zuno 1.0.1 → 1.0.2

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 7d6d4d8dd97c7ac8743ddc66545aeda41ed824357965c693ac244fcf46de5df2
-  data.tar.gz: d7e0a1a6029482af54dbf7bc6e5196b39dd60e424b1fb18b4003543cfbe79867
+  metadata.gz: 375f446f6d41ed1c8f05361915316c3da7fbc571ee5ec4d59316739f9fd6a14f
+  data.tar.gz: 641e201f64dc3076185e2a6593932e67847f05ce4c5a66866b200b8db4d38241
 SHA512:
-  metadata.gz: 4862131a38d657f175bbc488599d5cbcd97899b43bdac7a26117a437c081bfdaa365e4a2e5255e7e48464aa5e5d129e14106364727b19bae4b8166ff779cd621
-  data.tar.gz: 96340e7d0b4a2153d58b328cbdcbec5836ddd76724941fd8efa39343de14c052a51c2031e916562e0b66de5044832a429f2332c92e4325f40d6eefc2d1e67454
+  metadata.gz: 76d4be87dad8f35fc9e879a14705035344823f43c0ba45c6d8317d5b8e0a273c6349db11294b8e17c5dfb901dec9006fa488b8ebe8633baec33aa66e74be9166
+  data.tar.gz: 8555df550de0513330c086ec4b676d584a1908feb983d2ca59d32c8d685f626d2c767960e6e1dde23dbb17ac0cfc903aee251bef71519ffde27f8c1524cd2d68

data/CHANGELOG.md

@@ -0,0 +1,7 @@
+# Changelog
+
+## [1.0.2](https://github.com/dqnamo/zuno/compare/v1.0.1...v1.0.2) (2026-04-01)
+
+### Features
+
+* Add ElevenLabs Scribe batch transcription via `Zuno.transcribe` and `Zuno.elevenlabs` ([ElevenLabs speech-to-text API](https://elevenlabs.io/docs/api-reference/speech-to-text/convert)).

data/lib/zuno/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Zuno
-  VERSION = "1.0.1"
+  VERSION = "1.0.2"
 end

data/lib/zuno.rb

@@ -78,6 +78,7 @@ module Zuno
   OPENROUTER_ADAPTER_CONFIG_KEYS = %i[api_key app_url title timeout].freeze
   AI_GATEWAY_ADAPTER_CONFIG_KEYS = %i[api_key timeout base_url].freeze
   REPLICATE_ADAPTER_CONFIG_KEYS = %i[api_key timeout].freeze
+  ELEVENLABS_ADAPTER_CONFIG_KEYS = %i[api_key timeout base_url].freeze
   DEFAULT_MAX_ITERATIONS = 1
   REPLICATE_PREFER_WAIT_SECONDS = 60
   REPLICATE_POLL_INTERVAL_SECONDS = 1
@@ -126,6 +127,18 @@ module Zuno
     )
   end
 
+  def elevenlabs(
+    api_key: nil,
+    timeout: Providers::ElevenLabs::DEFAULT_TIMEOUT,
+    base_url: Providers::ElevenLabs::DEFAULT_BASE_URL
+  )
+    Providers::ElevenLabs.new(
+      api_key: api_key,
+      timeout: timeout,
+      base_url: base_url
+    )
+  end
+
   def tool(name:, description:, input_schema:, &execute)
     raise ToolError, "A block is required for tool '#{name}'" unless block_given?
 
@@ -137,6 +150,46 @@ module Zuno
     )
   end
 
+  def transcribe(
+    model_id: "scribe_v2",
+    file: nil,
+    cloud_storage_url: nil,
+    source_url: nil,
+    provider_options: {},
+    **options
+  )
+    validate_transcription_input!(
+      model_id: model_id,
+      file: file,
+      cloud_storage_url: cloud_storage_url,
+      source_url: source_url
+    )
+    resolved_provider_options = merge_provider_options({}, provider_options)
+    adapter = provider_adapter(:elevenlabs, resolved_provider_options)
+    response = adapter.transcribe(
+      model_id: model_id,
+      file: file,
+      cloud_storage_url: cloud_storage_url,
+      source_url: source_url,
+      options: options
+    )
+
+    result = {
+      text: response["text"],
+      language_code: response["language_code"],
+      language_probability: response["language_probability"],
+      words: response["words"],
+      transcripts: response["transcripts"],
+      transcription_id: response["transcription_id"],
+      raw_response: response
+    }
+    result.reject { |_key, value| value.nil? }
+  rescue ProviderError
+    raise
+  rescue StandardError => e
+    raise Error, e.message
+  end
+
   def generate(
     model:,
     messages: nil,
@@ -763,6 +816,25 @@ module Zuno
   end
   private_class_method :validate_no_webhook_support!
 
+  def validate_transcription_input!(model_id:, file:, cloud_storage_url:, source_url:)
+    model_id_value = model_id.to_s.strip
+    raise Error, "model_id is required" if model_id_value.empty?
+
+    inputs = []
+    inputs << :file unless file.nil?
+    inputs << :cloud_storage_url unless cloud_storage_url.nil? || cloud_storage_url.to_s.strip.empty?
+    inputs << :source_url unless source_url.nil? || source_url.to_s.strip.empty?
+
+    if inputs.empty?
+      raise Error, "transcribe requires one input: file, cloud_storage_url, or source_url"
+    end
+
+    return if inputs.length == 1
+
+    raise Error, "transcribe accepts exactly one input: file, cloud_storage_url, or source_url"
+  end
+  private_class_method :validate_transcription_input!
+
   def normalize_tools(tools)
     return {} if tools.nil?
 
@@ -840,6 +912,8 @@ module Zuno
       AI_GATEWAY_ADAPTER_CONFIG_KEYS
     when :replicate
       REPLICATE_ADAPTER_CONFIG_KEYS
+    when :elevenlabs
+      ELEVENLABS_ADAPTER_CONFIG_KEYS
     else
       []
     end
@@ -866,6 +940,9 @@ module Zuno
     when :replicate
       config = pick_keys(provider_options, REPLICATE_ADAPTER_CONFIG_KEYS)
       Providers::Replicate.new(**config)
+    when :elevenlabs
+      config = pick_keys(provider_options, ELEVENLABS_ADAPTER_CONFIG_KEYS)
+      Providers::ElevenLabs.new(**config)
     else
       raise ProviderError, "Unsupported provider: #{provider}"
     end
@@ -1506,6 +1583,106 @@ module Zuno
         raise ProviderError, "Replicate request failed: #{response.return_code}#{suffix}"
       end
     end
+
+    class ElevenLabs
+      DEFAULT_BASE_URL = "https://api.elevenlabs.io".freeze
+      SPEECH_TO_TEXT_PATH = "/v1/speech-to-text".freeze
+      DEFAULT_TIMEOUT = 120_000
+
+      def initialize(api_key: nil, timeout: DEFAULT_TIMEOUT, base_url: DEFAULT_BASE_URL)
+        @api_key = api_key
+        raise ProviderError, "ElevenLabs API key not configured" if @api_key.nil? || @api_key.to_s.empty?
+
+        @timeout = timeout
+        @base_url = base_url.to_s.empty? ? DEFAULT_BASE_URL : base_url.to_s
+      end
+
+      def transcribe(model_id:, file:, cloud_storage_url:, source_url:, options:)
+        opened_file = nil
+        body = build_body(
+          model_id: model_id,
+          file: file,
+          cloud_storage_url: cloud_storage_url,
+          source_url: source_url,
+          options: options
+        ) do |candidate|
+          opened_file = candidate
+        end
+
+        response = Typhoeus.post(
+          speech_to_text_url,
+          headers: headers,
+          body: body,
+          multipart: true,
+          timeout: @timeout
+        )
+
+        validate_response!(response)
+        parsed = JSON.parse(response.body)
+        raise ProviderError, "ElevenLabs returned invalid JSON" unless parsed.is_a?(Hash)
+
+        parsed
+      rescue JSON::ParserError => e
+        raise ProviderError, "Failed to parse ElevenLabs response: #{e.message}"
+      ensure
+        opened_file.close if opened_file.is_a?(File) && !opened_file.closed?
+      end
+
+      private
+
+      def speech_to_text_url
+        "#{@base_url}#{SPEECH_TO_TEXT_PATH}"
+      end
+
+      def headers
+        { "xi-api-key" => @api_key }
+      end
+
+      def build_body(model_id:, file:, cloud_storage_url:, source_url:, options:)
+        body = {
+          "model_id" => model_id.to_s
+        }
+
+        normalized_options = options.is_a?(Hash) ? options : {}
+        normalized_options.each do |key, value|
+          next if value.nil?
+
+          body[key.to_s] = value
+        end
+
+        if file.is_a?(String)
+          opened_file = File.open(file, "rb")
+          body["file"] = opened_file
+          yield(opened_file) if block_given?
+        elsif !file.nil?
+          body["file"] = file
+        end
+
+        body["cloud_storage_url"] = cloud_storage_url.to_s unless cloud_storage_url.nil? || cloud_storage_url.to_s.strip.empty?
+        body["source_url"] = source_url.to_s unless source_url.nil? || source_url.to_s.strip.empty?
+        body
+      rescue Errno::ENOENT => e
+        raise ProviderError, "Failed to open transcription file: #{e.message}"
+      end
+
+      def validate_response!(response)
+        raise ProviderError, "No response returned from ElevenLabs" if response.nil?
+        raise ProviderError, "ElevenLabs request timed out" if response.timed_out?
+
+        status = response.code.to_i
+        body = response.body.to_s
+        message = body.length > 300 ? "#{body[0, 300]}..." : body
+
+        return if status >= 200 && status < 300
+
+        if status.positive?
+          raise ProviderError, "ElevenLabs responded with HTTP #{status}: #{message}"
+        end
+
+        suffix = message.empty? ? "" : ": #{message}"
+        raise ProviderError, "ElevenLabs request failed: #{response.return_code}#{suffix}"
+      end
+    end
   end
 
   class SseParser

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: zuno
 version: !ruby/object:Gem::Version
-  version: 1.0.1
+  version: 1.0.2
 platform: ruby
 authors:
 - Hyperaide
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2026-03-31 00:00:00.000000000 Z
+date: 2026-04-01 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: typhoeus
@@ -53,13 +53,14 @@ dependencies:
       - !ruby/object:Gem::Version
         version: '3.13'
 description: Standalone Ruby SDK for AI generation across OpenRouter and Replicate,
-  with iterative tool loops and SSE streaming.
+  ElevenLabs speech-to-text, with iterative tool loops and SSE streaming.
 email:
 - team@hyperaide.dev
 executables: []
 extensions: []
 extra_rdoc_files: []
 files:
+- CHANGELOG.md
 - README.md
 - lib/zuno.rb
 - lib/zuno/version.rb