

Ruby code:

require"ruby-spacy"nlp=Spacy::Language.new("en_core_web_sm")sentence="Autonomous cars shift insurance liability toward manufacturers"doc=nlp.read(sentence)dep_svg=doc.displacy(style: "dep",compact: false)File.open(File.join("test_dep.svg"),"w")do |file|
file.write(dep_svg)end

Output:

Visualizing Dependency (Compact)

Ruby code:

require"ruby-spacy"nlp=Spacy::Language.new("en_core_web_sm")sentence="Autonomous cars shift insurance liability toward manufacturers"doc=nlp.read(sentence)dep_svg=doc.displacy(style: "dep",compact: true)File.open(File.join("test_dep_compact.svg"),"w")do |file|
file.write(dep_svg)end

Output:

Named Entity Recognition

→ spaCy: Named entities

Ruby code:

require"ruby-spacy"require"terminal-table"nlp=Spacy::Language.new("en_core_web_sm")doc=nlp.read("Apple is looking at buying U.K. startup for $1 billion")rows=[]doc.ents.eachdo |ent|
rows << [ent.text,ent.start_char,ent.end_char,ent.label]endheadings=["text","start_char","end_char","label"]table=Terminal::Table.newrows: rows,headings: headingsputstable

Output:

text	start_char	end_char	label
Apple	0	5	ORG
U.K.	27	31	GPE
$1 billion	44	54	MONEY

Named Entity Recognition (Japanese)

Ruby code:

require("ruby-spacy")require"terminal-table"nlp=Spacy::Language.new("ja_core_news_lg")sentence="任天堂は1983年にファミコンを14,800円で発売した。"doc=nlp.read(sentence)rows=[]doc.ents.eachdo |ent|
rows << [ent.text,ent.start_char,ent.end_char,ent.label]endheadings=["text","start","end","label"]table=Terminal::Table.newrows: rows,headings: headingsprinttable

Output:

text	start	end	label
任天堂	0	3	ORG
1983年	4	9	DATE
ファミコン	10	15	PRODUCT
14,800円	16	23	MONEY

Checking Availability of Word Vectors

→ spaCy: Word vectors and similarity

Ruby code:

require"ruby-spacy"require"terminal-table"nlp=Spacy::Language.new("en_core_web_lg")doc=nlp.read("dog cat banana afskfsd")rows=[]doc.eachdo |token|
rows << [token.text,token.has_vector,token.vector_norm,token.is_oov]endheadings=["text","has_vector","vector_norm","is_oov"]table=Terminal::Table.newrows: rows,headings: headingsputstable

Output:

text	has_vector	vector_norm	is_oov
dog	true	7.0336733	false
cat	true	6.6808186	false
banana	true	6.700014	false
afskfsd	false	0.0	true

Similarity Calculation

Ruby code:

require"ruby-spacy"nlp=Spacy::Language.new("en_core_web_lg")doc1=nlp.read("I like salty fries and hamburgers.")doc2=nlp.read("Fast food tastes very good.")puts"Doc 1: " + doc1.textputs"Doc 2: " + doc2.textputs"Similarity: #{doc1.similarity(doc2)}"

Output:

Doc 1: I like salty fries and hamburgers.
Doc 2: Fast food tastes very good.
Similarity: 0.7687607012190486

Similarity Calculation (Japanese)

Ruby code:

require"ruby-spacy"nlp=Spacy::Language.new("ja_core_news_lg")ja_doc1=nlp.read("今日は雨ばっかり降って、嫌な天気ですね。")puts"doc1: #{ja_doc1.text}"ja_doc2=nlp.read("あいにくの悪天候で残念です。")puts"doc2: #{ja_doc2.text}"puts"Similarity: #{ja_doc1.similarity(ja_doc2)}"

Output:

doc1: 今日は雨ばっかり降って、嫌な天気ですね。
doc2: あいにくの悪天候で残念です。
Similarity: 0.8684192637149641

Word Vector Calculation

Tokyo - Japan + France = Paris ?

Ruby code:

require"ruby-spacy"require"terminal-table"nlp=Spacy::Language.new("en_core_web_lg")tokyo=nlp.get_lexeme("Tokyo")japan=nlp.get_lexeme("Japan")france=nlp.get_lexeme("France")query=tokyo.vector - japan.vector + france.vectorheadings=["rank","text","score"]rows=[]results=nlp.most_similar(query,10)results.each_with_indexdo |lexeme,i|
index=(i + 1).to_srows << [index,lexeme.text,lexeme.score]endtable=Terminal::Table.newrows: rows,headings: headingsputstable

Output:

rank	text	score
1	FRANCE	0.8346999883651733
2	France	0.8346999883651733
3	france	0.8346999883651733
4	PARIS	0.7703999876976013
5	paris	0.7703999876976013
6	Paris	0.7703999876976013
7	TOULOUSE	0.6381999850273132
8	Toulouse	0.6381999850273132
9	toulouse	0.6381999850273132
10	marseille	0.6370999813079834

Word Vector Calculation (Japanese)

東京 - 日本 + フランス = パリ ?

Ruby code:

require"ruby-spacy"require"terminal-table"nlp=Spacy::Language.new("ja_core_news_lg")tokyo=nlp.get_lexeme("東京")japan=nlp.get_lexeme("日本")france=nlp.get_lexeme("フランス")query=tokyo.vector - japan.vector + france.vectorheadings=["rank","text","score"]rows=[]results=nlp.most_similar(query,10)results.each_with_indexdo |lexeme,i|
index=(i + 1).to_srows << [index,lexeme.text,lexeme.score]endtable=Terminal::Table.newrows: rows,headings: headingsputstable

Output:

rank	text	score
1	パリ	0.7376999855041504
2	フランス	0.7221999764442444
3	東京	0.6697999835014343
4	ストラスブール	0.631600022315979
5	リヨン	0.5939000248908997
6	Paris	0.574400007724762
7	ベルギー	0.5683000087738037
8	ニース	0.5679000020027161
9	アルザス	0.5644999742507935
10	南仏	0.5547999739646912

Matcher

Matcher finds token sequences with rule-based patterns.

require"ruby-spacy"nlp=Spacy::Language.new("en_core_web_sm")matcher=nlp.matchermatcher.add("GREETING",[[{LOWER: "hello"},{IS_PUNCT: true},{LOWER: "world"}]])doc=nlp.read("Hello, world!")matcher.match(doc).eachdo |match|
span=doc.span(match[:start_index]..match[:end_index])puts"#{match[:label]}: #{span.text}"end# => GREETING: Hello, world

Matcher#match returns an array of hashes with :match_id (the label's numeric id), :start_index, :end_index, and :label (the label string).

See examples/rule_based_matching/ for more examples.

PhraseMatcher

PhraseMatcher is more efficient than Matcher for matching large terminology lists. It's ideal for extracting known entities like product names, company names, or domain-specific terms.

Basic usage:

require"ruby-spacy"nlp=Spacy::Language.new("en_core_web_sm")# Create a phrase matchermatcher=nlp.phrase_matchermatcher.add("PRODUCT",["iPhone","MacBook Pro","iPad"])doc=nlp.read("I bought an iPhone and a MacBook Pro yesterday.")matches=matcher.match(doc)matches.eachdo |span|
puts"#{span.text} => #{span.label}"end# => iPhone => PRODUCT# => MacBook Pro => PRODUCT

Case-insensitive matching:

# Use attr: "LOWER" for case-insensitive matchingmatcher=nlp.phrase_matcher(attr: "LOWER")matcher.add("COMPANY",["apple","google","microsoft"])doc=nlp.read("Apple and GOOGLE are competitors of Microsoft.")matches=matcher.match(doc)matches.eachdo |span|
putsspan.textend# => Apple# => GOOGLE# => Microsoft

Multiple categories:

matcher=nlp.phrase_matcher(attr: "LOWER")matcher.add("TECH_COMPANY",["apple","google","microsoft","amazon"])matcher.add("PRODUCT",["iphone","pixel","surface","kindle"])doc=nlp.read("Apple released the new iPhone while Google announced Pixel updates.")matches=matcher.match(doc)matches.eachdo |span|
puts"#{span.text}: #{span.label}"end# => Apple: TECH_COMPANY# => iPhone: PRODUCT# => Google: TECH_COMPANY# => Pixel: PRODUCT

OpenAI API Integration

⚠️ This feature requires GPT-5 series models. Please refer to OpenAI's API reference for details.

ℹ️ The temperature parameter is sent to the API only when you specify it explicitly. If the model does not support it (e.g., GPT-5 series and o-series models), the request is automatically retried once without it — no per-model configuration is needed.

Easily leverage GPT models within ruby-spacy by using an OpenAI API key. When constructing prompts for the Doc::openai_query method, you can incorporate the following token properties of the document. These properties are retrieved through tool calls (made internally by GPT when necessary) and seamlessly integrated into your prompt. The available properties include:

• surface
• lemma
• tag
• pos (part of speech)
• dep (dependency)
• ent_type (entity type)
• morphology

GPT Prompting (Translation)

Ruby code:

require"ruby-spacy"api_key=ENV["OPENAI_API_KEY"]nlp=Spacy::Language.new("en_core_web_sm")doc=nlp.read("The Beatles released 12 studio albums")# default parameter values# max_completion_tokens: 1000# model: "gpt-5-mini"res1=doc.openai_query(access_token: api_key,prompt: "Translate the text to Japanese.")putsres1

Output:

ビートルズは12枚のスタジオアルバムをリリースしました。

GPT Prompting (Elaboration)

Ruby code:

require"ruby-spacy"api_key=ENV["OPENAI_API_KEY"]nlp=Spacy::Language.new("en_core_web_sm")doc=nlp.read("The Beatles were an English rock band formed in Liverpool in 1960.")# default parameter values# max_completion_tokens: 1000# model: "gpt-5-mini"res=doc.openai_query(access_token: api_key,prompt: "Extract the topic of the document and list 10 entities (names, concepts, locations, etc.) that are relevant to the topic.")

Output:

Topic: The Beatles

Relevant Entities:

1. The Beatles (PERSON)
2. Liverpool (GPE - Geopolitical Entity)
3. English (LANGUAGE)
4. Rock (MUSIC GENRE)
5. 1960 (DATE)
6. Band (MUSIC GROUP)
7. John Lennon (PERSON - key member)
8. Paul McCartney (PERSON - key member)
9. George Harrison (PERSON - key member)
10. Ringo Starr (PERSON - key member)

GPT Prompting (JSON Output Using RAG with Token Properties)

Ruby code:

require"ruby-spacy"api_key=ENV["OPENAI_API_KEY"]nlp=Spacy::Language.new("en_core_web_sm")doc=nlp.read("The Beatles released 12 studio albums")# default parameter values# max_completion_tokens: 1000# model: "gpt-5-mini"res=doc.openai_query(access_token: api_key,prompt: "List token data of each of the words used in the sentence. Add 'meaning' property and value (brief semantic definition) to each token data. Output as a JSON object.")

Output:

{
"tokens": [
{
"surface": "The",
"lemma": "the",
"pos": "DET",
"tag": "DT",
"dep": "det",
"ent_type": "",
"morphology": "{'Definite': 'Def', 'PronType': 'Art'}",
"meaning": "A definite article used to specify a noun."
},
{
"surface": "Beatles",
"lemma": "beatle",
"pos": "NOUN",
"tag": "NNS",
"dep": "nsubj",
"ent_type": "GPE",
"morphology": "{'Number': 'Plur'}",
"meaning": "A British rock band formed in Liverpool in 1960."
},
{
"surface": "released",
"lemma": "release",
"pos": "VERB",
"tag": "VBD",
"dep": "ROOT",
"ent_type": "",
"morphology": "{'Tense': 'Past', 'VerbForm': 'Fin'}",
"meaning": "To make something available to the public."
},
{
"surface": "12",
"lemma": "12",
"pos": "NUM",
"tag": "CD",
"dep": "nummod",
"ent_type": "CARDINAL",
"morphology": "{'NumType': 'Card'}",
"meaning": "A cardinal number representing the quantity of twelve."
},
{
"surface": "studio",
"lemma": "studio",
"pos": "NOUN",
"tag": "NN",
"dep": "compound",
"ent_type": "",
"morphology": "{'Number': 'Sing'}",
"meaning": "A place where recording or filming takes place."
},
{
"surface": "albums",
"lemma": "album",
"pos": "NOUN",
"tag": "NNS",
"dep": "dobj",
"ent_type": "",
"morphology": "{'Number': 'Plur'}",
"meaning": "Collections of music tracks or recordings."
}
]
}

GPT Prompting (Generate a Syntax Tree using Token Properties)

Ruby code:

require"ruby-spacy"api_key=ENV["OPENAI_API_KEY"]nlp=Spacy::Language.new("en_core_web_sm")doc=nlp.read("The Beatles released 12 studio albums")# default parameter values# max_completion_tokens: 1000# model: "gpt-5-mini"res=doc.openai_query(access_token: api_key,prompt: "Generate a tree diagram from the text using given token data. Use the following bracketing style: [S [NP [Det the] [N cat]] [VP [V sat] [PP [P on] [NP the mat]]]")putsres

Output:

[S
[NP
[Det The]
[N Beatles]
]
[VP
[V released]
[NP
[Num 12]
[N
[N studio]
[N albums]
]
]
]
]

GPT Text Completion

Ruby code:

require"ruby-spacy"api_key=ENV["OPENAI_API_KEY"]nlp=Spacy::Language.new("en_core_web_sm")doc=nlp.read("Vladimir Nabokov was a")# default parameter values# max_completion_tokens: 1000# model: "gpt-5-mini"res=doc.openai_completion(access_token: api_key)putsres

Output:

Vladimir Nabokov was a Russian-American novelist, poet, and entomologist, best known for his intricate prose style and innovative narrative techniques. He is most famously recognized for his controversial novel "Lolita," which explores themes of obsession and manipulation. Nabokov's works often reflect his fascination with language, memory, and the nature of art. In addition to his literary accomplishments, he was also a passionate lepidopterist, contributing to the field of butterfly studies. His literary career spanned several decades, and his influence continues to be felt in contemporary literature.

Text Embeddings

Ruby code:

require"ruby-spacy"api_key=ENV["OPENAI_API_KEY"]nlp=Spacy::Language.new("en_core_web_sm")doc=nlp.read("Vladimir Nabokov was a Russian-American novelist, poet, translator and entomologist.")# default model: text-embedding-3-smallres=doc.openai_embeddings(access_token: api_key)putsres

Output:

-0.0023891362
-0.016671216
0.010879759
0.012918914
0.0012281279
...

Block-based OpenAI API

The Language#with_openai block API provides a streamlined way to combine spaCy's linguistic analysis with OpenAI. The Doc#linguistic_summary method generates a JSON summary of spaCy's analysis (tokens, entities, noun chunks, etc.) that can be passed directly to the LLM as context.

Basic usage:

require"ruby-spacy"nlp=Spacy::Language.new("en_core_web_sm")doc=nlp.read("Apple Inc. was founded by Steve Jobs in California.")nlp.with_openai(model: "gpt-5-mini")do |ai|
result=ai.chat(system: "You are a linguistic analyst. Analyze the given linguistic data.",user: doc.linguistic_summary)putsresultend

Batch processing with pipe:

require"ruby-spacy"nlp=Spacy::Language.new("en_core_web_sm")texts=["The bank approved the loan.","I sat on the river bank."]nlp.with_openai(model: "gpt-5-mini")do |ai|
nlp.pipe(texts).eachdo |doc|
result=ai.chat(system: "Identify the meaning of 'bank' in one word based on the linguistic context.",user: doc.linguistic_summary)puts"#{doc.text} => #{result}"endend

Customizing linguistic summary:

# Include sentences and morphology, exclude noun chunkssummary=doc.linguistic_summary(sections: [:text,:tokens,:entities,:sentences],token_attributes: [:text,:lemma,:pos,:dep,:head,:morphology])

Embeddings:

nlp.with_openaido |ai|
vector=ai.embeddings("Hello world")putsvector.length# => 1536end

Multi-provider LLM API

The Language#with_llm block API generalizes with_openai to multiple providers. The helper yielded to the block has the same chat interface for every provider.

Anthropic (Claude):

# Requires the ANTHROPIC_API_KEY environment variable# Default model: claude-sonnet-5 (override with model: "...")nlp.with_llm(provider: :anthropic)do |ai|
result=ai.chat(system: "You are a linguistic analyst.",user: doc.linguistic_summary)putsresultend

Local models via Ollama (no API key needed):

# Requires a running Ollama server: https://ollama.comnlp.with_llm(provider: :ollama,model: "llama3.2")do |ai|
putsai.chat(user: doc.linguistic_summary)end

Any other OpenAI-compatible server (LM Studio, llama.cpp server, vLLM, OpenRouter, etc.) works via base_url::

nlp.with_llm(provider: :openai,base_url: "http://localhost:1234/v1",access_token: "not-needed",model: "your-model")do |ai|
putsai.chat(user: "Hello!")end

Structured outputs (schema:): pass a JSON Schema and receive a validated, parsed Ruby Hash — useful for comparing LLM output with spaCy's analysis programmatically. Works with both :openai and :anthropic. Objects in the schema must set additionalProperties: false.

schema={type: "object",properties: {entities: {type: "array",items: {type: "object",properties: {text: {type: "string"},label: {type: "string"}},required: %w[textlabel],additionalProperties: false}}},required: ["entities"],additionalProperties: false}result=nlp.with_llm(provider: :openai)do |ai|
ai.chat(system: "Extract named entities.",user: doc.text,schema: schema)endresult["entities"].each{ |ent| puts"#{ent["text"]} (#{ent["label"]})"}

Note on temperature: for all providers, temperature is omitted from requests unless you pass it explicitly (ai.chat(user: "...", temperature: 0.3)). Models that reject the parameter (e.g., GPT-5 series, o-series, and current Claude models) are automatically retried once without it, so any model works without per-model configuration.

See examples/llm/ for complete scripts, including a spaCy-vs-LLM NER comparison.

Advanced Usage

Thread Safety

All spaCy calls must be made from the same thread that initialized PyCall (normally the main thread). Calling the spaCy pipeline from another thread — e.g. nlp.read(text) inside a Thread.new block, a Rails multi-threaded server, or a Sidekiq worker — will hang the entire process.

Note that attribute access (such as nlp.pipe_names) works from other threads, so the failure mode is not obvious: the process only freezes when the pipeline actually runs. The root cause is currently unknown (it is specific to spaCy pipeline execution; other GIL-releasing Python calls work fine from other threads).

If you need concurrent processing, serialize all Python calls onto a single dedicated thread, for example with a worker thread and a queue.

Setting a Timeout

You can set a timeout for the Spacy::Language.new method:

nlp=Spacy::Language.new("en_core_web_sm",timeout: 120)# Set timeout to 120 seconds

If the model does not finish loading within the given seconds, a RuntimeError is raised. Pass timeout: nil to wait indefinitely.

Document Serialization

You can serialize processed documents to binary format for caching or storage. This is useful when you want to avoid re-processing the same text multiple times.

Saving a document:

require"ruby-spacy"nlp=Spacy::Language.new("en_core_web_sm")doc=nlp.read("Apple Inc. was founded by Steve Jobs in California.")# Serialize to binarybytes=doc.to_bytes# Save to fileFile.binwrite("doc_cache.bin",bytes)

Restoring a document:

nlp=Spacy::Language.new("en_core_web_sm")# Load from filebytes=File.binread("doc_cache.bin")# Restore the document (all annotations are preserved)restored_doc=Spacy::Doc.from_bytes(nlp,bytes)putsrestored_doc.text# => "Apple Inc. was founded by Steve Jobs in California."restored_doc.ents.eachdo |ent|
puts"#{ent.text} (#{ent.label})"end# => Apple Inc. (ORG)# => Steve Jobs (PERSON)# => California (GPE)

Author

Yoichiro Hasebe [yohasebe@gmail.com]

Acknowlegments

I would like to thank the following open source projects and their creators for making this project possible:

• explosion/spaCy
• mrkn/pycall.rb

License

This library is available as open source under the terms of the MIT License.