claude-mpm 4.0.19__py3-none-any.whl → 4.0.22__py3-none-any.whl

claude_mpm/services/agents/memory/agent_memory_manager.py

@@ -1,6 +1,6 @@
-from pathlib import Path
-
 #!/usr/bin/env python3
+
+from pathlib import Path
 """
 Agent Memory Manager Service
 ===========================
@@ -17,7 +17,7 @@ This service provides:
 - Directory initialization with README
 
 Memory files are stored in .claude-mpm/memories/ directory
-following the naming convention: {agent_id}_agent.md
+following the naming convention: {agent_id}_memories.md
 """
 
 import logging
@@ -28,9 +28,6 @@ from typing import Any, Dict, List, Optional, Tuple
 from claude_mpm.core.config import Config
 from claude_mpm.core.interfaces import MemoryServiceInterface
 from claude_mpm.core.unified_paths import get_path_manager
-from claude_mpm.services.project.analyzer import ProjectAnalyzer
-
-from .analyzer import MemoryAnalyzer
 from .content_manager import MemoryContentManager
 from .template_generator import MemoryTemplateGenerator
 
@@ -85,28 +82,24 @@ class AgentMemoryManager(MemoryServiceInterface):
         self.project_root = get_path_manager().project_root
         # Use current working directory by default, not project root
         self.working_directory = working_directory or Path(os.getcwd())
-        self.memories_dir = self.working_directory / ".claude-mpm" / "memories"
+        
+        # Use only project memory directory
+        self.project_memories_dir = self.working_directory / ".claude-mpm" / "memories"
+        
+        # Primary memories_dir points to project
+        self.memories_dir = self.project_memories_dir
+        
+        # Ensure project directory exists
         self._ensure_memories_directory()
 
         # Initialize memory limits from configuration
         self._init_memory_limits()
 
-        # Initialize project analyzer for context-aware memory creation
-        self.project_analyzer = ProjectAnalyzer(self.config, self.working_directory)
-
         # Initialize component services
         self.template_generator = MemoryTemplateGenerator(
-            self.config, self.working_directory, self.project_analyzer
+            self.config, self.working_directory
         )
         self.content_manager = MemoryContentManager(self.memory_limits)
-        self.analyzer = MemoryAnalyzer(
-            self.memories_dir,
-            self.memory_limits,
-            self.agent_overrides,
-            self._get_agent_limits,
-            self._get_agent_auto_learning,
-            self.content_manager,
-        )
 
     @property
     def logger(self):
@@ -202,31 +195,82 @@ class AgentMemoryManager(MemoryServiceInterface):
         # Fall back to global setting
         return self.auto_learning
 
+    def _get_memory_file_with_migration(self, directory: Path, agent_id: str) -> Path:
+        """Get memory file path, migrating from old naming if needed.
+        
+        WHY: Supports backward compatibility by automatically migrating from
+        the old {agent_id}_agent.md and {agent_id}.md formats to the new {agent_id}_memories.md format.
+        
+        Args:
+            directory: Directory containing memory files
+            agent_id: The agent identifier
+            
+        Returns:
+            Path: Path to the memory file (may not exist)
+        """
+        new_file = directory / f"{agent_id}_memories.md"
+        # Support migration from both old formats
+        old_file_agent = directory / f"{agent_id}_agent.md"
+        old_file_simple = directory / f"{agent_id}.md"
+        
+        # Migrate from old formats if needed
+        if not new_file.exists():
+            # Try migrating from {agent_id}_agent.md first
+            if old_file_agent.exists():
+                try:
+                    content = old_file_agent.read_text(encoding="utf-8")
+                    new_file.write_text(content, encoding="utf-8")
+                    
+                    # Delete old file for all agents
+                    old_file_agent.unlink()
+                    self.logger.info(f"Migrated memory file from {old_file_agent.name} to {new_file.name}")
+                except Exception as e:
+                    self.logger.error(f"Failed to migrate memory file for {agent_id}: {e}")
+                    return old_file_agent
+            # Try migrating from {agent_id}.md
+            elif old_file_simple.exists():
+                try:
+                    content = old_file_simple.read_text(encoding="utf-8")
+                    new_file.write_text(content, encoding="utf-8")
+                    
+                    # Delete old file for all agents
+                    old_file_simple.unlink()
+                    self.logger.info(f"Migrated memory file from {old_file_simple.name} to {new_file.name}")
+                except Exception as e:
+                    self.logger.error(f"Failed to migrate memory file for {agent_id}: {e}")
+                    return old_file_simple
+        
+        return new_file
+
     def load_agent_memory(self, agent_id: str) -> str:
-        """Load agent memory file content.
+        """Load agent memory file content from project directory.
 
         WHY: Agents need to read their accumulated knowledge before starting tasks
-        to apply learned patterns and avoid repeated mistakes.
+        to apply learned patterns and avoid repeated mistakes. All memories are
+        now stored at the project level for consistency.
 
         Args:
-            agent_id: The agent identifier (e.g., 'research', 'engineer')
+            agent_id: The agent identifier (e.g., 'PM', 'research', 'engineer')
 
         Returns:
             str: The memory file content, creating default if doesn't exist
         """
-        memory_file = self.memories_dir / f"{agent_id}_agent.md"
-
-        if not memory_file.exists():
-            self.logger.info(f"Creating default memory for agent: {agent_id}")
-            return self._create_default_memory(agent_id)
-
-        try:
-            content = memory_file.read_text(encoding="utf-8")
-            return self.content_manager.validate_and_repair(content, agent_id)
-        except Exception as e:
-            self.logger.error(f"Error reading memory file for {agent_id}: {e}")
-            # Return default memory on error - never fail
-            return self._create_default_memory(agent_id)
+        # All agents use project directory
+        project_memory_file = self._get_memory_file_with_migration(self.project_memories_dir, agent_id)
+        
+        # Load project-level memory if exists
+        if project_memory_file.exists():
+            try:
+                project_memory = project_memory_file.read_text(encoding="utf-8")
+                project_memory = self.content_manager.validate_and_repair(project_memory, agent_id)
+                self.logger.debug(f"Loaded project-level memory for {agent_id}")
+                return project_memory
+            except Exception as e:
+                self.logger.error(f"Error reading project memory file for {agent_id}: {e}")
+        
+        # Memory doesn't exist - create default in project directory
+        self.logger.info(f"Creating default memory for agent: {agent_id}")
+        return self._create_default_memory(agent_id)
 
     def update_agent_memory(self, agent_id: str, section: str, new_item: str) -> bool:
         """Add new learning item to specified section.
@@ -317,9 +361,10 @@ class AgentMemoryManager(MemoryServiceInterface):
         # Delegate to template generator
         template = self.template_generator.create_default_memory(agent_id, limits)
 
-        # Save default file
+        # Save default file to project directory
         try:
-            memory_file = self.memories_dir / f"{agent_id}_agent.md"
+            target_dir = self.memories_dir
+            memory_file = target_dir / f"{agent_id}_memories.md"
             memory_file.write_text(template, encoding="utf-8")
             self.logger.info(f"Created project-specific memory file for {agent_id}")
 
@@ -342,9 +387,16 @@ class AgentMemoryManager(MemoryServiceInterface):
             bool: True if save succeeded
         """
         try:
-            memory_file = self.memories_dir / f"{agent_id}_agent.md"
+            # All agents save to project directory
+            target_dir = self.project_memories_dir
+            
+            # Ensure directory exists
+            target_dir.mkdir(parents=True, exist_ok=True)
+            
+            memory_file = target_dir / f"{agent_id}_memories.md"
             memory_file.write_text(content, encoding="utf-8")
-            self.logger.debug(f"Saved memory for {agent_id}")
+            
+            self.logger.info(f"Saved {agent_id} memory to project directory: {memory_file}")
             return True
         except Exception as e:
             self.logger.error(f"Error saving memory for {agent_id}: {e}")
@@ -436,6 +488,375 @@ class AgentMemoryManager(MemoryServiceInterface):
             self.logger.error(f"Error routing memory command: {e}")
             return {"success": False, "error": str(e)}
 
+    def extract_and_update_memory(self, agent_id: str, response: str) -> bool:
+        """Extract memory updates from agent response and update memory file.
+
+        WHY: Agents provide memory updates in their responses that need to be
+        extracted and persisted. This method looks for "remember" field in JSON
+        responses and merges new learnings with existing memory.
+
+        Args:
+            agent_id: The agent identifier
+            response: The agent's response text (may contain JSON)
+
+        Returns:
+            bool: True if memory was updated, False otherwise
+        """
+        try:
+            import json
+            import re
+            
+            # Log that we're processing memory for this agent
+            is_pm = agent_id.upper() == "PM"
+            self.logger.debug(f"Extracting memory for {agent_id} (is_pm={is_pm})")
+            
+            # Look for JSON block in the response
+            # Pattern matches ```json ... ``` blocks
+            json_pattern = r'```json\s*(.*?)\s*```'
+            json_matches = re.findall(json_pattern, response, re.DOTALL)
+            
+            if not json_matches:
+                # Also try to find inline JSON objects
+                json_pattern2 = r'\{[^{}]*"(?:remember|Remember)"[^{}]*\}'
+                json_matches = re.findall(json_pattern2, response, re.DOTALL)
+            
+            for json_str in json_matches:
+                try:
+                    data = json.loads(json_str)
+                    
+                    # Check for memory updates in "remember" field
+                    memory_items = None
+                    
+                    # Check both "remember" and "Remember" fields
+                    if "remember" in data:
+                        memory_items = data["remember"]
+                    elif "Remember" in data:
+                        memory_items = data["Remember"]
+                    
+                    # Process memory items if found and not null
+                    if memory_items is not None and memory_items != "null":
+                        # Skip if explicitly null or empty list
+                        if isinstance(memory_items, list) and len(memory_items) > 0:
+                            # Filter out empty strings and None values
+                            valid_items = []
+                            for item in memory_items:
+                                if item and isinstance(item, str) and item.strip():
+                                    valid_items.append(item.strip())
+                            
+                            # Only proceed if we have valid items
+                            if valid_items:
+                                self.logger.info(f"Found {len(valid_items)} memory items for {agent_id}: {valid_items[:2]}...")
+                                success = self._add_learnings_to_memory(agent_id, valid_items)
+                                if success:
+                                    self.logger.info(f"Successfully saved {len(valid_items)} memories for {agent_id} to project directory")
+                                    return True
+                                else:
+                                    self.logger.error(f"Failed to save memories for {agent_id}")
+                    
+                except json.JSONDecodeError as je:
+                    # Not valid JSON, continue to next match
+                    self.logger.debug(f"JSON decode error for {agent_id}: {je}")
+                    continue
+            
+            self.logger.debug(f"No memory items found in response for {agent_id}")
+            return False
+            
+        except Exception as e:
+            self.logger.error(f"Error extracting memory from response for {agent_id}: {e}")
+            return False
+    
+    def _add_learnings_to_memory(self, agent_id: str, learnings: List[str]) -> bool:
+        """Add new learnings to existing agent memory.
+        
+        WHY: Instead of replacing all memory, we want to intelligently merge new
+        learnings with existing knowledge, avoiding duplicates and maintaining
+        the most relevant information. PM memories are always saved to user dir.
+        
+        Args:
+            agent_id: The agent identifier
+            learnings: List of new learning strings to add
+            
+        Returns:
+            bool: True if memory was successfully updated
+        """
+        try:
+            # Load existing memory
+            current_memory = self.load_agent_memory(agent_id)
+            
+            # Parse existing memory into sections
+            sections = self._parse_memory_sections(current_memory)
+            
+            # Clean sections - remove template placeholder text
+            sections = self._clean_template_placeholders(sections)
+            
+            # Determine which section to add learnings to based on content
+            for learning in learnings:
+                if not learning or not isinstance(learning, str):
+                    continue
+                    
+                learning = learning.strip()
+                if not learning:
+                    continue
+                
+                # Categorize the learning based on keywords
+                section = self._categorize_learning(learning)
+                
+                # Add to appropriate section if not duplicate
+                if section not in sections:
+                    sections[section] = []
+                
+                # Check for duplicates (case-insensitive) - FIXED LOGIC
+                normalized_learning = learning.lower()
+                # Strip bullet points from existing items for comparison
+                existing_normalized = [item.lstrip('- ').strip().lower() for item in sections[section]]
+                
+                if normalized_learning not in existing_normalized:
+                    # Add bullet point if not present
+                    if not learning.startswith("-"):
+                        learning = f"- {learning}"
+                    sections[section].append(learning)
+                    self.logger.info(f"Added new memory for {agent_id}: {learning[:50]}...")
+                else:
+                    self.logger.debug(f"Skipping duplicate memory for {agent_id}: {learning}")
+            
+            # Rebuild memory content
+            new_content = self._build_memory_content(agent_id, sections)
+            
+            # Validate and save
+            agent_limits = self._get_agent_limits(agent_id)
+            if self.content_manager.exceeds_limits(new_content, agent_limits):
+                self.logger.debug(f"Memory for {agent_id} exceeds limits, truncating")
+                new_content = self.content_manager.truncate_to_limits(new_content, agent_limits)
+            
+            # All memories go to project directory
+            return self._save_memory_file(agent_id, new_content)
+            
+        except Exception as e:
+            self.logger.error(f"Error adding learnings to memory for {agent_id}: {e}")
+            return False
+    
+    def _clean_template_placeholders(self, sections: Dict[str, List[str]]) -> Dict[str, List[str]]:
+        """Remove template placeholder text from sections.
+        
+        Args:
+            sections: Dict mapping section names to lists of items
+            
+        Returns:
+            Dict with placeholder text removed
+        """
+        # Template placeholder patterns to remove
+        placeholders = [
+            "Analyze project structure to understand architecture patterns",
+            "Observe codebase patterns and conventions during tasks",
+            "Extract implementation guidelines from project documentation",
+            "Learn from errors encountered during project work",
+            "Project analysis pending - gather context during tasks",
+            "claude-mpm: Software project requiring analysis"
+        ]
+        
+        cleaned = {}
+        for section_name, items in sections.items():
+            cleaned_items = []
+            for item in items:
+                # Remove bullet point for comparison
+                item_text = item.lstrip("- ").strip()
+                # Keep item if it's not a placeholder
+                if item_text and item_text not in placeholders:
+                    cleaned_items.append(item)
+            
+            # Only include section if it has real content
+            if cleaned_items:
+                cleaned[section_name] = cleaned_items
+        
+        return cleaned
+    
+    def _categorize_learning(self, learning: str) -> str:
+        """Categorize a learning item into appropriate section.
+        
+        Args:
+            learning: The learning string to categorize
+            
+        Returns:
+            str: The section name for this learning
+        """
+        learning_lower = learning.lower()
+        
+        # Check for keywords to categorize with improved patterns
+        # Order matters - more specific patterns should come first
+        
+        # Architecture keywords
+        if any(word in learning_lower for word in ["architecture", "structure", "design", "module", "component", "microservices", "service-oriented"]):
+            return "Project Architecture"
+            
+        # Integration keywords (check before patterns to avoid "use" conflict)
+        elif any(word in learning_lower for word in ["integration", "interface", "api", "connection", "database", "pooling", "via"]):
+            return "Integration Points"
+            
+        # Mistake keywords (check before patterns to avoid conflicts)
+        elif any(word in learning_lower for word in ["mistake", "error", "avoid", "don't", "never", "not"]):
+            return "Common Mistakes to Avoid"
+            
+        # Context keywords (check before patterns to avoid "working", "version" conflicts)
+        elif any(word in learning_lower for word in ["context", "current", "currently", "working", "version", "release", "candidate"]):
+            return "Current Technical Context"
+            
+        # Guideline keywords (check before patterns to avoid "must", "should" conflicts)
+        elif any(word in learning_lower for word in ["guideline", "rule", "standard", "practice", "docstring", "documentation", "must", "should", "include", "comprehensive"]):
+            return "Implementation Guidelines"
+            
+        # Pattern keywords (including dependency injection, conventions)
+        elif any(word in learning_lower for word in ["pattern", "convention", "style", "format", "dependency injection", "instantiation", "use", "implement"]):
+            return "Coding Patterns Learned"
+            
+        # Strategy keywords  
+        elif any(word in learning_lower for word in ["strategy", "approach", "method", "technique", "effective"]):
+            return "Effective Strategies"
+            
+        # Performance keywords
+        elif any(word in learning_lower for word in ["performance", "optimization", "speed", "efficiency"]):
+            return "Performance Considerations"
+            
+        # Domain keywords
+        elif any(word in learning_lower for word in ["domain", "business", "specific"]):
+            return "Domain-Specific Knowledge"
+            
+        else:
+            return "Recent Learnings"
+    
+    def _build_memory_content(self, agent_id: str, sections: Dict[str, List[str]]) -> str:
+        """Build memory content from sections.
+        
+        Args:
+            agent_id: The agent identifier
+            sections: Dict mapping section names to lists of items
+            
+        Returns:
+            str: The formatted memory content
+        """
+        lines = []
+        
+        # Add header
+        lines.append(f"# {agent_id.capitalize()} Agent Memory")
+        lines.append("")
+        lines.append(f"<!-- Last Updated: {datetime.now().isoformat()} -->")
+        lines.append("")
+        
+        # Add sections in consistent order
+        section_order = [
+            "Project Architecture",
+            "Implementation Guidelines",
+            "Common Mistakes to Avoid",
+            "Current Technical Context",
+            "Coding Patterns Learned",
+            "Effective Strategies",
+            "Integration Points",
+            "Performance Considerations",
+            "Domain-Specific Knowledge",
+            "Recent Learnings"
+        ]
+        
+        for section_name in section_order:
+            if section_name in sections and sections[section_name]:
+                lines.append(f"## {section_name}")
+                lines.append("")
+                for item in sections[section_name]:
+                    if item.strip():
+                        lines.append(item)
+                lines.append("")
+        
+        # Add any remaining sections
+        remaining = set(sections.keys()) - set(section_order)
+        for section_name in sorted(remaining):
+            if sections[section_name]:
+                lines.append(f"## {section_name}")
+                lines.append("")
+                for item in sections[section_name]:
+                    if item.strip():
+                        lines.append(item)
+                lines.append("")
+        
+        return '\n'.join(lines)
+    
+    def replace_agent_memory(self, agent_id: str, memory_sections: Dict[str, List[str]]) -> bool:
+        """Replace agent's memory with new content organized by sections.
+
+        WHY: When agents provide memory updates, they replace the existing memory
+        rather than appending to it. This ensures memories stay current and relevant.
+
+        Args:
+            agent_id: The agent identifier
+            memory_sections: Dict mapping section names to lists of memory items
+
+        Returns:
+            bool: True if memory was successfully replaced
+        """
+        try:
+            # Build new memory content
+            lines = []
+            
+            # Add header
+            lines.append(f"# {agent_id.capitalize()} Agent Memory")
+            lines.append("")
+            lines.append(f"<!-- Last Updated: {datetime.now().isoformat()} -->")
+            lines.append("")
+            
+            # Add sections in a consistent order
+            section_order = [
+                "Project Architecture",
+                "Implementation Guidelines", 
+                "Common Mistakes to Avoid",
+                "Current Technical Context",
+                "Coding Patterns Learned",
+                "Effective Strategies",
+                "Integration Points",
+                "Performance Considerations",
+                "Domain-Specific Knowledge",
+                "Recent Learnings"
+            ]
+            
+            # First add ordered sections that exist in memory_sections
+            for section_name in section_order:
+                if section_name in memory_sections and memory_sections[section_name]:
+                    lines.append(f"## {section_name}")
+                    lines.append("")
+                    for item in memory_sections[section_name]:
+                        if item.strip():  # Skip empty items
+                            # Add bullet point if not already present
+                            if not item.strip().startswith("-"):
+                                lines.append(f"- {item.strip()}")
+                            else:
+                                lines.append(item.strip())
+                    lines.append("")
+            
+            # Then add any remaining sections not in the order list
+            remaining_sections = set(memory_sections.keys()) - set(section_order)
+            for section_name in sorted(remaining_sections):
+                if memory_sections[section_name]:
+                    lines.append(f"## {section_name}")
+                    lines.append("")
+                    for item in memory_sections[section_name]:
+                        if item.strip():
+                            if not item.strip().startswith("-"):
+                                lines.append(f"- {item.strip()}")
+                            else:
+                                lines.append(item.strip())
+                    lines.append("")
+            
+            new_content = '\n'.join(lines)
+            
+            # Validate and save
+            agent_limits = self._get_agent_limits(agent_id)
+            if self.content_manager.exceeds_limits(new_content, agent_limits):
+                self.logger.debug(f"Memory for {agent_id} exceeds limits, truncating")
+                new_content = self.content_manager.truncate_to_limits(new_content, agent_limits)
+            
+            # Save the new memory
+            return self._save_memory_file(agent_id, new_content)
+            
+        except Exception as e:
+            self.logger.error(f"Error replacing memory for {agent_id}: {e}")
+            return False
+
     def get_memory_status(self) -> Dict[str, Any]:
         """Get comprehensive memory system status.
 
@@ -446,7 +867,32 @@ class AgentMemoryManager(MemoryServiceInterface):
         Returns:
             Dict containing comprehensive memory system status
         """
-        return self.analyzer.get_memory_status()
+        # Simplified status implementation without analyzer
+        status = {
+            "system_enabled": self.memory_enabled,
+            "auto_learning": self.auto_learning,
+            "memory_directory": str(self.memories_dir),
+            "total_agents": 0,
+            "total_size_kb": 0,
+            "agents": {},
+            "system_health": "healthy"
+        }
+        
+        if self.memories_dir.exists():
+            memory_files = list(self.memories_dir.glob("*_memories.md"))
+            status["total_agents"] = len(memory_files)
+            
+            for file_path in memory_files:
+                if file_path.name != "README.md":
+                    size_kb = file_path.stat().st_size / 1024
+                    status["total_size_kb"] += size_kb
+                    agent_id = file_path.stem.replace("_memories", "")
+                    status["agents"][agent_id] = {
+                        "file": file_path.name,
+                        "size_kb": round(size_kb, 2)
+                    }
+        
+        return status
 
     def cross_reference_memories(self, query: Optional[str] = None) -> Dict[str, Any]:
         """Find common patterns and cross-references across agent memories.
@@ -461,7 +907,12 @@ class AgentMemoryManager(MemoryServiceInterface):
         Returns:
             Dict containing cross-reference analysis results
         """
-        return self.analyzer.cross_reference_memories(query)
+        # Deprecated - return informative message
+        return {
+            "status": "deprecated",
+            "message": "Cross-reference analysis has been deprecated in favor of simplified memory management",
+            "suggestion": "Use get_memory_status() for memory overview"
+        }
 
     def get_all_memories_raw(self) -> Dict[str, Any]:
         """Get all agent memories in structured JSON format.
@@ -473,7 +924,12 @@ class AgentMemoryManager(MemoryServiceInterface):
         Returns:
             Dict containing structured memory data for all agents
         """
-        return self.analyzer.get_all_memories_raw()
+        # Deprecated - return informative message
+        return {
+            "status": "deprecated", 
+            "message": "Raw memory access has been deprecated in favor of simplified memory management",
+            "suggestion": "Use load_agent_memory() for specific agent memories"
+        }
 
     def _ensure_memories_directory(self):
         """Ensure memories directory exists with README.
@@ -531,6 +987,45 @@ Standard markdown with structured sections. Agents expect:
             self.logger.error(f"Error ensuring memories directory: {e}")
             # Continue anyway - memory system should not block operations
 
+
+    
+    def _parse_memory_sections(self, memory_content: str) -> Dict[str, List[str]]:
+        """Parse memory content into sections and items.
+        
+        Args:
+            memory_content: Raw memory file content
+            
+        Returns:
+            Dict mapping section names to lists of items
+        """
+        sections = {}
+        current_section = None
+        current_items = []
+        
+        for line in memory_content.split('\n'):
+            # Skip metadata lines
+            if line.startswith('<!-- ') and line.endswith(' -->'):
+                continue
+            # Check for section headers (## Level 2 headers)
+            elif line.startswith('## '):
+                # Save previous section if exists
+                if current_section and current_items:
+                    sections[current_section] = current_items
+                
+                # Start new section
+                current_section = line[3:].strip()  # Remove "## " prefix
+                current_items = []
+            # Collect non-empty lines as items (but not HTML comments)
+            elif line.strip() and current_section and not line.strip().startswith('<!--'):
+                # Keep the full line with its formatting
+                current_items.append(line.strip())
+        
+        # Save last section
+        if current_section and current_items:
+            sections[current_section] = current_items
+        
+        return sections
+
     # ================================================================================
     # Interface Adapter Methods
     # ================================================================================
@@ -570,7 +1065,7 @@ Standard markdown with structured sections. Agents expect:
             True if save successful
         """
         try:
-            memory_path = self.memories_dir / f"{agent_id}_agent.md"
+            memory_path = self.memories_dir / f"{agent_id}_memories.md"
 
             # Validate size before saving
             is_valid, error_msg = self.validate_memory_size(content)
@@ -613,7 +1108,45 @@ Standard markdown with structured sections. Agents expect:
         Returns:
             Dictionary with memory metrics
         """
-        return self.analyzer.get_memory_metrics(agent_id)
+        # Minimal implementation for interface compliance
+        metrics = {
+            "total_memory_kb": 0,
+            "agent_count": 0,
+            "agents": {}
+        }
+        
+        if self.memories_dir.exists():
+            if agent_id:
+                # Metrics for specific agent
+                memory_file = self.memories_dir / f"{agent_id}_memories.md"
+                if memory_file.exists():
+                    size_kb = memory_file.stat().st_size / 1024
+                    metrics["agents"][agent_id] = {
+                        "size_kb": round(size_kb, 2),
+                        "limit_kb": self._get_agent_limits(agent_id)["max_file_size_kb"],
+                        "usage_percent": round((size_kb / self._get_agent_limits(agent_id)["max_file_size_kb"]) * 100, 1)
+                    }
+                    metrics["total_memory_kb"] = round(size_kb, 2)
+                    metrics["agent_count"] = 1
+            else:
+                # Metrics for all agents
+                memory_files = list(self.memories_dir.glob("*_memories.md"))
+                for file_path in memory_files:
+                    if file_path.name != "README.md":
+                        agent_name = file_path.stem.replace("_memories", "")
+                        size_kb = file_path.stat().st_size / 1024
+                        limit_kb = self._get_agent_limits(agent_name)["max_file_size_kb"]
+                        metrics["agents"][agent_name] = {
+                            "size_kb": round(size_kb, 2),
+                            "limit_kb": limit_kb,
+                            "usage_percent": round((size_kb / limit_kb) * 100, 1)
+                        }
+                        metrics["total_memory_kb"] += size_kb
+                
+                metrics["total_memory_kb"] = round(metrics["total_memory_kb"], 2)
+                metrics["agent_count"] = len(metrics["agents"])
+        
+        return metrics
 
 
 # Convenience functions for external use