Module: PWN::AI::Ollama

Defined in:
lib/pwn/ai/ollama.rb

Overview

Direct client for a local/remote Ollama server REST API. No API key is required for a stock ollama serve (http://127.0.0.1:11434). Paths are native Ollama:

GET  /api/tags
POST /api/chat          (native tool_calls, options.num_ctx / num_predict)
POST /api/embed
POST /v1/chat/completions (OpenAI-compat shim)

Spec: https://github.com/ollama/ollama/blob/main/docs/api.md

Constant Summary collapse

DEFAULT_BASE_URI = 
'http://127.0.0.1:11434'

Class Method Summary collapse

Class Method Details

.authors ⇒ Object

Author(s)

0day Inc. support@0dayinc.com



647
648
649
650
651
 
# File 'lib/pwn/ai/ollama.rb', line 647

public_class_method def self.authors
  "AUTHOR(S):
    0day Inc. <support@0dayinc.com>
  "
end

.chat(opts = {}) ⇒ Object

Supported Method Parameters

response = PWN::AI::Ollama.chat( request: 'required - message to Ollama' model: 'optional - model to use for text generation (defaults to PWN::Env[:ollama][:model])', temp: 'optional - creative response float (defaults to PWN::Env[:ollama][:temp])', system_role_content: 'optional - context to set up the model behavior for conversation (Default: PWN::Env[:ollama][:system_role_content])', response_history: 'optional - pass response back in to have a conversation', speak_answer: 'optional speak answer using PWN::Plugins::Voice.text_to_speech (Default: nil)', timeout: 'optional timeout in seconds (defaults to 900)', spinner: 'optional - display spinner (defaults to false)' )



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
 
# File 'lib/pwn/ai/ollama.rb', line 554

public_class_method def self.chat(opts = {})
  engine = PWN::Env[:ai][:ollama]
  request = opts[:request]
  max_prompt_length = engine[:max_prompt_length] ||= 1_000_000
  request_trunc_idx = ((max_prompt_length - 1) / 3.36).floor
  request = request[0..request_trunc_idx]

  model = opts[:model] ||= engine[:model]
  raise 'ERROR: Model is required.  Call #get_models method for details' if model.nil?

  temp = opts[:temp].to_f ||= engine[:temp].to_f
  temp = 1 if temp.zero?

  # Native Ollama chat. Stock ollama serve and many reverse proxies
  # expose /api/chat, not the OpenAI-compat shim /v1/chat/completions
  # (that path 404s when the shim is disabled or not mounted).
  rest_call = 'api/chat'

  response_history = opts[:response_history]

  system_role_content = opts[:system_role_content] ||= engine[:system_role_content]

  system_role = {
    role: 'system',
    content: system_role_content
  }

  user_role = {
    role: 'user',
    content: request
  }

  response_history ||= { choices: [system_role] }

  num_ctx     = (engine[:num_ctx] || 32_768).to_i
  num_predict = (engine[:num_predict] || 4_096).to_i
  keep_alive  = engine[:keep_alive] || '30m'

  http_body = {
    model: model,
    messages: [system_role],
    stream: true,
    keep_alive: keep_alive,
    options: {
      num_ctx: num_ctx,
      num_predict: num_predict,
      temperature: temp
    }
  }

  if response_history[:choices].length > 1
    response_history[:choices][1..-1].each do |message|
      http_body[:messages].push(message)
    end
  end

  http_body[:messages].push(user_role)

  timeout = opts[:timeout]
   = opts[:spinner]

  response = ollama_rest_call(
    http_method: :post,
    rest_call: rest_call,
    http_body: http_body,
    timeout: timeout,
    spinner: 
  )

  json_resp = JSON.parse(response, symbolize_names: true)
  assistant_resp = json_resp[:message] || json_resp.dig(:choices, 0, :message)
  raise "ERROR: Ollama chat response missing message/choices: #{json_resp.inspect[0, 400]}" if assistant_resp.nil?

  json_resp[:choices] = http_body[:messages]
  json_resp[:choices].push(assistant_resp)

  speak_answer = true if opts[:speak_answer]

  if speak_answer
    answer = assistant_resp[:content]
    text_path = "/tmp/#{SecureRandom.hex}.pwn_voice"
    File.write(text_path, answer)
    PWN::Plugins::Voice.text_to_speech(text_path: text_path)
    File.unlink(text_path)
  end

  json_resp
rescue StandardError => e
  raise e
end

.chat_with_tools(opts = {}) ⇒ Object

Supported Method Parameters

response = PWN::AI::Ollama.chat_with_tools( messages: 'required - full OpenAI-format messages array (system/user/assistant/tool)', tools: 'optional - OpenAI tools array [function:{...}]', tool_choice: 'optional - "auto" | "none" | function:{name:..}', model: 'optional - overrides PWN::Env[:ollama][:model]', temp: 'optional - temperature (defaults to PWN::Env[:ollama][:temp] || 1)', timeout: 'optional - seconds (default 900)', spinner: 'optional - display spinner (default false)' )

Hits Ollama NATIVE POST /api/chat so options.num_ctx / num_predict / keep_alive take effect. Streaming is ON; ollama_rest_call assembles NDJSON chunks back into a single response.



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
 
# File 'lib/pwn/ai/ollama.rb', line 481

public_class_method def self.chat_with_tools(opts = {})
  engine   = PWN::Env[:ai][:ollama]
  messages = normalize_messages_for_ollama(messages: opts[:messages])
  raise 'ERROR: messages array is required' if messages.nil? || messages.empty?

  model = opts[:model] ||= engine[:model]
  raise 'ERROR: Model is required.  Call #get_models method for details' if model.nil?

  temp = opts[:temp].to_f
  temp = engine[:temp].to_f.nonzero? || 1 if temp.zero?

  tools_present = opts[:tools] && !opts[:tools].empty?
  tool_temp     = (engine[:tool_temp] || 0.1).to_f
  num_ctx       = (engine[:num_ctx]   || 32_768).to_i
  num_predict   = (engine[:num_predict] || 4_096).to_i
  keep_alive    = engine[:keep_alive] || '30m'

  http_body = {
    model: model,
    messages: messages,
    stream: true,
    keep_alive: keep_alive,
    options: {
      num_ctx: num_ctx,
      num_predict: num_predict,
      temperature: tools_present ? tool_temp : temp
    }
  }
  if tools_present
    http_body[:tools] = opts[:tools]
    fmt = engine[:format]
    http_body[:format] = fmt unless fmt.nil? || fmt.to_s.empty?
  end
  http_body[:tool_choice] = opts[:tool_choice] if opts[:tool_choice]

  response = ollama_rest_call(
    http_method: :post,
    rest_call: 'api/chat',
    http_body: http_body,
    timeout: opts[:timeout],
    spinner: opts[:spinner]
  )
  raise 'ERROR: Ollama chat_with_tools received empty response from ollama_rest_call' if response.nil? || (response.respond_to?(:empty?) && response.empty?)

  json_resp = JSON.parse(response, symbolize_names: true)
  msg = json_resp[:message] || json_resp.dig(:choices, 0, :message)
  if msg.is_a?(Hash)
    content = msg[:content].to_s
    thinking = msg[:thinking].to_s
    tcalls = Array(msg[:tool_calls])
    msg = msg.merge(content: visible_from_thinking(thinking: thinking)) if content.strip.empty? && !thinking.strip.empty? && tcalls.empty?
  end
  json_resp[:choices] = [{ message: msg }] if msg && !json_resp.key?(:choices)
  json_resp[:assistant_message] = msg
  raise "ERROR: Ollama response missing message/choices: #{json_resp.inspect[0, 400]}" if msg.nil?

  json_resp
rescue StandardError => e
  raise e
end

.get_models ⇒ Object

Supported Method Parameters

response = PWN::AI::Ollama.get_models



369
370
371
372
373
374
375
 
# File 'lib/pwn/ai/ollama.rb', line 369

public_class_method def self.get_models
  models = ollama_rest_call(rest_call: 'api/tags')

  JSON.parse(models, symbolize_names: true)[:models]
rescue StandardError => e
  raise e
end

.get_plan_usage(opts = {}) ⇒ Object

Supported Method Parameters

usage = PWN::AI::Ollama.get_plan_usage

Local Ollama has no subscription / plan-usage endpoint. Always unavailable so the PS1 shows the infinity glyph.



382
383
384
385
 
# File 'lib/pwn/ai/ollama.rb', line 382

public_class_method def self.get_plan_usage(opts = {})
  _unused = opts
  { available: false, engine: :ollama, unlimited: true }
end

.help ⇒ Object

Display Usage for this Module



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
 
# File 'lib/pwn/ai/ollama.rb', line 655

public_class_method def self.help
  puts "USAGE:
    models = #{self}.get_models

    usage = #{self}.get_plan_usage

    response = #{self}.chat(
      request: 'required - message to Ollama',
      model: 'optional - model to use for text generation (defaults to PWN::Env[:ai][:ollama][:model])',
      temp: 'optional - creative response float (defaults to PWN::Env[:ai][:ollama][:temp])',
      system_role_content: 'optional - context to set up the model behavior for conversation (Default: PWN::Env[:ai][:ollama][:system_role_content])',
      response_history: 'optional - pass response back in to have a conversation',
      speak_answer: 'optional speak answer using PWN::Plugins::Voice.text_to_speech (Default: nil)',
      timeout: 'optional - timeout in seconds (defaults to 900)',
      spinner: 'optional - display spinner (defaults to false)'
    )

    response = #{self}.chat_with_tools(
      messages: 'required - messages array',
      tools: 'optional - OpenAI tools array',
      model: 'optional - overrides PWN::Env[:ai][:ollama][:model]'
    )

    #{self}.authors
  "
end