Module: Zuno

Defined in:
lib/zuno.rb,
lib/zuno/version.rb

Defined Under Namespace

Modules: Providers Classes: CallbackContext, CallbackControl, Error, MaxIterationsExceeded, ModelDescriptor, ProviderError, SseParser, StreamingError, ToolDefinition, ToolError

Constant Summary collapse

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
REPLICATE_WAIT_TIMEOUT_SECONDS = 
600
REPLICATE_TERMINAL_STATUSES = 
%w[succeeded failed canceled aborted].freeze
VERSION = 
"1.1.3"

Class Method Summary collapse

Class Method Details

.ai_gateway(api_key: nil, timeout: Providers::AIGateway::DEFAULT_TIMEOUT, base_url: Providers::AIGateway::DEFAULT_BASE_URL) ⇒ Object 



205
206
207
208
209
210
211
212
213
214
215
 
# File 'lib/zuno.rb', line 205

def ai_gateway(
  api_key: nil,
  timeout: Providers::AIGateway::DEFAULT_TIMEOUT,
  base_url: Providers::AIGateway::DEFAULT_BASE_URL
)
  Providers::AIGateway.new(
    api_key: api_key,
    timeout: timeout,
    base_url: base_url
  )
end

.default_provider_options ⇒ Object 



177
178
179
 
# File 'lib/zuno.rb', line 177

def default_provider_options
  @default_provider_options ||= {}
end

.default_provider_options=(options) ⇒ Object 



181
182
183
 
# File 'lib/zuno.rb', line 181

def default_provider_options=(options)
  @default_provider_options = options.is_a?(Hash) ? options : {}
end

.elevenlabs(api_key: nil, timeout: Providers::ElevenLabs::DEFAULT_TIMEOUT, base_url: Providers::ElevenLabs::DEFAULT_BASE_URL) ⇒ Object 



217
218
219
220
221
222
223
224
225
226
227
 
# File 'lib/zuno.rb', line 217

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

.embed(model:, content:, dimensions: nil, encoding_format: nil, provider_options: {}, **options) ⇒ Object 



280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
 
# File 'lib/zuno.rb', line 280

def embed(
  model:,
  content:,
  dimensions: nil,
  encoding_format: nil,
  provider_options: {},
  **options
)
  validate_embedding_input!(content: content)

  model_descriptor = normalize_model(model)
  resolved_provider_options = merge_provider_options(
    model_descriptor.provider_options,
    provider_options
  )
  provider = model_descriptor.provider.to_sym

  case provider
  when :openrouter, :ai_gateway
    adapter = provider_adapter(provider, resolved_provider_options)
    response = adapter.embed(
      build_embedding_payload(
        model_id: model_descriptor.id,
        provider: provider,
        content: content,
        dimensions: dimensions,
        encoding_format: encoding_format,
        provider_options: resolved_provider_options,
        options: options
      )
    )

    embeddings = Array(response["data"]).filter_map do |item|
      item["embedding"] if item.is_a?(Hash)
    end
    result = {
      embeddings: embeddings,
      usage: response["usage"],
      model: response["model"],
      raw_response: response
    }
    result[:embedding] = embeddings.first if embeddings.length == 1
    result
  else
    raise ProviderError, "embed only supports openrouter or ai_gateway provider"
  end
rescue ProviderError
  raise
rescue StandardError => e
  raise Error, e.message
end

.generate(model:, messages: nil, system: nil, prompt: nil, input: nil, tools: {}, tool_choice: nil, temperature: nil, max_tokens: nil, provider_options: {}, before_tool_execution: nil, after_tool_execution: nil, before_generation: nil, after_generation: nil) ⇒ Object 



332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
 
# File 'lib/zuno.rb', line 332

def generate(
  model:,
  messages: nil,
  system: nil,
  prompt: nil,
  input: nil,
  tools: {},
  tool_choice: nil,
  temperature: nil,
  max_tokens: nil,
  provider_options: {},
  before_tool_execution: nil,
  after_tool_execution: nil,
  before_generation: nil,
  after_generation: nil
)
  callback_control = nil
  after_generation_called = false
  callback_control = CallbackControl.new

  model_descriptor = normalize_model(model)
  resolved_provider_options = merge_provider_options(
    model_descriptor.provider_options,
    provider_options
  )
  provider = model_descriptor.provider.to_sym

  call_callback!(
    before_generation,
    {
      model: model_descriptor,
      mode: "single",
      provider: provider
    },
    callback_control
  )

  if callback_control.stopped?
    result = callback_stopped_result(
      control: callback_control,
      iterations: [],
      message: {},
      usage: nil,
      raw_response: nil
    )
    after_generation_called = true
    call_callback!(after_generation, { ok: true, result: result }, callback_control)
    return result
  end

  result =
    case provider
    when :openrouter, :ai_gateway
      adapter = provider_adapter(provider, resolved_provider_options)
      generate_openrouter_single(
        model_descriptor: model_descriptor,
        adapter: adapter,
        messages: messages,
        system: system,
        prompt: prompt,
        tools: tools,
        tool_choice: tool_choice,
        temperature: temperature,
        max_tokens: max_tokens,
        provider_options: resolved_provider_options,
        before_tool_execution: before_tool_execution,
        after_tool_execution: after_tool_execution
      )
    when :replicate
      raise Error, "tools are not supported for replicate generate" unless normalize_tools(tools).empty?
      raise Error, "tool_choice is not supported for replicate generate" unless tool_choice.nil?

      validate_no_webhook_support!(resolved_provider_options)
      adapter = provider_adapter(provider, resolved_provider_options)
      generate_replicate_single(
        model_descriptor: model_descriptor,
        adapter: adapter,
        input: input,
        provider_options: resolved_provider_options
      )
    else
      raise ProviderError, "Unsupported provider: #{provider}"
    end

  after_generation_called = true
  call_callback!(after_generation, { ok: true, result: result }, callback_control)
  result
rescue ProviderError => e
  unless after_generation_called
    after_generation_called = true
    call_callback!(after_generation, { ok: false, error: e }, callback_control)
  end
  raise
rescue StandardError => e
  unless after_generation_called
    after_generation_called = true
    call_callback!(after_generation, { ok: false, error: e }, callback_control)
  end
  raise Error, e.message
end

.loop(model:, messages: nil, system: nil, prompt: nil, tools: {}, tool_choice: nil, stop_when: nil, max_iterations: DEFAULT_MAX_ITERATIONS, temperature: nil, max_tokens: nil, provider_options: {}, before_tool_execution: nil, after_tool_execution: nil, before_iteration: nil, after_iteration: nil, before_generation: nil, after_generation: nil) ⇒ Object 



433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
 
# File 'lib/zuno.rb', line 433

def loop(
  model:,
  messages: nil,
  system: nil,
  prompt: nil,
  tools: {},
  tool_choice: nil,
  stop_when: nil,
  max_iterations: DEFAULT_MAX_ITERATIONS,
  temperature: nil,
  max_tokens: nil,
  provider_options: {},
  before_tool_execution: nil,
  after_tool_execution: nil,
  before_iteration: nil,
  after_iteration: nil,
  before_generation: nil,
  after_generation: nil
)
  callback_control = nil
  callback_context = nil
  model_descriptor = normalize_model(model)
  unless i[openrouter ai_gateway].include?(model_descriptor.provider.to_sym)
    raise Error, "loop only supports openrouter or ai_gateway provider"
  end

  resolved_provider_options = merge_provider_options(
    model_descriptor.provider_options,
    provider_options
  )
  adapter = provider_adapter(model_descriptor.provider, resolved_provider_options)
  callback_context = build_callback_context(
    messages: messages,
    system: system,
    prompt: prompt,
    tools: tools
  )
  resolved_stop_when = normalize_stop_when(stop_when)
  resolved_max_iterations = normalize_max_iterations(max_iterations)
  after_generation_called = false
  callback_control = CallbackControl.new

  call_callback!(
    before_generation,
    {
      model: model_descriptor,
      messages: materialize_messages(messages: callback_context.messages, system: callback_context.system),
      tool_names: callback_context.tools.keys,
      tool_choice: normalize_tool_choice(
        explicit_tool_choice: tool_choice,
        provider_options: resolved_provider_options,
        tools: callback_context.tools
      ),
      max_iterations: resolved_max_iterations,
      stop_when: resolved_stop_when
    },
    callback_control,
    callback_context
  )
  normalize_callback_context!(callback_context)
  if callback_control.stopped?
    result = callback_stopped_result(
      control: callback_control,
      iterations: [],
      message: {},
      usage: nil,
      raw_response: nil
    )
    after_generation_called = true
    call_callback!(after_generation, { ok: true, result: result }, callback_control, callback_context)
    return result
  end

  iterations = []
  iteration_count = 0

  infinite_iterations = resolved_max_iterations == :infinite

  while infinite_iterations || iteration_count < resolved_max_iterations
    current_iteration = iteration_count + 1
    call_callback!(
      before_iteration,
      {
        iteration_index: current_iteration,
        messages: materialize_messages(messages: callback_context.messages, system: callback_context.system)
      },
      callback_control,
      callback_context
    )
    normalize_callback_context!(callback_context)
    if callback_control.stopped?
      result = callback_stopped_result(
        control: callback_control,
        iterations: iterations,
        message: {},
        usage: nil,
        raw_response: nil
      )
      after_generation_called = true
      call_callback!(after_generation, { ok: true, result: result }, callback_control, callback_context)
      return result
    end

    resolved_tool_choice = normalize_tool_choice(
      explicit_tool_choice: tool_choice,
      provider_options: resolved_provider_options,
      tools: callback_context.tools
    )
    payload = build_payload(
      model_id: model_descriptor.id,
      provider: model_descriptor.provider,
      messages: materialize_messages(messages: callback_context.messages, system: callback_context.system),
      tools: callback_context.tools,
      tool_choice: resolved_tool_choice,
      temperature: temperature,
      max_tokens: max_tokens,
      provider_options: resolved_provider_options
    )

    response = adapter.chat(payload)
    message = response.dig("choices", 0, "message") || {}
    tool_calls = Array(message["tool_calls"])

    iteration_record = {
      index: current_iteration,
      message: message,
      tool_calls: tool_calls,
      usage: response["usage"],
      finish_reason: response.dig("choices", 0, "finish_reason"),
      tool_results: []
    }

    if tool_calls.empty?
      iterations << iteration_record
      call_callback!(
        after_iteration,
        {
          iteration_index: current_iteration,
          iteration: iteration_record
        },
        callback_control,
        callback_context
      )
      normalize_callback_context!(callback_context)
      if callback_control.stopped?
        result = callback_stopped_result(
          control: callback_control,
          iterations: iterations,
          message: message,
          usage: response["usage"],
          raw_response: response
        )
        after_generation_called = true
        call_callback!(after_generation, { ok: true, result: result }, callback_control, callback_context)
        return result
      end

      result = {
        text: extract_message_text(message),
        message: message,
        usage: response["usage"],
        finish_reason: response.dig("choices", 0, "finish_reason"),
        iterations: iterations,
        raw_response: response
      }

      after_generation_called = true
      call_callback!(after_generation, { ok: true, result: result }, callback_control, callback_context)
      return result
    end

    callback_context.add_message(build_assistant_tool_call_message(message: message, tool_calls: tool_calls))
    stop_triggered = false
    stop_triggered_tool_name = nil

    tool_calls.each do |tool_call|
      tool_call_id = normalize_tool_call_id(tool_call["id"])
      arguments = parse_arguments(tool_call.dig("function", "arguments"))
      tool_name = tool_call.dig("function", "name").to_s

      call_callback!(
        before_tool_execution,
        {
          iteration_index: current_iteration,
          tool_call_id: tool_call_id,
          tool_name: tool_name,
          input: arguments,
          raw_tool_call: tool_call
        },
        callback_control,
        callback_context
      )
      normalize_callback_context!(callback_context)

      tool_result = execute_tool_call(
        tool_call: tool_call,
        tools: callback_context.tools,
        tool_call_id: tool_call_id,
        arguments: arguments
      )

      iteration_record[:tool_results] << tool_result
      call_callback!(
        after_tool_execution,
        tool_result.merge(iteration_index: current_iteration),
        callback_control,
        callback_context
      )
      normalize_callback_context!(callback_context)
      if tool_stop_condition_met?(resolved_stop_when, tool_result)
        stop_triggered = true
        stop_triggered_tool_name ||= tool_result[:tool_name]
      end

      callback_context.add_message(
        "role" => "tool",
        "tool_call_id" => tool_result[:tool_call_id],
        "content" => serialize_tool_content(tool_result[:output])
      )
    end

    iterations << iteration_record
    call_callback!(
      after_iteration,
        {
          iteration_index: current_iteration,
          iteration: iteration_record
        },
        callback_control,
        callback_context
      )
    normalize_callback_context!(callback_context)
    if callback_control.stopped?
      result = callback_stopped_result(
        control: callback_control,
        iterations: iterations,
        message: message,
        usage: response["usage"],
        raw_response: response
      )
      after_generation_called = true
      call_callback!(after_generation, { ok: true, result: result }, callback_control, callback_context)
      return result
    end

    if stop_triggered
      result = {
        text: extract_message_text(message),
        message: message,
        usage: response["usage"],
        finish_reason: "stop_when_tool_called",
        stop_reason: {
          type: "tool_called",
          tool_name: stop_triggered_tool_name
        },
        iterations: iterations,
        raw_response: response
      }

      after_generation_called = true
      call_callback!(after_generation, { ok: true, result: result }, callback_control, callback_context)
      return result
    end

    iteration_count += 1
  end

  raise MaxIterationsExceeded,
    "Reached max_iterations=#{resolved_max_iterations} without a final assistant response" unless infinite_iterations
rescue ProviderError, MaxIterationsExceeded => e
  unless after_generation_called
    after_generation_called = true
    call_callback!(after_generation, { ok: false, error: e }, callback_control, callback_context)
  end
  raise
rescue StandardError => e
  unless after_generation_called
    after_generation_called = true
    call_callback!(after_generation, { ok: false, error: e }, callback_control, callback_context)
  end
  raise Error, e.message
end

.model(id, provider: :openrouter, provider_options: {}) ⇒ Object 



185
186
187
 
# File 'lib/zuno.rb', line 185

def model(id, provider: :openrouter, provider_options: {})
  ModelDescriptor.new(id: id, provider: provider, provider_options: provider_options)
end

.openrouter(api_key: nil, app_url: nil, title: nil, timeout: Providers::OpenRouter::DEFAULT_TIMEOUT) ⇒ Object 



189
190
191
192
193
194
195
196
 
# File 'lib/zuno.rb', line 189

def openrouter(api_key: nil, app_url: nil, title: nil, timeout: Providers::OpenRouter::DEFAULT_TIMEOUT)
  Providers::OpenRouter.new(
    api_key: api_key,
    app_url: app_url,
    title: title,
    timeout: timeout
  )
end

.replicate(api_key: nil, timeout: Providers::Replicate::DEFAULT_TIMEOUT) ⇒ Object 



198
199
200
201
202
203
 
# File 'lib/zuno.rb', line 198

def replicate(api_key: nil, timeout: Providers::Replicate::DEFAULT_TIMEOUT)
  Providers::Replicate.new(
    api_key: api_key,
    timeout: timeout
  )
end

.stream(model:, messages: nil, system: nil, prompt: nil, temperature: nil, max_tokens: nil, provider_options: {}, &block) ⇒ Object 



716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
 
# File 'lib/zuno.rb', line 716

def stream(
  model:,
  messages: nil,
  system: nil,
  prompt: nil,
  temperature: nil,
  max_tokens: nil,
  provider_options: {},
  &block
)
  raise ArgumentError, "stream requires a block callback" unless block_given?

  model_descriptor = normalize_model(model)
  unless i[openrouter ai_gateway].include?(model_descriptor.provider.to_sym)
    raise ProviderError, "stream only supports openrouter or ai_gateway provider"
  end

  resolved_provider_options = merge_provider_options(
    model_descriptor.provider_options,
    provider_options
  )
  adapter = provider_adapter(model_descriptor.provider, resolved_provider_options)
  llm_messages = normalize_messages(messages: messages, system: system, prompt: prompt)

  payload = build_payload(
    model_id: model_descriptor.id,
    provider: model_descriptor.provider,
    messages: llm_messages,
    tools: {},
    tool_choice: nil,
    temperature: temperature,
    max_tokens: max_tokens,
    provider_options: resolved_provider_options
  ).merge("stream" => true)

  block.call(type: :start, model: model_descriptor.id, provider: model_descriptor.provider)

  emitted_finish = false

  adapter.stream(payload) do |raw_data|
    next if raw_data == "[DONE]"

    parsed = parse_json(raw_data)
    raise StreamingError, "Malformed SSE payload: #{raw_data}" if parsed.nil?

    usage = parsed["usage"]
    block.call(type: :usage, usage: usage, raw: parsed) if usage.is_a?(Hash)

    choice = Array(parsed["choices"]).first || {}
    delta = choice["delta"] || {}

    text_delta = extract_text_delta(delta)
    block.call(type: :text_delta, text: text_delta, raw: parsed) if text_delta && !text_delta.empty?

    Array(delta["tool_calls"]).each do |tool_call_delta|
      tool_delta = {
        index: tool_call_delta["index"],
        id: tool_call_delta["id"],
        type: tool_call_delta["type"],
        name: tool_call_delta.dig("function", "name"),
        arguments_delta: tool_call_delta.dig("function", "arguments")
      }
      block.call(type: :tool_call_delta, tool_call: tool_delta, raw: parsed)
    end

    finish_reason = choice["finish_reason"]
    next if finish_reason.nil? || finish_reason.to_s.empty?

    emitted_finish = true
    block.call(type: :finish, finish_reason: finish_reason, raw: parsed)
  end

  block.call(type: :finish, finish_reason: "stop") unless emitted_finish
  true
rescue ProviderError => e
  wrapped_error = StreamingError.new(e.message)
  block.call(type: :error, error: wrapped_error.message) if block_given?
  raise wrapped_error
rescue StreamingError => e
  block.call(type: :error, error: e.message) if block_given?
  raise
rescue StandardError => e
  wrapped_error = StreamingError.new(e.message)
  block.call(type: :error, error: wrapped_error.message) if block_given?
  raise wrapped_error
end

.tool(name:, description:, input_schema:, &execute) ⇒ Object

Raises:



229
230
231
232
233
234
235
236
237
238
 
# File 'lib/zuno.rb', line 229

def tool(name:, description:, input_schema:, &execute)
  raise ToolError, "A block is required for tool '#{name}'" unless block_given?

  ToolDefinition.new(
    name: name,
    description: description,
    input_schema: input_schema,
    execute_proc: execute
  )
end

.transcribe(model_id: "scribe_v2", file: nil, cloud_storage_url: nil, source_url: nil, provider_options: {}, **options) ⇒ Object 



240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
 
# File 'lib/zuno.rb', line 240

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