smart_prompt 0.4.4 → 0.5.1

data/workers/video_generation_workers.rb

@@ -0,0 +1,264 @@
+# Video Generation Workers for SmartPrompt
+# These workers demonstrate the new video generation capabilities
+
+# Text-to-video generation worker
+SmartPrompt.define_worker :video_generator do
+  use "video_gen"
+  model "Wan-AI/Wan2.2-T2V-A14B"
+
+  # Prepare parameters for video generation
+  generation_params = {
+    prompt: params[:prompt],
+    duration: params[:duration] || 4,
+    resolution: params[:resolution] || "720p",
+    fps: params[:fps] || 24,
+    seed: params[:seed]
+  }
+
+  # Call the video generation adapter directly
+  adapter = engine.llms["video_gen"]
+  video_data = adapter.generate_video(
+    generation_params[:prompt],
+    model: params[:model],
+    duration: generation_params[:duration],
+    resolution: generation_params[:resolution],
+    fps: generation_params[:fps],
+    seed: generation_params[:seed]
+  )
+
+  # Wait for completion and download if requested
+  if params[:wait_for_completion]
+    completed_video = adapter.wait_for_video_completion(
+      video_data[:job_id],
+      check_interval: params[:check_interval] || 10,
+      timeout: params[:timeout] || 600
+    )
+
+    # Download video if URL is available
+    if completed_video[:video_url] && params[:download_to_file]
+      output_dir = params[:output_dir] || "./generated_videos"
+      filename_prefix = params[:filename_prefix] || "generated_video"
+      output_path = File.join(output_dir, "#{filename_prefix}_#{video_data[:job_id]}.mp4")
+
+      downloaded_file = adapter.download_video(completed_video[:video_url], output_path)
+      { video_data: completed_video, downloaded_file: downloaded_file }
+    else
+      { video_data: completed_video }
+    end
+  else
+    { video_data: video_data }
+  end
+end
+
+# Image-to-video generation worker
+SmartPrompt.define_worker :image_to_video_generator do
+  use "video_gen"
+  model "Wan-AI/Wan2.2-I2V-A14B"
+
+  # Prepare parameters for image-to-video generation
+  generation_params = {
+    image_file: params[:image_file],
+    prompt: params[:prompt],
+    duration: params[:duration] || 4,
+    resolution: params[:resolution] || "720p",
+    fps: params[:fps] || 24,
+    seed: params[:seed]
+  }
+
+  # Call the image-to-video adapter directly
+  adapter = engine.llms["video_gen"]
+  video_data = adapter.create_video_from_image(
+    generation_params[:image_file],
+    generation_params[:prompt],
+    model: params[:model],
+    duration: generation_params[:duration],
+    resolution: generation_params[:resolution],
+    fps: generation_params[:fps],
+    seed: generation_params[:seed]
+  )
+
+  # Wait for completion and download if requested
+  if params[:wait_for_completion]
+    completed_video = adapter.wait_for_video_completion(
+      video_data[:job_id],
+      check_interval: params[:check_interval] || 10,
+      timeout: params[:timeout] || 600
+    )
+
+    # Download video if URL is available
+    if completed_video[:video_url] && params[:download_to_file]
+      output_dir = params[:output_dir] || "./generated_videos"
+      filename_prefix = params[:filename_prefix] || "image_to_video"
+      output_path = File.join(output_dir, "#{filename_prefix}_#{video_data[:job_id]}.mp4")
+
+      downloaded_file = adapter.download_video(completed_video[:video_url], output_path)
+      { video_data: completed_video, downloaded_file: downloaded_file }
+    else
+      { video_data: completed_video }
+    end
+  else
+    { video_data: video_data }
+  end
+end
+
+# Video status checker worker
+SmartPrompt.define_worker :video_status_checker do
+  use "video_gen"
+
+  # Check video generation status
+  adapter = engine.llms["video_gen"]
+  status_data = adapter.check_video_status(params[:job_id])
+
+  # Download video if completed and requested
+  if status_data[:status] == "completed" && status_data[:video_url] && params[:download_to_file]
+    output_dir = params[:output_dir] || "./generated_videos"
+    filename_prefix = params[:filename_prefix] || "video"
+    output_path = File.join(output_dir, "#{filename_prefix}_#{params[:job_id]}.mp4")
+
+    downloaded_file = adapter.download_video(status_data[:video_url], output_path)
+    { status_data: status_data, downloaded_file: downloaded_file }
+  else
+    { status_data: status_data }
+  end
+end
+
+# Creative video generation worker
+SmartPrompt.define_worker :creative_video_generator do
+  use "video_gen"
+  model "Wan-AI/Wan2.2-T2V-A14B"
+
+  # Prepare creative parameters
+  creative_params = {
+    prompt: params[:prompt] || "A beautiful animated scene",
+    duration: params[:duration] || 4,
+    resolution: params[:resolution] || "720p",
+    fps: params[:fps] || 24,
+    seed: params[:seed]
+  }
+
+  # Add creative style to prompt if specified
+  style_prompt = creative_params[:prompt]
+  if params[:video_style]
+    style_prompt = "#{style_prompt}, in the style of #{params[:video_style]}"
+  end
+
+  adapter = engine.llms["video_gen"]
+  video_data = adapter.generate_video(
+    style_prompt,
+    model: params[:model],
+    duration: creative_params[:duration],
+    resolution: creative_params[:resolution],
+    fps: creative_params[:fps],
+    seed: creative_params[:seed]
+  )
+
+  # Wait for completion and download if requested
+  if params[:wait_for_completion]
+    completed_video = adapter.wait_for_video_completion(
+      video_data[:job_id],
+      check_interval: params[:check_interval] || 10,
+      timeout: params[:timeout] || 600
+    )
+
+    if completed_video[:video_url] && params[:download_to_file]
+      output_dir = params[:output_dir] || "./creative_videos"
+      filename_prefix = params[:filename_prefix] || "creative_video"
+      output_path = File.join(output_dir, "#{filename_prefix}_#{video_data[:job_id]}.mp4")
+
+      downloaded_file = adapter.download_video(completed_video[:video_url], output_path)
+      { video_data: completed_video, downloaded_file: downloaded_file }
+    else
+      { video_data: completed_video }
+    end
+  else
+    { video_data: video_data }
+  end
+end
+
+# Product video generation worker
+SmartPrompt.define_worker :product_video_generator do
+  use "video_gen"
+  model "Wan-AI/Wan2.2-T2V-A14B"
+
+  # Prepare product-specific parameters
+  product_params = {
+    prompt: params[:prompt] || "Professional product showcase",
+    duration: params[:duration] || 4,
+    resolution: params[:resolution] || "720p",
+    fps: params[:fps] || 24,
+    seed: params[:seed]
+  }
+
+  # Enhance prompt for product video
+  enhanced_prompt = "Professional product video, smooth animation, cinematic quality, #{product_params[:prompt]}"
+
+  adapter = engine.llms["video_gen"]
+  video_data = adapter.generate_video(
+    enhanced_prompt,
+    model: params[:model],
+    duration: product_params[:duration],
+    resolution: product_params[:resolution],
+    fps: product_params[:fps],
+    seed: product_params[:seed]
+  )
+
+  # Wait for completion and download if requested
+  if params[:wait_for_completion]
+    completed_video = adapter.wait_for_video_completion(
+      video_data[:job_id],
+      check_interval: params[:check_interval] || 10,
+      timeout: params[:timeout] || 600
+    )
+
+    if completed_video[:video_url] && params[:download_to_file]
+      output_dir = params[:output_dir] || "./product_videos"
+      filename_prefix = params[:filename_prefix] || "product_video"
+      output_path = File.join(output_dir, "#{filename_prefix}_#{video_data[:job_id]}.mp4")
+
+      downloaded_file = adapter.download_video(completed_video[:video_url], output_path)
+      { video_data: completed_video, downloaded_file: downloaded_file }
+    else
+      { video_data: completed_video }
+    end
+  else
+    { video_data: video_data }
+  end
+end
+
+# Batch video generation worker
+SmartPrompt.define_worker :batch_video_generator do
+  use "video_gen"
+  model "Wan-AI/Wan2.2-T2V-A14B"
+
+  results = []
+
+  # Process multiple prompts
+  prompts = params[:prompts] || [params[:prompt]]
+  prompts.each_with_index do |prompt, index|
+    generation_params = {
+      prompt: prompt,
+      duration: params[:duration] || 4,
+      resolution: params[:resolution] || "720p",
+      fps: params[:fps] || 24,
+      seed: params[:seed] ? params[:seed] + index : nil
+    }
+
+    adapter = engine.llms["video_gen"]
+    video_data = adapter.generate_video(
+      generation_params[:prompt],
+      model: params[:model],
+      duration: generation_params[:duration],
+      resolution: generation_params[:resolution],
+      fps: generation_params[:fps],
+      seed: generation_params[:seed]
+    )
+
+    results << {
+      prompt: prompt,
+      video_data: video_data,
+      index: index
+    }
+  end
+
+  { batch_results: results }
+end

data/workers/zhipu_workers.rb

@@ -0,0 +1,113 @@
+# 智谱 AI (Zhipu BigModel) workers for SmartPrompt
+#
+# One worker per model category, reusing the standard DSL (`use`, `model`, `sys_msg`,
+# `prompt`, `send_msg`) and the media helpers. Chat/vision/embed/image go through
+# Conversation-delegated methods; video/tts/asr reach the adapter directly via
+# engine.llms[...] (same pattern as workers/video_generation_workers.rb).
+#
+# `send_msg` transparently becomes streaming when the engine invokes the worker via
+# call_worker_by_stream — so :glm_chat serves both sync and stream callers.
+
+# 1. 文本对话 (sync + stream)
+SmartPrompt.define_worker :glm_chat do
+  use "glm"
+  model params[:model] if params[:model]
+  sys_msg(params[:system] || "你是一个有帮助的中文助手，回答简洁准确。", params)
+  prompt(params[:prompt] || "你好，请介绍一下智谱GLM。")
+  send_msg
+end
+
+# 2. 图文多模态 (single image_url or image_urls array)
+SmartPrompt.define_worker :glm_vision do
+  use "glm_vision"
+  model params[:model] if params[:model]
+  sys_msg("你是一个专业的多模态图像分析助手。", params)
+
+  content = [{ type: "text", text: params[:question] || "请描述这张图片中的内容。" }]
+  images = params[:image_urls] || [params[:image_url]]
+  images.each do |url|
+    content << { type: "image_url", image_url: { url: url } }
+  end
+  add_message({ role: "user", content: content })
+
+  send_msg
+end
+
+# 3. 向量模型
+SmartPrompt.define_worker :glm_embed do
+  use "embedding"
+  model params[:model] if params[:model]
+  prompt(params[:text] || "智谱GLM大模型")
+  embeddings(params[:length] || 1024)
+end
+
+# 4. 文生图 (CogView / GLM-Image)
+SmartPrompt.define_worker :cogview_image do
+  use "cogview"
+  model params[:model] if params[:model]
+  images = generate_image(params[:prompt], params)
+  if params[:save_to_file]
+    saved = save_image(images, params[:output_dir] || "./generated_images", params[:filename_prefix] || "zhipu")
+    { images: images, saved_files: saved }
+  else
+    images
+  end
+end
+
+# 5. 文生视频 (async: submit -> wait -> download)
+SmartPrompt.define_worker :cogvideo_video do
+  use "cogvideo"
+  model params[:model] if params[:model]
+  adapter = engine.llms["cogvideo"]
+
+  submitted = adapter.generate_video(params[:prompt], params)
+  result = { submitted: submitted }
+
+  if params[:wait_for_completion]
+    completed = adapter.wait_for_video_completion(
+      submitted[:task_id],
+      check_interval: params[:check_interval] || 10,
+      timeout: params[:timeout] || 600
+    )
+    if completed[:video_url] && params[:download_to_file]
+      output_dir = params[:output_dir] || "./generated_videos"
+      prefix = params[:filename_prefix] || "zhipu_video"
+      output_path = File.join(output_dir, "#{prefix}_#{submitted[:task_id]}.mp4")
+      downloaded = adapter.download_video(completed[:video_url], output_path)
+      result = { submitted: submitted, video: completed, downloaded_file: downloaded }
+    else
+      result = { submitted: submitted, video: completed }
+    end
+  end
+  result
+end
+
+# 6. 语音合成 (GLM-TTS)
+SmartPrompt.define_worker :glm_tts do
+  use "glm_tts"
+  model params[:model] if params[:model]
+  adapter = engine.llms["glm_tts"]
+
+  output_path = params[:output_path] || "./generated_audio/zhipu_tts.wav"
+  info = adapter.synthesize_to_file(
+    params[:text],
+    output_path,
+    voice: params[:voice],
+    model: params[:model],
+    response_format: params[:response_format] || "wav"
+  )
+  info
+end
+
+# 7. 语音识别 (GLM-ASR-2512)
+SmartPrompt.define_worker :glm_asr do
+  use "glm_asr"
+  model params[:model] if params[:model]
+  adapter = engine.llms["glm_asr"]
+
+  adapter.transcribe_audio(
+    params[:audio_file],
+    model: params[:model],
+    language: params[:language]
+  )
+end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: smart_prompt
 version: !ruby/object:Gem::Version
-  version: 0.4.4
+  version: 0.5.1
 platform: ruby
 authors:
 - zhuang biaowei
@@ -107,6 +107,34 @@ dependencies:
     - - "~>"
       - !ruby/object:Gem::Version
         version: 0.3.0
+- !ruby/object:Gem::Dependency
+  name: tiktoken_ruby
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: 0.0.6
+  type: :runtime
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: 0.0.6
+- !ruby/object:Gem::Dependency
+  name: anthropic
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: '1.14'
+  type: :runtime
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: '1.14'
 description: SmartPrompt provides a flexible DSL for managing prompts, interacting
   with multiple LLMs, and creating composable task workers.
 email:
@@ -120,19 +148,78 @@ files:
 - README.cn.md
 - README.md
 - Rakefile
+- config/anthropic_config.yml
+- config/image_generation_config.yml
+- config/multimodal_config.yml
+- config/sensenova_config.yml
+- config/zhipu_config.yml
+- docs/ANTHROPIC_EXAMPLES.md
+- docs/CONVERSATION_INTEGRATION_SUMMARY.md
+- docs/HISTORY_EXAMPLES_README.md
+- docs/HISTORY_MANAGEMENT_GUIDE.md
+- docs/MONITORING_GUIDE.md
+- docs/MULTIMODAL_README.md
+- docs/RELEVANCE_BASED_STRATEGY_IMPLEMENTATION.md
+- docs/STT_README.md
+- docs/TTS_README.md
+- docs/VIDEO_GENERATION_README.md
+- docs/delete_files_list.md
+- examples/anthropic_basic_chat.rb
+- examples/anthropic_example.rb
+- examples/anthropic_multimodal.rb
+- examples/anthropic_streaming.rb
+- examples/anthropic_tool_calling.rb
+- examples/automatic_cleanup_example.rb
+- examples/history_management_examples.rb
+- examples/image_generation_example.rb
+- examples/monitoring_example.rb
+- examples/multimodal_example.rb
+- examples/relevance_based_strategy_example.rb
+- examples/sensenova_example.rb
+- examples/stt_example.rb
+- examples/tts_example.rb
+- examples/video_generation_example.rb
+- examples/zhipu_example.rb
 - lib/smart_prompt.rb
 - lib/smart_prompt/anthropic_adapter.rb
 - lib/smart_prompt/api_handler.rb
+- lib/smart_prompt/compression_engine.rb
+- lib/smart_prompt/context_strategy.rb
 - lib/smart_prompt/conversation.rb
 - lib/smart_prompt/db_adapter.rb
 - lib/smart_prompt/engine.rb
+- lib/smart_prompt/history_manager.rb
+- lib/smart_prompt/hybrid_strategy.rb
+- lib/smart_prompt/image_generation_adapter.rb
 - lib/smart_prompt/llamacpp_adapter.rb
 - lib/smart_prompt/llm_adapter.rb
+- lib/smart_prompt/lru_cache.rb
+- lib/smart_prompt/message.rb
+- lib/smart_prompt/multimodal_adapter.rb
 - lib/smart_prompt/openai_adapter.rb
+- lib/smart_prompt/persistence_layer.rb
 - lib/smart_prompt/prompt_template.rb
+- lib/smart_prompt/relevance_based_strategy.rb
+- lib/smart_prompt/sensenova_adapter.rb
+- lib/smart_prompt/session.rb
+- lib/smart_prompt/sliding_window_strategy.rb
+- lib/smart_prompt/stt_adapter.rb
+- lib/smart_prompt/summary_based_strategy.rb
+- lib/smart_prompt/token_counter.rb
+- lib/smart_prompt/tts_adapter.rb
 - lib/smart_prompt/version.rb
+- lib/smart_prompt/video_generation_adapter.rb
 - lib/smart_prompt/worker.rb
+- lib/smart_prompt/zhipu_adapter.rb
 - sig/smart_prompt.rbs
+- workers/history_management_examples.rb
+- workers/image_generation_workers.rb
+- workers/multimodal_workers.rb
+- workers/sensenova_workers.rb
+- workers/stt_workers.rb
+- workers/tts_workers.rb
+- workers/video_generation_workers.rb
+- workers/zhipu_workers.rb
 homepage: https://github.com/zhuangbiaowei/smart_prompt
 licenses:
 - MIT