gliner 0.1.1 → 0.2.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 40c13e552a887bd1d77fc718be8708b228c9e213c9bb49c3591b113738346bd4
-  data.tar.gz: e6dce5cda3fcef66bb9a627ffe23c5324b9395af6bd5eeb3c45a4ea400f52c63
+  metadata.gz: 749c6a152abe40b2735e56724d01bcb9df57df3ed883f7b7a7a735c7d4dbe613
+  data.tar.gz: dd8add045f98be4876bea92faadd68b2da4a64613502c710c3a6769d224016bd
 SHA512:
-  metadata.gz: 723f039e065cd9b86111de3d9478cf711d01efdea5a67ce426cf1bcfb4e84396abe9f9969150b743a8f0e54be5caea0401d25a1dcf191c0f09a932189f19fbed
-  data.tar.gz: 1566d0fc8446ee1463c0dfbeb37d7eb7021ca0f0d04bdf3d83148f1e7b33eb1d2eea4ddea44dcef91bf9b11f7cd2508cdcfd9120ade6fb72ac70aa5510d22896
+  metadata.gz: 996bd63e35f9302f1a01a78d92d39ca47013deb42e58ff3c332ac260c2351f172d5a1897bdb3f673a4a1cf1f63c2ef1b1a7c04c2025ff94be6eb71548c79bf97
+  data.tar.gz: 246d3939daa2ca82df524ae6a89dbd734792d5c6952857d517634f50cc59176f26fd1bef15677b655b1c2f9cb1b35c5cb700bdbbc764d6949b6b941b7c83e77f

data/README.md

@@ -1,4 +1,6 @@
-# Gliner
+# GLiNER
+[![tests](https://github.com/elcuervo/gliner/actions/workflows/tests.yml/badge.svg)](https://github.com/elcuervo/gliner/actions/workflows/tests.yml)
+![Gem Version](https://img.shields.io/gem/v/gliner)
 
 ![](https://images.unsplash.com/photo-1625768376503-68d2495d78c5?q=80&w=2225&auto=format&fit=crop&ixlib=rb-4.1.0&ixid=M3wxMjA3fDB8MHxwaG90by1wYWdlfHx8fGVufDB8fHx8fA%3D%3D)
 
@@ -17,53 +19,47 @@ require "gliner"
 
 Gliner.configure do |config|
   config.threshold = 0.2
-  config.model_dir = "/path/to/gliner2-multi-v1"
-  config.model_file = "model.onnx"
+  # By default, the gem downloads the default model to .cache/
+  # Or set a local path explicitly:
+  # config.model = "/path/to/gliner2-multi-v1"
+  config.variant = :fp16
 end
 
-Gliner.load("path/to/gliner2-multi-v1")
-
 text = "Apple CEO Tim Cook announced iPhone 15 in Cupertino yesterday."
 labels = ["company", "person", "product", "location"]
 
 model = Gliner[labels]
 pp model[text]
-```
-
-Expected shape:
 
-```ruby
-{"entities"=>{"company"=>["Apple"], "person"=>["Tim Cook"], "product"=>["iPhone 15"], "location"=>["Cupertino"]}}
+# => {"company"=>["Apple"], "person"=>["Tim Cook"], "product"=>["iPhone 15"], "location"=>["Cupertino"]}
 ```
 
 You can also pass per-entity configs:
 
 ```ruby
 labels = {
-  "email" => { "description" => "Email addresses", "dtype" => "list", "threshold" => 0.9 },
-  "person" => { "description" => "Person names", "dtype" => "str" }
+  email: { description: "Email addresses", dtype: "list", threshold: 0.9 },
+  person: { description: "Person names", dtype: "str" }
 }
 
 model = Gliner[labels]
 pp model["Email John Doe at john@example.com.", threshold: 0.5]
+
+# => {"email"=>["john@example.com"], "person"=>"John Doe"}
 ```
 
 ### Classification
 
 ```ruby
 model = Gliner.classify[
-  { "sentiment" => %w[positive negative neutral] }
+  { sentiment: %w[positive negative neutral] }
 ]
 
 result = model["This laptop has amazing performance but terrible battery life!"]
 
 pp result
-```
 
-Expected shape:
-
-```ruby
-{"sentiment"=>"negative"}
+# => {"sentiment"=>"negative"}
 ```
 
 ### Structured extraction
@@ -72,7 +68,7 @@ Expected shape:
 text = "iPhone 15 Pro Max with 256GB storage, A17 Pro chip, priced at $1199."
 
 structure = {
-  "product" => [
+  product: [
     "name::str::Full product name and model",
     "storage::str::Storage capacity",
     "processor::str::Chip or processor information",
@@ -83,18 +79,16 @@ structure = {
 result = Gliner[structure][text]
 
 pp result
-```
 
-Expected shape:
-
-```ruby
-{"product"=>[{"name"=>"iPhone 15 Pro Max", "storage"=>"256GB", "processor"=>"A17 Pro chip", "price"=>"$1199"}]}
+# => {"product"=>[{"name"=>"iPhone 15 Pro Max", "storage"=>"256GB", "processor"=>"A17 Pro", "price"=>"1199"}]}
 ```
 
 Choices can be included in field specs:
 
 ```ruby
-result = Gliner[{ "order" => ["status::[pending|processing|shipped]::str"] }]["Status: shipped"]
+result = Gliner[{ order: ["status::[pending|processing|shipped]::str"] }]["Status: shipped"]
+
+# => {"order"=>[{"status"=>"shipped"}]}
 ```
 
 ## Model files
@@ -102,18 +96,19 @@ result = Gliner[{ "order" => ["status::[pending|processing|shipped]::str"] }]["S
 This implementation expects a directory containing:
 
 - `tokenizer.json`
-- `model.onnx` or `model_int8.onnx`
+- `model.onnx`, `model_fp16.onnx`, or `model_int8.onnx`
 - (optional) `config.json` with `max_width` and `max_seq_len`
 
 One publicly available ONNX export is `cuerbot/gliner2-multi-v1` on Hugging Face.
-By default, `model_int8.onnx` is used; set `config.model_file` or `GLINER_MODEL_FILE` to override.
+By default, `model_fp16.onnx` is used; set `config.variant` (or `GLINER_MODEL_FILE`) to override.
+Variants map to files as: `:fp16` → `model_fp16.onnx`, `:fp32` → `model.onnx`, `:int8` → `model_int8.onnx`.
 
-You can also configure the model directory in code:
+You can also configure the model source directly:
 
 ```ruby
 Gliner.configure do |config|
-  config.model_dir = "/path/to/model_dir"
-  config.model_file = "model_int8.onnx"
+  config.model = "/path/to/model_dir"
+  config.variant = :int8
 end
 ```
 
@@ -150,7 +145,7 @@ If you omit `MODEL_DIR`, the console auto-downloads a public test model (configu
 ```bash
 rake console
 # or:
-GLINER_REPO_ID=cuerbot/gliner2-multi-v1 GLINER_MODEL_FILE=model_int8.onnx rake console
+GLINER_REPO_ID=cuerbot/gliner2-multi-v1 GLINER_MODEL_FILE=model_fp16.onnx rake console
 ```
 
 Or:

data/bin/console

@@ -1,4 +1,5 @@
 #!/usr/bin/env ruby
+
 # frozen_string_literal: true
 
 begin
@@ -7,50 +8,20 @@ rescue LoadError
 end
 
 require "gliner"
-require "fileutils"
-require "httpx"
 require "irb"
 
-DEFAULT_REPO_ID = "cuerbot/gliner2-multi-v1"
-DEFAULT_MODEL_FILE = "model_int8.onnx"
-
-def ensure_model_dir!(repo_id:, model_file:)
-  dir = File.expand_path("../tmp/models/#{repo_id.tr('/', '__')}", __dir__)
-  FileUtils.mkdir_p(dir)
-
-  base = "https://huggingface.co/#{repo_id}/resolve/main"
-  files = ["tokenizer.json", "config.json", model_file]
-
-  files.each do |file|
-    dest = File.join(dir, file)
-    next if File.exist?(dest) && File.size?(dest)
-    download("#{base}/#{file}", dest)
-  end
-
-  dir
-end
-
-def download(url, dest)
-  response = HTTPX.get(url)
-  raise "Download failed: #{url} (status: #{response.status})" unless response.status.between?(200, 299)
-
-  File.binwrite(dest, response.body.to_s)
-end
-
 model_dir = ARGV[0] || ENV["GLINER_MODEL_DIR"]
-repo_id = ENV["GLINER_REPO_ID"] || DEFAULT_REPO_ID
-model_file = ENV["GLINER_MODEL_FILE"] || DEFAULT_MODEL_FILE
+model_file = ENV["GLINER_MODEL_FILE"]
 
 if model_dir && !model_dir.empty?
-  $gliner_model = Gliner.load(model_dir, file: model_file)
+  $gliner_model = model_file ? Gliner.load(model_dir, file: model_file) : Gliner.load(model_dir)
 else
   begin
-    require "fileutils"
-    model_dir = ensure_model_dir!(repo_id: repo_id, model_file: model_file)
-    $gliner_model = Gliner.load(model_dir, file: model_file)
+    Gliner.configure { |config| config.auto = true }
+    $gliner_model = Gliner.model
   rescue => e
     warn "No model loaded (auto-download failed: #{e.class}: #{e.message})"
-    warn "Set GLINER_MODEL_DIR to a local model dir, or set GLINER_REPO_ID/GLINER_MODEL_FILE for auto-download."
+    warn "Set GLINER_MODEL_DIR or configure Gliner.config.model to a local model dir."
   end
 end
 
@@ -76,6 +47,5 @@ puts "- helper: gliner_classify(text, tasks)"
 puts "- helper: gliner_extract_json(text, structures)"
 puts "- model variable: $gliner_model"
 puts "- model dir: #{model_dir.inspect}"
-puts "- auto-download env: GLINER_REPO_ID=#{repo_id.inspect} GLINER_MODEL_FILE=#{model_file.inspect}" unless $gliner_model
 
 IRB.start(__FILE__)

data/gliner.gemspec

@@ -16,11 +16,14 @@ Gem::Specification.new do |spec|
   spec.files = Dir.glob('lib/**/*') + Dir.glob('bin/*') + %w[README.md LICENSE gliner.gemspec]
   spec.require_paths = ['lib']
 
+  spec.add_dependency 'httpx', '~> 1.0'
   spec.add_dependency 'onnxruntime', '~> 0.10'
   spec.add_dependency 'tokenizers', '~> 0.6'
 
-  spec.add_development_dependency 'httpx', '~> 1.0'
   spec.add_development_dependency 'rake', '~> 13.0'
+  spec.add_development_dependency 'irb', '~> 1.16.0'
   spec.add_development_dependency 'rspec', '~> 3.13'
   spec.add_development_dependency 'rubocop', '~> 1.50'
+
+  spec.metadata['rubygems_mfa_required'] = 'true'
 end

data/lib/gliner/classifier.rb

@@ -5,6 +5,7 @@ require 'gliner/position_iteration'
 module Gliner
   class Classifier
     include PositionIteration
+
     def initialize(inference, max_width:)
       @inference = inference
       @max_width = max_width
@@ -63,6 +64,5 @@ module Gliner
     def format_label(label, score, include_confidence)
       include_confidence ? { 'label' => label, 'confidence' => score } : label
     end
-
   end
 end

data/lib/gliner/configuration.rb

@@ -4,12 +4,30 @@ module Gliner
   class Configuration
     DEFAULT_THRESHOLD = 0.5
 
-    attr_accessor :threshold, :model_dir, :model_file
+    attr_accessor :threshold, :model
+    attr_reader :variant
 
     def initialize
       @threshold = DEFAULT_THRESHOLD
-      @model_dir = nil
-      @model_file = nil
+      @model = nil
+      @variant = :fp16
+      @auto = true
+    end
+
+    def variant=(value)
+      @variant = value&.to_sym
+    end
+
+    def auto!(value = true)
+      @auto = !!value
+    end
+
+    def auto=(value)
+      @auto = !!value
+    end
+
+    def auto?
+      @auto
     end
   end
 end

data/lib/gliner/model.rb

@@ -22,7 +22,7 @@ module Gliner
     DEFAULT_MAX_WIDTH = 8
     DEFAULT_MAX_SEQ_LEN = 512
 
-    def self.from_dir(dir, file: 'model_int8.onnx')
+    def self.from_dir(dir, file: 'model_fp16.onnx')
       config_path = File.join(dir, 'config.json')
       config = File.exist?(config_path) ? JSON.parse(File.read(config_path)) : {}
 

data/lib/gliner/runners/classification_runner.rb

@@ -3,6 +3,8 @@
 module Gliner
   module Runners
     class ClassificationRunner
+      include Inspectable
+
       def self.[](tasks)
         new(Gliner.model!, tasks)
       end
@@ -21,6 +23,11 @@ module Gliner
       end
 
       alias call []
+
+      private
+
+      def inspect_label = 'Classification'
+      def inspect_items = @tasks.keys
     end
   end
 end

data/lib/gliner/runners/entity_runner.rb

@@ -3,8 +3,12 @@
 module Gliner
   module Runners
     class EntityRunner
+      include Inspectable
+
       def initialize(model, config)
         parsed = model.entity_task.parse_config(config)
+
+        @labels = parsed[:labels]
         @task = PreparedTask.new(model.entity_task, parsed)
       end
 
@@ -14,6 +18,11 @@ module Gliner
       end
 
       alias call []
+
+      private
+
+      def inspect_label = 'Entity'
+      def inspect_items = @labels
     end
   end
 end

data/lib/gliner/runners/inspectable.rb

@@ -0,0 +1,13 @@
+# frozen_string_literal: true
+
+module Gliner
+  module Runners
+    module Inspectable
+      def inspect
+        items = Array(inspect_items).map(&:to_s)
+
+        "#<Gliner(#{inspect_label}) input=#{items.inspect}>"
+      end
+    end
+  end
+end

data/lib/gliner/runners/structured_runner.rb

@@ -3,6 +3,8 @@
 module Gliner
   module Runners
     class StructuredRunner
+      include Inspectable
+
       def initialize(model, config)
         @tasks = build_tasks(model, config)
       end
@@ -17,6 +19,9 @@ module Gliner
 
       private
 
+      def inspect_label = 'Structure'
+      def inspect_items = @tasks.keys
+
       def build_tasks(model, config)
         raise Error, 'structures must be a Hash' unless config.is_a?(Hash)
 

data/lib/gliner/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Gliner
-  VERSION = '0.1.1'
+  VERSION = '0.2.1'
 end

data/lib/gliner.rb

@@ -1,14 +1,22 @@
 # frozen_string_literal: true
 
+require 'fileutils'
+require 'httpx'
 require 'gliner/version'
 require 'gliner/configuration'
 require 'gliner/model'
 require 'gliner/runners/prepared_task'
+require 'gliner/runners/inspectable'
 require 'gliner/runners/entity_runner'
 require 'gliner/runners/structured_runner'
 require 'gliner/runners/classification_runner'
 
 module Gliner
+  HF_REPO = 'cuerbot/gliner2-multi-v1'
+  HF_DIR = 'onnx'
+
+  DEFAULT_MODEL_BASE = "https://huggingface.co/#{HF_REPO}/resolve/main/#{HF_DIR}".freeze
+
   Error = Class.new(StandardError)
 
   PreparedInput = Data.define(
@@ -41,11 +49,13 @@ module Gliner
   end
 
   class << self
-    attr_writer :model
-    attr_writer :config
+    attr_writer :model, :config
 
     def configure
       yield(config)
+
+      reset_model!
+      apply_model_source!
     end
 
     def config
@@ -53,53 +63,51 @@ module Gliner
     end
 
     def load(dir, file: nil)
-      file ||= ENV['GLINER_MODEL_FILE']
-      file ||= config.model_file
+      file ||= ENV['GLINER_MODEL_FILE'] || model_file_for_variant(config.variant)
 
-      self.model = Model.from_dir(dir, file: file || 'model_int8.onnx')
+      self.model = Model.from_dir(dir, file: file)
     end
 
     def model
-      @model ||= model_from_config || model_from_env
-    end
-
-    def model!
-      fetch_model!
+      @model ||= begin
+        apply_model_source!
+        model_from_config || model_from_env
+      end
     end
 
     def [](config)
-      runner_for(config).new(fetch_model!, config)
+      runner_for(config).new(model!, config)
     end
 
     def classify
       Runners::ClassificationRunner
     end
 
+    def model!
+      model = self.model
+
+      return model if model
+
+      raise Error, 'No model loaded. Call Gliner.load("/path/to/model"), set config.model, or set GLINER_MODEL_DIR.'
+    end
+
     private
 
     def model_from_config
-      dir = config.model_dir
-      return nil if dir.nil? || dir.empty?
-
-      file = config.model_file
-      return Model.from_dir(dir) if file.nil? || file.empty?
+      source = config.model
+      return nil if source.nil?
 
-      Model.from_dir(dir, file: file)
+      file = model_file_for_variant(config.variant)
+      Model.from_dir(source, file: file)
     end
 
     def model_from_env
-      dir = ENV.fetch('GLINER_MODEL_DIR', nil)
-      return nil if dir.nil? || dir.empty?
+      dir = env_model_dir
+      return if dir.nil?
 
-      file = ENV['GLINER_MODEL_FILE'] || 'model_int8.onnx'
-      Model.from_dir(dir, file: file)
-    end
-
-    def fetch_model!
-      model = self.model
-      return model if model
+      file = ENV['GLINER_MODEL_FILE'] || model_file_for_variant(config.variant)
 
-      raise Error, 'No model loaded. Call Gliner.load("/path/to/model"), set config.model_dir, or set GLINER_MODEL_DIR.'
+      Model.from_dir(dir, file: file)
     end
 
     def runner_for(config)
@@ -112,9 +120,71 @@ module Gliner
       return false unless config.is_a?(Hash)
 
       keys = config.transform_keys(&:to_s)
+
       return true if keys.key?('name') && keys.key?('fields')
 
       config.values.all? { |value| value.is_a?(Array) }
     end
+
+    def reset_model!
+      @model = nil
+    end
+
+    def apply_model_source!
+      return unless config.auto?
+
+      source = config.model
+      return unless source.nil?
+      return if env_model_dir
+
+      config.model = download_default_model
+    end
+
+    def download_default_model
+      model_file = model_file_for_variant(config.variant)
+      root = File.expand_path('..', __dir__)
+      dir = File.join(root, '.cache', 'models', HF_REPO.tr('/', '__'))
+
+      FileUtils.mkdir_p(dir)
+
+      files = ['tokenizer.json', 'config.json', model_file]
+      client = HTTPX.plugin(:follow_redirects).with(max_redirects: 5)
+
+      files.each do |file|
+        dest = File.join(dir, file)
+        next if File.exist?(dest) && File.size?(dest)
+        download_file!(client, "#{DEFAULT_MODEL_BASE}/#{file}", dest)
+      end
+
+      dir
+    end
+
+    def download_file!(client, url, dest)
+      response = client.get(url)
+      status = response.status
+
+      unless status && status.between?(200, 299)
+        raise Error, "Download failed: #{url} (status: #{status || 'unknown'})"
+      end
+
+      File.binwrite(dest, response.body.to_s)
+    end
+
+    def env_model_dir
+      dir = ENV.fetch('GLINER_MODEL_DIR', nil)
+      return nil if dir.nil? || dir.empty?
+
+      dir
+    end
+
+    def model_file_for_variant(variant = :fp16)
+      case variant.to_sym
+      when :fp16 then 'model_fp16.onnx'
+      when :fp32 then 'model.onnx'
+      when :int8 then 'model_int8.onnx'
+      else
+        raise Error, "Unknown model variant: #{variant.inspect}"
+      end
+    end
   end
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: gliner
 version: !ruby/object:Gem::Version
-  version: 0.1.1
+  version: 0.2.1
 platform: ruby
 authors:
 - elcuervo
@@ -9,6 +9,20 @@ bindir: bin
 cert_chain: []
 date: 1980-01-01 00:00:00.000000000 Z
 dependencies:
+- !ruby/object:Gem::Dependency
+  name: httpx
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: '1.0'
+  type: :runtime
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: '1.0'
 - !ruby/object:Gem::Dependency
   name: onnxruntime
   requirement: !ruby/object:Gem::Requirement
@@ -38,33 +52,33 @@ dependencies:
       - !ruby/object:Gem::Version
         version: '0.6'
 - !ruby/object:Gem::Dependency
-  name: httpx
+  name: rake
   requirement: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: '1.0'
+        version: '13.0'
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: '1.0'
+        version: '13.0'
 - !ruby/object:Gem::Dependency
-  name: rake
+  name: irb
   requirement: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: '13.0'
+        version: 1.16.0
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: '13.0'
+        version: 1.16.0
 - !ruby/object:Gem::Dependency
   name: rspec
   requirement: !ruby/object:Gem::Requirement
@@ -117,6 +131,7 @@ files:
 - lib/gliner/position_iteration.rb
 - lib/gliner/runners/classification_runner.rb
 - lib/gliner/runners/entity_runner.rb
+- lib/gliner/runners/inspectable.rb
 - lib/gliner/runners/prepared_task.rb
 - lib/gliner/runners/structured_runner.rb
 - lib/gliner/span_extractor.rb
@@ -130,7 +145,8 @@ files:
 homepage: https://github.com/elcuervo/gliner
 licenses:
 - MIT
-metadata: {}
+metadata:
+  rubygems_mfa_required: 'true'
 rdoc_options: []
 require_paths:
 - lib
@@ -145,7 +161,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.7.2
+rubygems_version: 3.6.9
 specification_version: 4
 summary: Schema-based information extraction (GLiNER2) via ONNX Runtime
 test_files: []