claude-mpm 4.15.6__py3-none-any.whl → 4.21.3__py3-none-any.whl

claude_mpm/services/infrastructure/resume_log_generator.py

@@ -0,0 +1,439 @@
+"""Resume Log Generator Service.
+
+Automatically generates session resume logs when approaching or hitting token limits.
+Integrates with session management and response tracking infrastructure.
+
+Triggers:
+- model_context_window_exceeded (stop_reason)
+- Manual pause command
+- 95% token threshold reached
+- Session end with high token usage (>85%)
+
+Design Principles:
+- Atomic file operations (via state_storage)
+- Non-blocking generation
+- Graceful degradation if generation fails
+- Integration with existing session state
+"""
+
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any, Dict, List, Optional
+
+from claude_mpm.core.logging_utils import get_logger
+from claude_mpm.models.resume_log import ContextMetrics, ResumeLog
+from claude_mpm.storage.state_storage import StateStorage
+
+logger = get_logger(__name__)
+
+
+class ResumeLogGenerator:
+    """Service for generating session resume logs."""
+
+    def __init__(
+        self,
+        storage_dir: Optional[Path] = None,
+        config: Optional[Dict[str, Any]] = None,
+    ):
+        """Initialize resume log generator.
+
+        Args:
+            storage_dir: Directory for resume logs (default: .claude-mpm/resume-logs)
+            config: Configuration dictionary
+        """
+        self.storage_dir = storage_dir or Path.home() / ".claude-mpm" / "resume-logs"
+        self.storage_dir.mkdir(parents=True, exist_ok=True)
+
+        # State storage for atomic writes
+        self.state_storage = StateStorage(
+            storage_dir=self.storage_dir.parent / "storage"
+        )
+
+        # Configuration
+        self.config = config or {}
+        self.enabled = (
+            self.config.get("context_management", {})
+            .get("resume_logs", {})
+            .get("enabled", True)
+        )
+        self.auto_generate = (
+            self.config.get("context_management", {})
+            .get("resume_logs", {})
+            .get("auto_generate", True)
+        )
+        self.max_tokens = (
+            self.config.get("context_management", {})
+            .get("resume_logs", {})
+            .get("max_tokens", 10000)
+        )
+
+        # Trigger thresholds
+        thresholds = self.config.get("context_management", {}).get("thresholds", {})
+        self.threshold_caution = thresholds.get("caution", 0.70)
+        self.threshold_warning = thresholds.get("warning", 0.85)
+        self.threshold_critical = thresholds.get("critical", 0.95)
+
+        logger.info(
+            f"ResumeLogGenerator initialized (enabled={self.enabled}, auto_generate={self.auto_generate})"
+        )
+
+    def should_generate(
+        self,
+        stop_reason: Optional[str] = None,
+        token_usage_pct: Optional[float] = None,
+        manual_trigger: bool = False,
+    ) -> bool:
+        """Determine if resume log should be generated.
+
+        Args:
+            stop_reason: Claude API stop_reason
+            token_usage_pct: Current token usage percentage (0.0-1.0)
+            manual_trigger: Manual pause/stop command
+
+        Returns:
+            True if resume log should be generated
+        """
+        if not self.enabled or not self.auto_generate:
+            return manual_trigger  # Only generate on manual trigger if auto is disabled
+
+        # Trigger conditions
+        triggers = [
+            stop_reason == "max_tokens",
+            stop_reason == "model_context_window_exceeded",
+            manual_trigger,
+            token_usage_pct and token_usage_pct >= self.threshold_critical,
+            token_usage_pct
+            and token_usage_pct >= self.threshold_warning,  # Generate at 85% too
+        ]
+
+        should_gen = any(triggers)
+
+        if should_gen:
+            reason = "unknown"
+            if stop_reason:
+                reason = f"stop_reason={stop_reason}"
+            elif manual_trigger:
+                reason = "manual_trigger"
+            elif token_usage_pct:
+                reason = f"token_usage={token_usage_pct:.1%}"
+
+            logger.info(f"Resume log generation triggered: {reason}")
+
+        return should_gen
+
+    def generate_from_session_state(
+        self,
+        session_id: str,
+        session_state: Dict[str, Any],
+        stop_reason: Optional[str] = None,
+    ) -> Optional[ResumeLog]:
+        """Generate resume log from session state data.
+
+        Args:
+            session_id: Current session ID
+            session_state: Session state dictionary
+            stop_reason: Claude API stop_reason
+
+        Returns:
+            Generated ResumeLog or None if generation failed
+        """
+        try:
+            # Extract context metrics
+            context_data = session_state.get("context_metrics", {})
+            context_metrics = ContextMetrics(
+                total_budget=context_data.get("total_budget", 200000),
+                used_tokens=context_data.get("used_tokens", 0),
+                remaining_tokens=context_data.get("remaining_tokens", 0),
+                percentage_used=context_data.get("percentage_used", 0.0),
+                stop_reason=stop_reason or context_data.get("stop_reason"),
+                model=context_data.get("model", "claude-sonnet-4.5"),
+                session_id=session_id,
+            )
+
+            # Extract content from session state
+            mission_summary = session_state.get("mission_summary", "")
+            accomplishments = session_state.get("accomplishments", [])
+            key_findings = session_state.get("key_findings", [])
+            decisions_made = session_state.get("decisions_made", [])
+            next_steps = session_state.get("next_steps", [])
+            critical_context = session_state.get("critical_context", {})
+
+            # Extract metadata
+            files_modified = session_state.get("files_modified", [])
+            agents_used = session_state.get("agents_used", {})
+            errors_encountered = session_state.get("errors_encountered", [])
+            warnings = session_state.get("warnings", [])
+
+            # Create resume log
+            resume_log = ResumeLog(
+                session_id=session_id,
+                previous_session_id=session_state.get("previous_session_id"),
+                context_metrics=context_metrics,
+                mission_summary=mission_summary,
+                accomplishments=accomplishments,
+                key_findings=key_findings,
+                decisions_made=decisions_made,
+                next_steps=next_steps,
+                critical_context=critical_context,
+                files_modified=files_modified,
+                agents_used=agents_used,
+                errors_encountered=errors_encountered,
+                warnings=warnings,
+            )
+
+            logger.info(f"Generated resume log for session {session_id}")
+            return resume_log
+
+        except Exception as e:
+            logger.error(
+                f"Failed to generate resume log from session state: {e}", exc_info=True
+            )
+            return None
+
+    def generate_from_todo_list(
+        self,
+        session_id: str,
+        todos: List[Dict[str, Any]],
+        context_metrics: Optional[ContextMetrics] = None,
+    ) -> Optional[ResumeLog]:
+        """Generate resume log from TODO list.
+
+        Useful when session state is minimal but TODO list has rich information.
+
+        Args:
+            session_id: Current session ID
+            todos: TODO list items
+            context_metrics: Context metrics (optional)
+
+        Returns:
+            Generated ResumeLog or None if generation failed
+        """
+        try:
+            # Categorize todos
+            completed = [t for t in todos if t.get("status") == "completed"]
+            in_progress = [t for t in todos if t.get("status") == "in_progress"]
+            pending = [t for t in todos if t.get("status") == "pending"]
+
+            # Build accomplishments from completed tasks
+            accomplishments = [f"✓ {task['content']}" for task in completed]
+
+            # Build next steps from in-progress and pending
+            next_steps = []
+            for task in in_progress:
+                next_steps.append(f"[IN PROGRESS] {task['content']}")
+            for task in pending:
+                next_steps.append(f"[PENDING] {task['content']}")
+
+            # Create mission summary
+            mission_summary = f"Working on {len(todos)} tasks: {len(completed)} completed, {len(in_progress)} in progress, {len(pending)} pending."
+
+            # Use provided context metrics or create default
+            if context_metrics is None:
+                context_metrics = ContextMetrics(session_id=session_id)
+
+            # Create resume log
+            resume_log = ResumeLog(
+                session_id=session_id,
+                context_metrics=context_metrics,
+                mission_summary=mission_summary,
+                accomplishments=accomplishments,
+                next_steps=next_steps,
+                critical_context={
+                    "total_tasks": len(todos),
+                    "completed_tasks": len(completed),
+                    "in_progress_tasks": len(in_progress),
+                    "pending_tasks": len(pending),
+                },
+            )
+
+            logger.info(f"Generated resume log from TODO list for session {session_id}")
+            return resume_log
+
+        except Exception as e:
+            logger.error(
+                f"Failed to generate resume log from TODO list: {e}", exc_info=True
+            )
+            return None
+
+    def save_resume_log(self, resume_log: ResumeLog) -> Optional[Path]:
+        """Save resume log to storage.
+
+        Args:
+            resume_log: ResumeLog instance to save
+
+        Returns:
+            Path to saved file or None if save failed
+        """
+        try:
+            # Save as markdown (primary format)
+            md_path = resume_log.save(storage_dir=self.storage_dir)
+
+            # Also save as JSON for programmatic access
+            json_path = self.storage_dir / f"session-{resume_log.session_id}.json"
+            self.state_storage.write_json(
+                data=resume_log.to_dict(),
+                file_path=json_path,
+                atomic=True,
+            )
+
+            logger.info(f"Resume log saved: {md_path}")
+            return md_path
+
+        except Exception as e:
+            logger.error(f"Failed to save resume log: {e}", exc_info=True)
+            return None
+
+    def load_resume_log(self, session_id: str) -> Optional[str]:
+        """Load resume log markdown content.
+
+        Args:
+            session_id: Session ID to load
+
+        Returns:
+            Markdown content or None if not found
+        """
+        try:
+            md_path = self.storage_dir / f"session-{session_id}.md"
+
+            if not md_path.exists():
+                logger.debug(f"Resume log not found for session {session_id}")
+                return None
+
+            content = md_path.read_text(encoding="utf-8")
+            logger.info(f"Loaded resume log for session {session_id}")
+            return content
+
+        except Exception as e:
+            logger.error(f"Failed to load resume log: {e}", exc_info=True)
+            return None
+
+    def list_resume_logs(self) -> List[Dict[str, Any]]:
+        """List all available resume logs.
+
+        Returns:
+            List of resume log metadata
+        """
+        try:
+            logs = []
+
+            for md_file in self.storage_dir.glob("session-*.md"):
+                # Extract session ID from filename
+                session_id = md_file.stem.replace("session-", "")
+
+                # Check if JSON metadata exists
+                json_file = md_file.with_suffix(".json")
+                metadata = {}
+                if json_file.exists():
+                    json_data = self.state_storage.read_json(json_file)
+                    if json_data:
+                        metadata = {
+                            "session_id": session_id,
+                            "created_at": json_data.get("created_at"),
+                            "previous_session_id": json_data.get("previous_session_id"),
+                            "context_metrics": json_data.get("context_metrics", {}),
+                            "file_path": str(md_file),
+                        }
+
+                if metadata:
+                    logs.append(metadata)
+                else:
+                    # Fallback to file metadata
+                    logs.append(
+                        {
+                            "session_id": session_id,
+                            "file_path": str(md_file),
+                            "modified_at": datetime.fromtimestamp(
+                                md_file.stat().st_mtime, tz=timezone.utc
+                            ).isoformat(),
+                        }
+                    )
+
+            # Sort by creation time (newest first)
+            logs.sort(
+                key=lambda x: x.get("created_at", x.get("modified_at", "")),
+                reverse=True,
+            )
+
+            logger.debug(f"Found {len(logs)} resume logs")
+            return logs
+
+        except Exception as e:
+            logger.error(f"Failed to list resume logs: {e}", exc_info=True)
+            return []
+
+    def cleanup_old_logs(self, keep_count: int = 10) -> int:
+        """Clean up old resume logs, keeping only the most recent.
+
+        Args:
+            keep_count: Number of logs to keep
+
+        Returns:
+            Number of logs deleted
+        """
+        try:
+            logs = self.list_resume_logs()
+
+            if len(logs) <= keep_count:
+                logger.debug(
+                    f"No cleanup needed ({len(logs)} logs <= {keep_count} keep)"
+                )
+                return 0
+
+            # Delete old logs
+            deleted = 0
+            for log in logs[keep_count:]:
+                try:
+                    md_path = Path(log["file_path"])
+                    json_path = md_path.with_suffix(".json")
+
+                    if md_path.exists():
+                        md_path.unlink()
+                        deleted += 1
+
+                    if json_path.exists():
+                        json_path.unlink()
+
+                except Exception as e:
+                    logger.warning(f"Failed to delete log {log['session_id']}: {e}")
+
+            logger.info(f"Cleaned up {deleted} old resume logs (kept {keep_count})")
+            return deleted
+
+        except Exception as e:
+            logger.error(f"Failed to cleanup old logs: {e}", exc_info=True)
+            return 0
+
+    def get_stats(self) -> Dict[str, Any]:
+        """Get resume log statistics.
+
+        Returns:
+            Dictionary with statistics
+        """
+        try:
+            logs = self.list_resume_logs()
+
+            total_size = 0
+            for log in logs:
+                path = Path(log["file_path"])
+                if path.exists():
+                    total_size += path.stat().st_size
+
+            return {
+                "enabled": self.enabled,
+                "auto_generate": self.auto_generate,
+                "total_logs": len(logs),
+                "storage_dir": str(self.storage_dir),
+                "total_size_kb": round(total_size / 1024, 2),
+                "thresholds": {
+                    "caution": f"{self.threshold_caution:.0%}",
+                    "warning": f"{self.threshold_warning:.0%}",
+                    "critical": f"{self.threshold_critical:.0%}",
+                },
+            }
+
+        except Exception as e:
+            logger.error(f"Failed to get stats: {e}", exc_info=True)
+            return {
+                "enabled": self.enabled,
+                "error": str(e),
+            }

claude_mpm/services/local_ops/__init__.py

@@ -86,6 +86,7 @@ from claude_mpm.services.core.models.process import (
     PROTECTED_PORT_RANGES,
     DeploymentState,
     ProcessInfo,
+    ProcessStatus,
     StartConfig,
     is_port_protected,
 )
@@ -147,6 +148,7 @@ __all__ = [
     "PortConflictError",
     "ProcessInfo",
     "ProcessSpawnError",
+    "ProcessStatus",
     # Data models - Process
     "ResourceMonitor",
     "ResourceUsage",

claude_mpm/services/mcp_config_manager.py

@@ -1130,139 +1130,15 @@ class MCPConfigManager:
         """
         Detect and fix corrupted MCP service installations.
 
-        This method:
-        1. Tests each MCP service for import/execution errors
-        2. Automatically reinstalls corrupted services
-        3. Fixes missing dependencies (like mcp-ticketer's gql)
-        4. Validates fixes worked
+        NOTE: Proactive health checking has been disabled.
+        Each MCP service should stand on its own and handle its own issues.
+        This function now only returns success without checking services.
 
         Returns:
             Tuple of (success, message)
         """
-        services_to_fix = []
-        fixed_services = []
-        failed_services = []
-
-        # Check each service for issues
-        for service_name in self.PIPX_SERVICES:
-            # Check if service is enabled in config
-            if not self.should_enable_service(service_name):
-                self.logger.debug(f"Skipping {service_name} (disabled in config)")
-                continue
-
-            self.logger.debug(f"🔍 Checking {service_name} for issues...")
-            issue_type = self._detect_service_issue(service_name)
-            if issue_type:
-                services_to_fix.append((service_name, issue_type))
-                self.logger.debug(f"  ⚠️  Found issue with {service_name}: {issue_type}")
-            else:
-                self.logger.debug(f"  ✅ {service_name} is functioning correctly")
-
-        if not services_to_fix:
-            return True, "All MCP services are functioning correctly"
-
-        # Fix each problematic service
-        for service_name, issue_type in services_to_fix:
-            self.logger.info(f"🔧 Fixing {service_name}: {issue_type}")
-
-            if issue_type == "not_installed":
-                # Install the service
-                success, method = self._install_service_with_fallback(service_name)
-                if success:
-                    fixed_services.append(f"{service_name} (installed via {method})")
-                else:
-                    failed_services.append(f"{service_name} (installation failed)")
-
-            elif issue_type == "import_error":
-                # Reinstall to fix corrupted installation
-                self.logger.info(
-                    f"  Reinstalling {service_name} to fix import errors..."
-                )
-                success = self._reinstall_service(service_name)
-                if success:
-                    # NOTE: Removed automatic dependency injection workaround
-                    # Package maintainers should fix dependency declarations
-                    fixed_services.append(f"{service_name} (reinstalled)")
-                else:
-                    failed_services.append(f"{service_name} (reinstall failed)")
-
-            elif issue_type == "missing_dependency":
-                # Fix missing dependencies - try injection first, then reinstall if needed
-                self.logger.info(
-                    f"  {service_name} has missing dependencies - attempting fix..."
-                )
-
-                # First try to inject dependencies without reinstalling
-                injection_success = self._inject_missing_dependencies(service_name)
-
-                if injection_success:
-                    # Verify the fix worked
-                    issue_after_injection = self._detect_service_issue(service_name)
-                    if issue_after_injection is None:
-                        fixed_services.append(f"{service_name} (dependencies injected)")
-                        self.logger.info(
-                            f"  ✅ Fixed {service_name} with dependency injection"
-                        )
-                        continue  # Move to next service
-
-                # If injection alone didn't work, try full reinstall
-                self.logger.info(
-                    "  Dependency injection insufficient, trying full reinstall..."
-                )
-                success = self._auto_reinstall_mcp_service(service_name)
-                if success:
-                    fixed_services.append(
-                        f"{service_name} (auto-reinstalled with dependencies)"
-                    )
-                else:
-                    # Provide specific manual fix for known services
-                    if service_name == "mcp-ticketer":
-                        self.logger.warning(
-                            f"  Auto-fix failed for {service_name}. Manual fix: "
-                            f"pipx uninstall {service_name} && pipx install {service_name} && pipx inject {service_name} gql"
-                        )
-                    else:
-                        self.logger.warning(
-                            f"  Auto-reinstall failed for {service_name}. Manual fix: "
-                            f"pipx uninstall {service_name} && pipx install {service_name}"
-                        )
-                    failed_services.append(f"{service_name} (auto-reinstall failed)")
-
-            elif issue_type == "path_issue":
-                # Path issues are handled by config updates
-                self.logger.info(
-                    f"  Path issue for {service_name} will be fixed by config update"
-                )
-                fixed_services.append(f"{service_name} (config updated)")
-
-        # Build result message
-        messages = []
-        if fixed_services:
-            messages.append(f"✅ Fixed: {', '.join(fixed_services)}")
-        if failed_services:
-            messages.append(f"❌ Failed: {', '.join(failed_services)}")
-
-        # Return success if at least some services were fixed
-        success = len(fixed_services) > 0 or len(failed_services) == 0
-        message = " | ".join(messages) if messages else "No services needed fixing"
-
-        # Provide manual fix instructions if auto-fix failed
-        if failed_services:
-            message += "\n\n💡 Manual fix instructions:"
-            for failed in failed_services:
-                service = failed.split(" ")[0]
-                if service in self.SERVICE_MISSING_DEPENDENCIES:
-                    deps = " ".join(
-                        [
-                            f"&& pipx inject {service} {dep}"
-                            for dep in self.SERVICE_MISSING_DEPENDENCIES[service]
-                        ]
-                    )
-                    message += f"\n  • {service}: pipx uninstall {service} && pipx install {service} {deps}"
-                else:
-                    message += f"\n  • {service}: pipx uninstall {service} && pipx install {service}"
-
-        return success, message
+        # Services should stand on their own - no proactive health checking
+        return True, "MCP services managing their own health"
 
     def _detect_service_issue(self, service_name: str) -> Optional[str]:
         """
@@ -1497,7 +1373,7 @@ class MCPConfigManager:
                 )
 
                 if result.returncode == 0:
-                    self.logger.info(f"    ✅ Successfully injected {dep}")
+                    self.logger.debug(f"    ✅ Successfully injected {dep}")
                 # Check if already injected (pipx will complain if package already exists)
                 elif (
                     "already satisfied" in result.stderr.lower()
@@ -1582,7 +1458,7 @@ class MCPConfigManager:
                     )
 
             # Verify the reinstall worked
-            self.logger.info(f"  → Verifying {service_name} installation...")
+            self.logger.debug(f"  → Verifying {service_name} installation...")
             issue = self._detect_service_issue(service_name)
 
             if issue is None: