claude-mpm 4.21.0__py3-none-any.whl → 4.21.3__py3-none-any.whl

claude_mpm/VERSION

@@ -1 +1 @@
-4.21.0
+4.21.3

claude_mpm/commands/mpm-help.md

@@ -87,6 +87,9 @@ Available Commands:
 /mpm-init [update]
   Initialize or update project documentation
 
+/mpm-resume
+  Create session resume files for easy work resumption
+
 /mpm-monitor [start|stop|restart|status|port]
   Manage Socket.IO monitoring server and dashboard
 

claude_mpm/commands/mpm-resume.md

@@ -0,0 +1,372 @@
+# /mpm-resume - Load Previous Session
+
+Load and display context from the most recent paused session to seamlessly continue your work.
+
+## Usage
+
+```
+/mpm-resume
+```
+
+## Description
+
+This command **loads and displays** the most recent paused session from automatic session saves, allowing you to resume work with full context restoration.
+
+Unlike `/mpm-init pause` which *creates* a new session save, this command *loads* existing session data that was automatically created when context usage reached 70% (140k/200k tokens).
+
+**Key Points:**
+- Reads from `.claude-mpm/sessions/` directory (automatically created at 70% context)
+- Displays up to ~40,000 tokens (~20% of 200k context budget)
+- Shows session summary, completed work, in-progress tasks, and git context
+- Does NOT create any new files - only reads and displays
+
+## What Gets Displayed
+
+When you run `/mpm-resume`, PM will display:
+
+### Session Summary
+- **Time Elapsed**: How long ago the session was paused
+- **Context Usage**: Token usage at time of pause (e.g., "67.6% - 135,259/200,000 tokens")
+- **Working On**: What you were working on when paused
+- **Session Duration**: How long the previous session lasted
+
+### Completed Work
+- List of accomplishments from the paused session
+- What was delivered and committed
+- Key milestones achieved
+
+### Current Tasks
+- **In Progress**: Tasks that were being worked on
+- **Pending**: Tasks that were planned but not started
+- **Completed**: Recently finished tasks for context
+
+### Git Context
+- **Branch**: Current branch name
+- **Recent Commits**: Last 5-10 commits with SHAs and messages
+- **Status**: Clean/dirty working directory
+- **Changes Since Pause**: New commits made since session was saved
+
+### Next Recommended Actions
+- Priority-ordered list of next steps
+- Estimated time for each task
+- Status and blockers for pending work
+
+## Session Storage Location
+
+Sessions are automatically saved to:
+```
+<project-root>/.claude-mpm/sessions/session-YYYYMMDD-HHMMSS.json
+```
+
+**Legacy location** (backward compatible):
+```
+<project-root>/.claude-mpm/sessions/pause/session-YYYYMMDD-HHMMSS.json
+```
+
+The system automatically checks both locations and uses the most recent session.
+
+## Example Output
+
+```
+================================================================================
+📋 PAUSED SESSION FOUND
+================================================================================
+
+Paused: 2 hours ago
+
+Last working on: Week 2 Skills Integration - Content Preparation
+
+Completed:
+  ✓ Week 1: Complete infrastructure - 8,900 lines of production-ready code
+  ✓ Week 1: Skills loading system with automatic progressive disclosure
+  ✓ Week 2: 15 of 23 skills downloaded (65% complete)
+  ✓ Week 2: 2 Tier 1 skills refactored to progressive disclosure
+  ✓ Code quality: All CRITICAL and HIGH issues resolved
+
+Next steps:
+  • Refactor Tier 2 skills (verification-before-completion, webapp-testing)
+  • Download remaining 8 skills from community repositories
+  • Refactor remaining 13 skills to progressive disclosure
+  • Generate license attributions for all bundled skills
+
+Git changes since pause: 3 commits
+
+Recent commits:
+  ac765731 - feat(skills): Week 2 progress - 15 skills downloaded (Bob Matsuoka)
+  205e532e - fix(skills): address CRITICAL and HIGH priority issues (Bob Matsuoka)
+  06a6d6a0 - feat: add automated pre-publish cleanup to release (Bob Matsuoka)
+
+Context Usage: 67.6% (135,259/200,000 tokens)
+Session Duration: 12 hours
+
+================================================================================
+Use this context to resume work, or start fresh if not relevant.
+================================================================================
+```
+
+## Implementation
+
+When PM receives `/mpm-resume`, it should:
+
+1. **Check for Sessions**
+   ```python
+   from claude_mpm.services.cli.session_resume_helper import SessionResumeHelper
+
+   helper = SessionResumeHelper()
+   if not helper.has_paused_sessions():
+       return "No paused sessions found"
+   ```
+
+2. **Load Most Recent Session**
+   ```python
+   session_data = helper.get_most_recent_session()
+   if not session_data:
+       return "Failed to load session data"
+   ```
+
+3. **Format and Display Context**
+   ```python
+   # Extract key information
+   conversation = session_data.get("conversation", {})
+   git_context = session_data.get("git_context", {})
+   context_usage = session_data.get("context_usage", {})
+   todos = session_data.get("todos", {})
+
+   # Display formatted output (see Example Output above)
+   ```
+
+4. **Calculate Git Changes**
+   ```python
+   # Get commits since pause
+   paused_at = session_data.get("paused_at")
+   new_commits = helper.get_git_changes_since_pause(paused_at, [])
+   ```
+
+5. **Limit Output to ~40k Tokens**
+   - Session summary: ~2k tokens
+   - Accomplishments (first 10): ~3k tokens
+   - Next steps (first 10): ~3k tokens
+   - Git context: ~5k tokens
+   - Todos: ~2k tokens
+   - Recent commits (up to 10): ~5k tokens
+   - **Total**: ~20k tokens (well under 40k limit)
+
+## Token Budget Management
+
+**Context Budget**: 200,000 tokens total
+**Resume Load**: ~20,000-40,000 tokens (10-20% of context)
+
+This leaves 160,000+ tokens for actual work after loading session context.
+
+## Session Data Format
+
+Sessions are stored as JSON with this structure:
+
+```json
+{
+  "session_id": "session-YYYYMMDD-HHMMSS",
+  "paused_at": "ISO-8601 timestamp",
+  "duration_hours": 12,
+  "context_usage": {
+    "tokens_used": 135259,
+    "tokens_total": 200000,
+    "percentage": 67.6
+  },
+  "conversation": {
+    "primary_task": "What user was working on",
+    "current_phase": "Current phase of work",
+    "summary": "Brief summary",
+    "accomplishments": ["list of completed items"],
+    "next_steps": [
+      {
+        "priority": 1,
+        "task": "Task description",
+        "estimated_hours": "8-12",
+        "status": "ready"
+      }
+    ]
+  },
+  "git_context": {
+    "branch": "main",
+    "recent_commits": [...],
+    "status": {...}
+  },
+  "todos": {
+    "active": [...],
+    "completed": [...]
+  }
+}
+```
+
+## When to Use This Command
+
+Use `/mpm-resume` when:
+- **Starting a new session**: After closing and reopening Claude CLI
+- **Context unclear**: You need to remember what you were working on
+- **After a break**: Coming back after hours or days
+- **Team handoff**: Another developer wants to understand current state
+- **Lost context**: Accidentally closed CLI and need to recover
+
+## Differences from Automatic Resume
+
+| Feature | Automatic Resume (70% context) | /mpm-resume Command |
+|---------|-------------------------------|---------------------|
+| **Trigger** | Automatic at 70% context | Manual user command |
+| **When** | PM startup (if session exists) | Anytime during session |
+| **Creates Files** | No (reads existing) | No (reads existing) |
+| **Session Source** | Same (`.claude-mpm/sessions/`) | Same (`.claude-mpm/sessions/`) |
+| **Display Format** | Identical | Identical |
+| **Token Usage** | ~20-40k tokens | ~20-40k tokens |
+
+Both features use the **same underlying system** (`SessionResumeHelper`), just triggered differently.
+
+## No Files Created
+
+**IMPORTANT**: This command does NOT create any new files.
+
+It ONLY reads from existing session files that were automatically created by the system at 70% context usage.
+
+If you want to manually create a session save (for example, at 50% context before hitting 70%), use a different workflow or wait for automatic save at 70%.
+
+## Related Features
+
+- **Automatic Session Save**: System creates sessions at 70% context automatically
+- **Automatic Session Resume**: PM startup hook displays sessions automatically
+- `/mpm-init pause`: Manual session pause workflow (if available)
+- `/mpm-init context`: Analyze git history for intelligent resumption
+- `/mpm-status`: Check current MPM status
+
+## Error Handling
+
+### No Sessions Found
+```
+No paused sessions found in .claude-mpm/sessions/
+
+To create a session save, continue working until context reaches 70% (140k tokens),
+at which point the system will automatically save your session state.
+```
+
+### Failed to Load Session
+```
+Paused session file found but failed to load.
+
+File: .claude-mpm/sessions/session-20251107-152740.json
+Error: Invalid JSON format
+
+You may need to manually inspect or delete this file.
+```
+
+### Invalid Session Format
+```
+Session file loaded but missing required fields.
+
+The session file may be corrupted or from an older version.
+Consider running /mpm-doctor to check system health.
+```
+
+## Benefits
+
+- **Instant Context**: Get full context in seconds without reading git logs
+- **No Mental Load**: Don't need to remember what you were doing
+- **Zero File Creation**: Pure read operation, no side effects
+- **Team Collaboration**: Share context with team members
+- **Graceful Recovery**: Recover from accidental CLI closures
+- **Smart Filtering**: Only shows relevant information (~20k tokens)
+- **Git Awareness**: See what changed since pause
+
+## Best Practices
+
+1. **Use Early**: Run `/mpm-resume` at start of each session if sessions exist
+2. **Check Git Changes**: Pay attention to commits made since pause
+3. **Validate Context**: Verify the session is still relevant before continuing
+4. **Clear Old Sessions**: Periodically clean up old session files
+5. **Combine with Git**: Use alongside `git log` for complete picture
+
+## Technical Details
+
+**Implementation Files:**
+- Service: `/src/claude_mpm/services/cli/session_resume_helper.py`
+- Hook: `/src/claude_mpm/hooks/session_resume_hook.py`
+- Command: This file
+
+**Key Functions:**
+- `SessionResumeHelper.has_paused_sessions()` - Check if sessions exist
+- `SessionResumeHelper.get_most_recent_session()` - Load latest session
+- `SessionResumeHelper.format_resume_prompt()` - Format display output
+- `SessionResumeHelper.get_git_changes_since_pause()` - Calculate git delta
+
+**Token Estimation:**
+- Session metadata: 1-2k tokens
+- Accomplishments (10 items): 2-4k tokens
+- Next steps (10 items): 2-4k tokens
+- Git commits (10 commits): 3-5k tokens
+- Todos (20 items): 2-3k tokens
+- Formatting/structure: 1-2k tokens
+- **Total**: 11-20k tokens (safely under 40k limit)
+
+## Troubleshooting
+
+### Session Not Found
+**Problem**: Command reports no sessions exist
+
+**Solutions:**
+1. Check directory exists: `ls .claude-mpm/sessions/`
+2. Check for legacy location: `ls .claude-mpm/sessions/pause/`
+3. Verify session files: `ls .claude-mpm/sessions/session-*.json`
+4. Session auto-saves at 70% context - may not exist yet
+
+### Git Changes Not Showing
+**Problem**: "No git changes since pause" but commits were made
+
+**Solutions:**
+1. Verify git repository: `git status`
+2. Check commit timestamps: `git log --since="<pause_time>"`
+3. Ensure session timestamp is correct
+4. Check timezone issues
+
+### Display Too Large
+**Problem**: Session context exceeds token budget
+
+**Solutions:**
+1. System automatically limits to first 10 items
+2. Full session details available in JSON file
+3. Use `cat .claude-mpm/sessions/session-*.json` for complete data
+4. Summary is optimized for 20k tokens max
+
+## Example Session Resume Workflow
+
+```bash
+# User starts new Claude CLI session
+$ claude-code
+
+# PM automatically checks for sessions on startup
+# (Automatic resume hook displays session if found)
+
+# OR user manually requests resume
+User: "/mpm-resume"
+
+# PM loads and displays session context
+PM: [Displays formatted session context as shown in Example Output]
+
+# User decides to continue work
+User: "Let's continue with the next priority task"
+
+# PM uses session context to understand what to do next
+PM: "Based on the paused session, the next priority is to refactor
+     the verification-before-completion skill. I'll delegate this to Engineer..."
+```
+
+## Version History
+
+- **v4.21.1**: Fixed command behavior - now loads sessions instead of creating files
+- **v4.21.0**: Added `/mpm-resume` command (incorrect behavior - created files)
+- **v4.19.0**: Automatic session resume infrastructure implemented
+- **v4.18.x**: Session pause/resume foundation
+
+## Support
+
+For issues or questions:
+- Run `/mpm-doctor` to check system health
+- Check logs: `.claude-mpm/logs/claude-mpm.log`
+- Verify session files: `ls -la .claude-mpm/sessions/`
+- Review documentation: `/docs/features/session-auto-resume.md`

claude_mpm/commands/mpm.md

@@ -8,6 +8,7 @@ Available MPM commands:
 - /mpm-help - Show command help
 - /mpm-status - Show MPM status
 - /mpm-config - Manage configuration
+- /mpm-resume - Create session resume files
 - /mpm-version - Display version information for project, agents, and skills
 
 Claude MPM extends Claude Code with:

claude_mpm/services/core/base.py

@@ -9,6 +9,7 @@ and lifecycle management.
 Part of TSK-0046: Service Layer Architecture Reorganization
 """
 
+import threading
 from abc import ABC, abstractmethod
 from typing import Any, Dict, Optional
 
@@ -221,29 +222,43 @@ class SingletonService(SyncBaseService):
     """
     Base class for singleton services.
 
-    Ensures only one instance of the service exists.
+    Ensures only one instance of the service exists with thread-safe initialization.
+    Uses double-checked locking pattern to prevent race conditions.
     """
 
     _instances: Dict[type, "SingletonService"] = {}
+    _lock = threading.Lock()
 
     def __new__(cls, *args, **kwargs):
-        """Ensure only one instance exists."""
+        """Ensure only one instance exists with thread-safe initialization."""
+        # Fast path - check without lock
         if cls not in cls._instances:
-            cls._instances[cls] = super().__new__(cls)
+            # Slow path - acquire lock and double-check
+            with cls._lock:
+                if cls not in cls._instances:
+                    cls._instances[cls] = super().__new__(cls)
         return cls._instances[cls]
 
     @classmethod
     def get_instance(cls) -> "SingletonService":
-        """Get the singleton instance."""
+        """Get the singleton instance with thread-safe initialization."""
+        # Fast path - check without lock
         if cls not in cls._instances:
-            cls._instances[cls] = cls()
+            # Slow path - acquire lock and double-check
+            with cls._lock:
+                if cls not in cls._instances:
+                    cls._instances[cls] = cls()
         return cls._instances[cls]
 
     @classmethod
     def clear_instance(cls) -> None:
-        """Clear the singleton instance (useful for testing)."""
-        if cls in cls._instances:
-            instance = cls._instances[cls]
-            if hasattr(instance, "shutdown") and not instance.is_shutdown:
-                instance.shutdown()
-            del cls._instances[cls]
+        """Clear the singleton instance (useful for testing).
+
+        Thread-safe implementation ensures proper cleanup.
+        """
+        with cls._lock:
+            if cls in cls._instances:
+                instance = cls._instances[cls]
+                if hasattr(instance, "shutdown") and not instance.is_shutdown:
+                    instance.shutdown()
+                del cls._instances[cls]

claude_mpm/services/event_bus/relay.py

@@ -9,6 +9,7 @@ WHY separate relay component:
 """
 
 import os
+import threading
 import time
 from datetime import datetime, timezone
 from typing import Any, Dict, Optional
@@ -271,11 +272,15 @@ class SocketIORelay:
 
 # Global relay instance
 _relay_instance: Optional[SocketIORelay] = None
+_relay_lock = threading.Lock()
 
 
 def get_relay(port: Optional[int] = None) -> SocketIORelay:
     """Get or create the global SocketIO relay instance.
 
+    Thread-safe implementation using double-checked locking pattern to
+    prevent race conditions during concurrent initialization.
+
     Args:
         port: Optional port number
 
@@ -283,9 +288,16 @@ def get_relay(port: Optional[int] = None) -> SocketIORelay:
         SocketIORelay: The relay instance
     """
     global _relay_instance
-    if _relay_instance is None:
-        _relay_instance = SocketIORelay(port)
-    return _relay_instance
+
+    # Fast path - check without lock
+    if _relay_instance is not None:
+        return _relay_instance
+
+    # Slow path - acquire lock and double-check
+    with _relay_lock:
+        if _relay_instance is None:
+            _relay_instance = SocketIORelay(port)
+        return _relay_instance
 
 
 def start_relay(port: Optional[int] = None) -> SocketIORelay:
@@ -303,8 +315,12 @@ def start_relay(port: Optional[int] = None) -> SocketIORelay:
 
 
 def stop_relay() -> None:
-    """Stop the global SocketIO relay."""
+    """Stop the global SocketIO relay.
+
+    Thread-safe implementation ensures proper cleanup.
+    """
     global _relay_instance
-    if _relay_instance:
-        _relay_instance.stop()
-        _relay_instance = None
+    with _relay_lock:
+        if _relay_instance:
+            _relay_instance.stop()
+            _relay_instance = None

claude_mpm/services/memory/failure_tracker.py

@@ -29,6 +29,7 @@ Example flow:
 
 import logging
 import re
+import threading
 from dataclasses import dataclass, field
 from datetime import datetime, timezone
 from typing import Dict, List, Optional, Tuple
@@ -536,6 +537,7 @@ class FailureTracker:
 
 # Singleton instance for session-level tracking
 _tracker_instance: Optional[FailureTracker] = None
+_tracker_lock = threading.Lock()
 
 
 def get_failure_tracker() -> FailureTracker:
@@ -544,13 +546,23 @@ def get_failure_tracker() -> FailureTracker:
     WHY: Session-level tracking requires a singleton to maintain state
     across multiple hook invocations during the same session.
 
+    Thread-safe implementation using double-checked locking pattern to
+    prevent race conditions during concurrent initialization.
+
     Returns:
         The FailureTracker singleton instance
     """
     global _tracker_instance
-    if _tracker_instance is None:
-        _tracker_instance = FailureTracker()
-    return _tracker_instance
+
+    # Fast path - check without lock
+    if _tracker_instance is not None:
+        return _tracker_instance
+
+    # Slow path - acquire lock and double-check
+    with _tracker_lock:
+        if _tracker_instance is None:
+            _tracker_instance = FailureTracker()
+        return _tracker_instance
 
 
 def reset_failure_tracker() -> None:
@@ -558,6 +570,9 @@ def reset_failure_tracker() -> None:
 
     WHY: Tests need to reset state between runs. Also useful for
     explicitly starting a new tracking session.
+
+    Thread-safe implementation ensures proper cleanup.
     """
     global _tracker_instance
-    _tracker_instance = None
+    with _tracker_lock:
+        _tracker_instance = None

claude_mpm/tools/code_tree_analyzer/__init__.py

@@ -0,0 +1,45 @@
+#!/usr/bin/env python3
+"""
+Code Tree Analyzer
+==================
+
+Analyzes source code using AST to extract structure and metrics,
+supporting multiple languages and emitting incremental events for visualization.
+
+This module has been refactored from a single 1,825-line file into focused,
+maintainable components while preserving complete backward compatibility.
+
+Public API:
+-----------
+- CodeTreeAnalyzer: Main analyzer class
+- CodeNode: Data structure for code nodes
+- GitignoreManager: Gitignore pattern matching
+- PythonAnalyzer: Python AST analysis
+- MultiLanguageAnalyzer: Multi-language support
+
+Example Usage:
+--------------
+    from claude_mpm.tools.code_tree_analyzer import CodeTreeAnalyzer
+
+    analyzer = CodeTreeAnalyzer(emit_events=True)
+    result = analyzer.analyze_directory(Path("/path/to/code"))
+    print(result['stats'])
+"""
+
+# Public API - Backward compatible imports
+from .core import CodeTreeAnalyzer
+from .gitignore import GitignoreManager
+from .models import CodeNode
+from .multilang_analyzer import MultiLanguageAnalyzer
+from .python_analyzer import PythonAnalyzer
+
+__all__ = [
+    "CodeNode",
+    "CodeTreeAnalyzer",
+    "GitignoreManager",
+    "MultiLanguageAnalyzer",
+    "PythonAnalyzer",
+]
+
+# Version info
+__version__ = "2.0.0"  # Major refactoring while maintaining API compatibility