massgen 0.1.1__py3-none-any.whl → 0.1.3__py3-none-any.whl

massgen/configs/tools/custom_tools/computer_use_docker_example.yaml

@@ -0,0 +1,65 @@
+# MassGen Configuration: Computer Use - Docker Environment
+#
+# This configuration uses a Docker container for OS-level automation.
+#
+# Usage:
+#   massgen --config @examples/tools/custom_tools/computer_use_docker_example "Open calculator and compute 123 + 456"
+#
+# Prerequisites:
+#   1. Set OPENAI_API_KEY in your .env file
+#   2. Have Docker installed and running
+#   3. Start the Docker container (see README for instructions)
+
+agents:
+  - id: "docker_automation_agent"
+    backend:
+      type: "openai"
+      model: "computer-use-preview"
+      custom_tools:
+        - name: ["computer_use"]
+          category: "automation"
+          path: "massgen/tool/_computer_use/computer_use_tool.py"
+          function: ["computer_use"]
+          # Default parameters for Docker environment
+          default_params:
+            environment: "ubuntu"
+            display_width: 1280
+            display_height: 800
+            max_iterations: 25
+            include_reasoning: true
+            environment_config:
+              container_name: "cua-container"
+              display: ":99"
+
+    system_message: |
+      You are a computer automation specialist with access to the computer_use tool.
+
+      You can control a virtual Ubuntu environment running in Docker to:
+      - Launch and use desktop applications
+      - Perform file operations
+      - Execute system-level tasks
+      - Automate GUI interactions
+
+      The environment includes:
+      - Ubuntu 22.04 desktop (Xfce)
+      - Firefox browser
+      - Standard desktop applications
+      - X11 display system
+
+      When using the computer_use tool:
+      1. Tasks are executed in a sandboxed Docker container
+      2. You have full desktop access via xdotool commands
+      3. Screenshots are captured after each action
+      4. Be specific about coordinates when clicking
+
+      Best practices:
+      - Allow time for applications to launch
+      - Use wait actions between steps
+      - Verify GUI elements are visible before clicking
+      - Consider screen resolution (1280x800)
+
+ui:
+  display_type: "detailed"
+  logging_enabled: true
+  show_screenshots: true
+  show_reasoning: true

massgen/configs/tools/custom_tools/computer_use_example.yaml

@@ -0,0 +1,50 @@
+# MassGen Configuration: Computer Use Tool Example
+#
+# This configuration demonstrates how to use the computer_use tool for automating
+# browser and computer interactions using OpenAI's computer-use-preview model.
+#
+# Usage:
+#   massgen --config @examples/tools/custom_tools/computer_use_example "Search for Python documentation on Google"
+#
+# Prerequisites:
+#   1. Set OPENAI_API_KEY in your .env file
+#   2. For browser environment: pip install playwright && playwright install
+#   3. For Docker environment: Have Docker installed and running
+
+agents:
+  - id: "computer_use_agent"
+    backend:
+      type: "openai"
+      model: "gpt-4.1"  # You can also use "computer-use-preview" for the main model
+      custom_tools:
+        - name: ["computer_use"]
+          category: "automation"
+          path: "massgen/tool/_computer_use/computer_use_tool.py"
+          function: ["computer_use"]
+
+    system_message: |
+      You are an AI assistant with access to computer automation capabilities.
+
+      The computer_use tool is available to you. This tool allows you to:
+      - Control a web browser (click, type, scroll, etc.)
+      - Automate computer tasks
+      - Search the web, fill forms, navigate websites
+      - Perform multi-step workflows
+
+      When a user asks you to perform a task that requires browser or computer interaction,
+      use the computer_use tool with a clear task description.
+
+      Important:
+      - Always provide clear, specific task descriptions
+      - The tool will execute actions step-by-step
+      - You will receive screenshots and action logs
+      - Safety checks may be triggered - acknowledge them when appropriate
+
+      Example usage:
+      - "Search for the latest AI news on Google"
+      - "Navigate to example.com and fill out the contact form"
+      - "Find Python documentation and save the URL"
+
+ui:
+  display_type: "simple"
+  logging_enabled: true

massgen/configs/tools/custom_tools/crawl4ai_mcp_example.yaml

@@ -0,0 +1,67 @@
+# MassGen Configuration: Crawl4AI Web Scraping via MCP
+#
+# Prerequisites:
+#   1. Start crawl4ai Docker container (one-time setup):
+#      docker pull unclecode/crawl4ai:latest
+#      docker run -d -p 11235:11235 --name crawl4ai --shm-size=1g unclecode/crawl4ai:latest
+#
+#   2. Verify container is running:
+#      docker ps | grep crawl4ai
+#
+#   3. Test MCP endpoint (optional):
+#      curl http://localhost:11235/mcp/schema
+#
+# Usage:
+#   massgen --config massgen/configs/tools/custom_tools/crawl4ai_mcp_example.yaml "Scrape https://example.com and summarize the content"
+#
+# Available Tools (via MCP):
+#   - md: Generate markdown from web content
+#   - html: Extract preprocessed HTML
+#   - screenshot: Capture webpage screenshots
+#   - pdf: Generate PDF documents
+#   - execute_js: Run JavaScript on web pages
+#   - crawl: Perform multi-URL crawling
+#   - ask: Query the Crawl4AI library context
+#
+# Note: Multiple agents can connect to the same crawl4ai container.
+#       The server handles up to 5 concurrent crawls by default.
+
+orchestrator:
+  snapshot_storage: "snapshots"
+  agent_temporary_workspace: "temp_workspaces"
+
+agents:
+  - id: "web_scraper_agent"
+    backend:
+      type: "claude_code"
+      model: "claude-sonnet-4-20250514"
+      cwd: "workspace1"
+
+      # Connect to crawl4ai MCP server
+      mcp_servers:
+        - name: "crawl4ai"
+          type: "sse"  # Server-Sent Events transport
+          url: "http://localhost:11235/mcp/sse"
+
+    append_system_prompt: |
+      You are a web scraping specialist with access to the Crawl4AI toolset via MCP.
+
+      Available tools:
+      - md: Convert webpages to clean markdown (best for LLM consumption)
+      - html: Extract preprocessed HTML
+      - screenshot: Capture webpage as image
+      - pdf: Generate PDF from webpage
+      - execute_js: Run JavaScript on pages (for dynamic content)
+      - crawl: Scrape multiple URLs in parallel
+
+      When users ask to scrape, analyze, or extract web content:
+      1. Use 'md' tool for text-based content (articles, docs, etc.)
+      2. Use 'screenshot' for visual content or layout analysis
+      3. Use 'execute_js' for JavaScript-heavy sites
+      4. Use 'crawl' for multiple pages
+
+      Always provide clear summaries of scraped content.
+
+ui:
+  display_type: "rich_terminal"
+  logging_enabled: true

massgen/configs/tools/custom_tools/crawl4ai_multi_agent_example.yaml

@@ -0,0 +1,68 @@
+# MassGen Configuration: Multi-Agent Web Research with Crawl4AI
+#
+# This example demonstrates multiple agents sharing a single crawl4ai MCP server
+# for collaborative web research and analysis.
+#
+# Prerequisites:
+#   1. Start crawl4ai Docker container:
+#      docker run -d -p 11235:11235 --name crawl4ai --shm-size=1g unclecode/crawl4ai:latest
+#
+# Usage:
+#   massgen --config massgen/configs/tools/custom_tools/crawl4ai_multi_agent_example.yaml "Research AI safety best practices by scraping relevant documentation sites and academic papers"
+#
+# What happens:
+#   - Multiple agents can scrape different URLs concurrently
+#   - The crawl4ai server handles up to 5 parallel crawls
+#   - Agents collaborate and vote on the synthesized research
+
+orchestrator:
+  snapshot_storage: "snapshots"
+  agent_temporary_workspace: "temp_workspaces"
+  voting_sensitivity: "balanced"
+  answer_novelty_requirement: "balanced"
+  max_new_answers_per_agent: 3
+
+agents:
+  - id: "research_agent_1"
+    backend:
+      type: "openai"
+      model: "gpt-5-nano"
+      cwd: "workspace1"
+
+      # All agents connect to the same crawl4ai container
+      mcp_servers:
+        - name: "crawl4ai"
+          type: "sse"
+          url: "http://localhost:11235/mcp/sse"
+
+      enable_web_search: true
+
+    system_message: |
+      You are a research specialist focused on finding authoritative sources and extracting key insights.
+
+      Use the crawl4ai tools to scrape documentation, articles, and research papers.
+      Prioritize official documentation and academic sources.
+
+  - id: "research_agent_2"
+    backend:
+      type: "gemini"
+      model: "gemini-2.5-pro"
+      cwd: "workspace2"
+
+      # Connects to the SAME crawl4ai container as agent_1
+      mcp_servers:
+        - name: "crawl4ai"
+          type: "sse"
+          url: "http://localhost:11235/mcp/sse"
+
+      enable_web_search: true
+
+    system_message: |
+      You are a synthesis specialist who combines information from multiple sources.
+
+      Use crawl4ai to gather diverse perspectives and cross-reference information.
+      Focus on finding patterns and connections across sources.
+
+ui:
+  display_type: "rich_terminal"
+  logging_enabled: true

massgen/configs/tools/custom_tools/multimodal_tools/playwright_with_img_understanding.yaml

@@ -0,0 +1,98 @@
+# massgen --config massgen/configs/tools/custom_tools/multimodal_tools/playwright_with_img_understanding.yaml "Analyze docs.massgen.ai and tell me how to improve its design."
+
+agents:
+- id: agent_a
+  backend:
+    type: openai
+    model: gpt-5-codex
+    text:
+      verbosity: medium
+    reasoning:
+      effort: medium
+      summary: auto
+    cwd: workspace1
+    enable_mcp_command_line: true
+    command_line_execution_mode: docker
+    command_line_docker_network_mode: "bridge"      # Enable network access (default: none)
+    enable_web_search: true
+    custom_tools:
+      - name: ["understand_image"]
+        category: "multimodal"
+        path: "massgen/tool/_multimodal_tools/understand_image.py"
+        function: ["understand_image"]
+    mcp_servers:
+        playwright:
+          type: "stdio"
+          command: "npx"
+          args: [
+            "@playwright/mcp@latest",
+            "--browser=chrome",  # Use Chrome browser
+            "--caps=vision,pdf", # Enable vision and PDF capabilities
+            "--user-data-dir=${cwd}/playwright-profile", # Persistent browser profile within workspace
+            "--output-dir=${cwd}", # Save screenshots/PDFs directly to workspace
+            # "--save-trace"       # Save Playwright traces for debugging
+          ]
+
+- id: agent_b
+  backend:
+    type: claude_code
+    model: claude-sonnet-4-5-20250929
+    cwd: workspace2
+    enable_mcp_command_line: true
+    command_line_execution_mode: docker
+    command_line_docker_network_mode: "bridge"      # Enable network access (default: none)
+    custom_tools:
+      - name: ["understand_image"]
+        category: "multimodal"
+        path: "massgen/tool/_multimodal_tools/understand_image.py"
+        function: ["understand_image"]
+    mcp_servers:
+        playwright:
+          type: "stdio"
+          command: "npx"
+          args: [
+            "@playwright/mcp@latest",
+            "--browser=chrome",  # Use Chrome browser
+            "--caps=vision,pdf", # Enable vision and PDF capabilities
+            "--user-data-dir=${cwd}/playwright-profile", # Persistent browser profile within workspace
+            "--output-dir=${cwd}", # Save screenshots/PDFs directly to workspace
+            # "--save-trace"       # Save Playwright traces for debugging
+          ]
+
+- id: agent_c
+  backend:
+    type: chatcompletion
+    base_url: "https://openrouter.ai/api/v1"
+    model: qwen/qwen3-coder
+    cwd: workspace3
+    enable_mcp_command_line: true
+    command_line_execution_mode: docker
+    command_line_docker_network_mode: "bridge"      # Enable network access (default: none)
+    custom_tools:
+      - name: ["understand_image"]
+        category: "multimodal"
+        path: "massgen/tool/_multimodal_tools/understand_image.py"
+        function: ["understand_image"]
+    mcp_servers:
+        playwright:
+          type: "stdio"
+          command: "npx"
+          args: [
+            "@playwright/mcp@latest",
+            "--browser=chrome",  # Use Chrome browser
+            "--caps=vision,pdf", # Enable vision and PDF capabilities
+            "--user-data-dir=${cwd}/playwright-profile", # Persistent browser profile within workspace
+            "--output-dir=${cwd}", # Save screenshots/PDFs directly to workspace
+            # "--save-trace"       # Save Playwright traces for debugging
+          ]
+
+ui:
+  display_type: rich_terminal
+  logging_enabled: true
+orchestrator:
+  snapshot_storage: snapshots
+  agent_temporary_workspace: temp_workspaces
+  session_storage: sessions
+  # voting_sensitivity: balanced
+  max_new_answers_per_agent: 5
+  # answer_novelty_requirement: balanced

massgen/configs/tools/custom_tools/multimodal_tools/understand_audio.yaml

@@ -0,0 +1,33 @@
+# MassGen Configuration: Understand Audio Tool
+# Usage:
+#   massgen --config massgen/configs/tools/custom_tools/multimodal_tools/understand_audio.yaml "Please summarize the content in this audio."
+agents:
+  - id: "understand_audio_tool"
+    backend:
+      type: "openai"
+      model: "gpt-5-nano"
+      cwd: "workspace1"
+      custom_tools:
+        - name: ["understand_audio"]
+          category: "multimodal"
+          path: "massgen/tool/_multimodal_tools/understand_audio.py"
+          function: ["understand_audio"]
+    system_message: |
+      You are an AI assistant with access to audio transcription capabilities.
+
+      The understand_audio tool is available to transcribe audio files to text using OpenAI's Transcription API.
+      Do not output tool call syntax or function declarations. Focus on answering the user's question clearly.
+
+      When users ask about transcribing or understanding audio files, use the understand_audio tool to
+      process the audio and provide the transcription.
+
+orchestrator:
+  snapshot_storage: "snapshots"
+  agent_temporary_workspace: "temp_workspaces"
+  context_paths:
+    - path: "massgen/configs/resources/v0.1.3-example/Sherlock_Holmes.mp3"
+      permission: "read"
+
+ui:
+  display_type: "rich_terminal"
+  logging_enabled: true

massgen/configs/tools/custom_tools/multimodal_tools/understand_file.yaml

@@ -0,0 +1,34 @@
+# MassGen Configuration: Understand File Tool
+# Usage:
+#   massgen --config massgen/configs/tools/custom_tools/multimodal_tools/understand_file.yaml "Please summarize the content in this file."
+agents:
+  - id: "understand_file_tool"
+    backend:
+      type: "openai"
+      model: "gpt-5-nano"
+      cwd: "workspace1"
+      custom_tools:
+        - name: ["understand_file"]
+          category: "multimodal"
+          path: "massgen/tool/_multimodal_tools/understand_file.py"
+          function: ["understand_file"]
+    system_message: |
+      You are an AI assistant with access to file understanding capabilities.
+
+      The understand_file tool is available to analyze and understand file contents using OpenAI's gpt-4.1 API.
+      It supports text files, PDF, DOCX, XLSX, PPTX, and more.
+      Do not output tool call syntax or function declarations. Focus on answering the user's question clearly.
+
+      When users ask about analyzing or understanding files, use the understand_file tool to process
+      the file and provide detailed descriptions or answers to their questions.
+
+orchestrator:
+  snapshot_storage: "snapshots"
+  agent_temporary_workspace: "temp_workspaces"
+  context_paths:
+    - path: "massgen/configs/resources/v0.1.3-example/TUMIX.pdf"
+      permission: "read"
+
+ui:
+  display_type: "rich_terminal"
+  logging_enabled: true

massgen/configs/tools/custom_tools/multimodal_tools/understand_image.yaml

@@ -0,0 +1,33 @@
+# MassGen Configuration: Understand Image Tool
+# Usage:
+#   massgen --config massgen/configs/tools/custom_tools/multimodal_tools/understand_image.yaml "Please summarize the content in this image."
+agents:
+  - id: "understand_image_tool"
+    backend:
+      type: "openai"
+      model: "gpt-5-nano"
+      cwd: "workspace1"
+      custom_tools:
+        - name: ["understand_image"]
+          category: "multimodal"
+          path: "massgen/tool/_multimodal_tools/understand_image.py"
+          function: ["understand_image"]
+    system_message: |
+      You are an AI assistant with access to image understanding capabilities.
+
+      The understand_image tool is available to analyze and understand images using OpenAI's gpt-4.1 API.
+      Do not output tool call syntax or function declarations. Focus on answering the user's question clearly.
+
+      When users ask about analyzing or understanding images, use the understand_image tool to process
+      the image and provide detailed descriptions or answers to their questions.
+
+orchestrator:
+  snapshot_storage: "snapshots"
+  agent_temporary_workspace: "temp_workspaces"
+  context_paths:
+    - path: "massgen/configs/resources/v0.1.3-example/multimodality.jpg"
+      permission: "read"
+
+ui:
+  display_type: "rich_terminal"
+  logging_enabled: true

massgen/configs/tools/custom_tools/multimodal_tools/understand_video.yaml

@@ -0,0 +1,34 @@
+# MassGen Configuration: Understand Video Tool
+# Usage:
+#   massgen --config massgen/configs/tools/custom_tools/multimodal_tools/understand_video.yaml "What's happening in this video?"
+agents:
+  - id: "understand_video_tool"
+    backend:
+      type: "openai"
+      model: "gpt-5-nano"
+      cwd: "workspace1"
+      custom_tools:
+        - name: ["understand_video"]
+          category: "multimodal"
+          path: "massgen/tool/_multimodal_tools/understand_video.py"
+          function: ["understand_video"]
+    system_message: |
+      You are an AI assistant with access to video understanding capabilities.
+
+      The understand_video tool is available to analyze and understand videos by extracting key frames
+      and using OpenAI's gpt-4.1 API.
+      Do not output tool call syntax or function declarations. Focus on answering the user's question clearly.
+
+      When users ask about analyzing or understanding videos, use the understand_video tool to process
+      the video and provide detailed descriptions or answers to their questions.
+
+orchestrator:
+  snapshot_storage: "snapshots"
+  agent_temporary_workspace: "temp_workspaces"
+  context_paths:
+    - path: "massgen/configs/resources/v0.1.3-example/oppenheimer_trailer_1920.mp4"
+      permission: "read"
+
+ui:
+  display_type: "rich_terminal"
+  logging_enabled: true

massgen/configs/tools/custom_tools/multimodal_tools/understand_video_example.yaml

@@ -0,0 +1,54 @@
+# MassGen Configuration: Understand Video Example
+#
+# Use Case: Analyze a specific video file using the understand_video tool
+#
+# This demonstrates direct video analysis without needing to download.
+# The video file is provided as a context path for agents to analyze.
+#
+# Run with:
+#   uv run massgen --config massgen/configs/tools/custom_tools/multimodal_tools/understand_video_example.yaml "What is shown in this video?"
+
+agents:
+  - id: "agent_a"
+    backend:
+      type: "openai"
+      model: "gpt-5-mini"
+      text:
+        verbosity: "medium"
+      reasoning:
+        effort: "medium"
+        summary: "auto"
+      custom_tools:
+        - name: ["understand_video"]
+          category: "multimodal"
+          path: "massgen/tool/_multimodal_tools/understand_video.py"
+          function: ["understand_video"]
+      cwd: "workspace1"
+
+  - id: "agent_b"
+    backend:
+      type: "gemini"
+      model: "gemini-2.5-pro"
+      custom_tools:
+        - name: ["understand_video"]
+          category: "multimodal"
+          path: "massgen/tool/_multimodal_tools/understand_video.py"
+          function: ["understand_video"]
+      cwd: "workspace2"
+
+orchestrator:
+  snapshot_storage: "snapshots"
+  agent_temporary_workspace: "temp_workspaces"
+  context_paths:
+    - path: "massgen/configs/resources/v0.1.3-example/case-study-videos/Dp2oldJJImw.mp4"
+      permission: "read"
+
+ui:
+  display_type: "rich_terminal"
+  logging_enabled: true
+
+# What happens:
+# 1. Agents have read access to the video file
+# 2. They can use understand_video tool to analyze it
+# 3. Tool extracts 8 frames and analyzes with GPT-4.1
+# 4. Agents collaborate to provide comprehensive insights

massgen/configs/tools/custom_tools/multimodal_tools/youtube_video_analysis.yaml

@@ -0,0 +1,59 @@
+# MassGen Configuration: YouTube Video Analysis with Multimodal Understanding
+#
+# Use Case: Download and analyze YouTube videos from MassGen case studies
+#
+# This demonstrates MassGen's self-evolution capabilities by having agents:
+# 1. Read local case study documentation to discover video URLs
+# 2. Download YouTube videos using yt-dlp via command-line execution
+# 3. Analyze video content using the understand_video multimodal tool
+# 4. Extract insights that could inform future feature development
+#
+# Run with:
+#   uv run massgen --config massgen/configs/tools/custom_tools/multimodal_tools/youtube_video_analysis.yaml "Download recent MassGen case study videos listed in the case study md files, analyze them, find out how to improve them and automate their creation."
+
+agents:
+  - id: "agent_a"
+    backend:
+      type: "openai"
+      model: "gpt-5-mini"
+      text:
+        verbosity: "medium"
+      reasoning:
+        effort: "medium"
+        summary: "auto"
+      custom_tools:
+        - name: ["understand_video"]
+          category: "multimodal"
+          path: "massgen/tool/_multimodal_tools/understand_video.py"
+          function: ["understand_video"]
+      enable_mcp_command_line: true
+      command_line_execution_mode: docker
+      command_line_docker_enable_sudo: true
+      command_line_docker_network_mode: "bridge"
+      cwd: "workspace1"
+
+  - id: "agent_b"
+    backend:
+      type: "claude_code"
+      model: "claude-sonnet-4-5-20250929"
+      custom_tools:
+        - name: ["understand_video"]
+          category: "multimodal"
+          path: "massgen/tool/_multimodal_tools/understand_video.py"
+          function: ["understand_video"]
+      enable_mcp_command_line: true
+      command_line_execution_mode: docker
+      command_line_docker_enable_sudo: true
+      command_line_docker_network_mode: "bridge"
+      cwd: "workspace2"
+
+orchestrator:
+  snapshot_storage: "snapshots"
+  agent_temporary_workspace: "temp_workspaces"
+  context_paths:
+    - path: "docs/case_studies"
+      permission: "read"
+
+ui:
+  display_type: "rich_terminal"
+  logging_enabled: true