claude-mpm 3.7.4__py3-none-any.whl → 3.8.1__py3-none-any.whl

claude_mpm/services/agent/management.py

@@ -0,0 +1,598 @@
+#!/usr/bin/env python3
+"""
+Agent Management Service
+========================
+
+Comprehensive service for managing agent definitions with CRUD operations,
+section extraction/updates, and version management.
+
+Uses python-frontmatter and mistune for markdown parsing as recommended.
+"""
+
+import os
+import re
+import json
+import yaml
+import logging
+from pathlib import Path
+from typing import Dict, List, Optional, Tuple, Any
+from datetime import datetime
+
+import frontmatter
+import mistune
+
+from claude_mpm.models.agent_definition import (
+    AgentDefinition, AgentMetadata, AgentType, 
+    AgentSection, AgentWorkflow, AgentPermissions
+)
+from ..deployment.agent_versioning import AgentVersionManager
+from claude_mpm.services.memory.cache.shared_prompt_cache import SharedPromptCache
+from claude_mpm.utils.paths import PathResolver
+from claude_mpm.core.config_paths import ConfigPaths
+
+logger = logging.getLogger(__name__)
+
+
+class AgentManager:
+    """Manages agent definitions with CRUD operations and versioning."""
+    
+    def __init__(self, framework_dir: Optional[Path] = None, project_dir: Optional[Path] = None):
+        """
+        Initialize AgentManager.
+        
+        Args:
+            framework_dir: Path to agents templates directory
+            project_dir: Path to project-specific agents directory
+        """
+        # Use PathResolver for consistent path discovery
+        if framework_dir is None:
+            try:
+                # Use agents templates directory
+                self.framework_dir = Path(__file__).parent.parent / "agents" / "templates"
+            except Exception:
+                # Fallback to agents directory
+                self.framework_dir = PathResolver.get_agents_dir()
+        else:
+            self.framework_dir = framework_dir
+            
+        if project_dir is None:
+            project_root = PathResolver.get_project_root()
+            # Use direct agents directory without subdirectory to match deployment expectations
+            self.project_dir = project_root / ConfigPaths.CONFIG_DIR / "agents"
+        else:
+            self.project_dir = project_dir
+        self.version_manager = AgentVersionManager()
+        self.cache = SharedPromptCache.get_instance()
+        self._markdown = mistune.create_markdown()
+        
+    def create_agent(self, name: str, definition: AgentDefinition, location: str = "project") -> Path:
+        """
+        Create a new agent definition file.
+        
+        Args:
+            name: Agent name (e.g., "performance-agent")
+            definition: Agent definition object
+            location: "project" or "framework"
+            
+        Returns:
+            Path to created file
+        """
+        # Determine target directory
+        target_dir = self.project_dir if location == "project" else self.framework_dir
+        target_dir.mkdir(parents=True, exist_ok=True)
+        
+        # Generate markdown content
+        content = self._definition_to_markdown(definition)
+        
+        # Write file
+        file_path = target_dir / f"{name}.md"
+        file_path.write_text(content, encoding='utf-8')
+        
+        # Clear cache
+        self._clear_agent_cache(name)
+        
+        logger.info(f"Created agent '{name}' at {file_path}")
+        return file_path
+    
+    def read_agent(self, name: str) -> Optional[AgentDefinition]:
+        """
+        Read an agent definition.
+        
+        Args:
+            name: Agent name (without .md extension)
+            
+        Returns:
+            AgentDefinition or None if not found
+        """
+        # Try to find the agent file
+        agent_path = self._find_agent_file(name)
+        if not agent_path:
+            logger.warning(f"Agent '{name}' not found")
+            return None
+        
+        try:
+            # Read and parse the file
+            content = agent_path.read_text(encoding='utf-8')
+            return self._parse_agent_markdown(content, name, str(agent_path))
+        except Exception as e:
+            logger.error(f"Error reading agent '{name}': {e}")
+            return None
+    
+    def update_agent(self, name: str, updates: Dict[str, Any], 
+                    increment_version: bool = True) -> Optional[AgentDefinition]:
+        """
+        Update an agent definition.
+        
+        Args:
+            name: Agent name
+            updates: Dictionary of updates to apply
+            increment_version: Whether to increment serial version
+            
+        Returns:
+            Updated AgentDefinition or None if failed
+        """
+        # Read current definition
+        agent_def = self.read_agent(name)
+        if not agent_def:
+            return None
+        
+        # Apply updates
+        for key, value in updates.items():
+            if hasattr(agent_def, key):
+                setattr(agent_def, key, value)
+            elif key in ["type", "model_preference", "tags", "specializations"]:
+                setattr(agent_def.metadata, key, value)
+        
+        # Increment version if requested
+        if increment_version:
+            agent_def.metadata.increment_serial_version()
+            agent_def.metadata.last_updated = datetime.now()
+        
+        # Write back
+        agent_path = self._find_agent_file(name)
+        if agent_path:
+            content = self._definition_to_markdown(agent_def)
+            agent_path.write_text(content, encoding='utf-8')
+            
+            # Clear cache
+            self._clear_agent_cache(name)
+            
+            logger.info(f"Updated agent '{name}' to version {agent_def.metadata.version}")
+            return agent_def
+        
+        return None
+    
+    def update_section(self, name: str, section: AgentSection, content: str,
+                      increment_version: bool = True) -> Optional[AgentDefinition]:
+        """
+        Update a specific section of an agent.
+        
+        Args:
+            name: Agent name
+            section: Section to update
+            content: New section content
+            increment_version: Whether to increment version
+            
+        Returns:
+            Updated AgentDefinition or None
+        """
+        agent_def = self.read_agent(name)
+        if not agent_def:
+            return None
+        
+        # Map section to attribute
+        section_map = {
+            AgentSection.PRIMARY_ROLE: "primary_role",
+            AgentSection.CAPABILITIES: "capabilities",
+            AgentSection.TOOLS: "tools_commands",
+            AgentSection.ESCALATION: "escalation_triggers",
+            AgentSection.KPI: "kpis",
+            AgentSection.DEPENDENCIES: "dependencies"
+        }
+        
+        if section in section_map:
+            attr_name = section_map[section]
+            if section in [AgentSection.CAPABILITIES, AgentSection.ESCALATION, 
+                          AgentSection.KPI, AgentSection.DEPENDENCIES]:
+                # Parse list content
+                setattr(agent_def, attr_name, self._parse_list_content(content))
+            else:
+                setattr(agent_def, attr_name, content.strip())
+        
+        # Special handling for complex sections
+        elif section == AgentSection.WHEN_TO_USE:
+            agent_def.when_to_use = self._parse_when_to_use(content)
+        elif section == AgentSection.AUTHORITY:
+            agent_def.authority = self._parse_authority(content)
+        elif section == AgentSection.WORKFLOWS:
+            agent_def.workflows = self._parse_workflows(content)
+        
+        # Update raw section
+        agent_def.raw_sections[section.value] = content
+        
+        # Increment version
+        if increment_version:
+            agent_def.metadata.increment_serial_version()
+            agent_def.metadata.last_updated = datetime.now()
+        
+        # Write back
+        return self.update_agent(name, {}, increment_version=False)
+    
+    def delete_agent(self, name: str) -> bool:
+        """
+        Delete an agent definition.
+        
+        Args:
+            name: Agent name
+            
+        Returns:
+            True if deleted, False otherwise
+        """
+        agent_path = self._find_agent_file(name)
+        if not agent_path:
+            return False
+        
+        try:
+            agent_path.unlink()
+            self._clear_agent_cache(name)
+            logger.info(f"Deleted agent '{name}'")
+            return True
+        except Exception as e:
+            logger.error(f"Error deleting agent '{name}': {e}")
+            return False
+    
+    def list_agents(self, location: Optional[str] = None) -> Dict[str, Dict[str, Any]]:
+        """
+        List all available agents.
+        
+        Args:
+            location: Filter by location ("project", "framework", or None for all)
+            
+        Returns:
+            Dictionary of agent info
+        """
+        agents = {}
+        
+        # Check framework agents
+        if location in [None, "framework"]:
+            for agent_file in self.framework_dir.glob("*.md"):
+                if agent_file.name != "base_agent.md":
+                    agent_name = agent_file.stem
+                    agent_def = self.read_agent(agent_name)
+                    if agent_def:
+                        agents[agent_name] = {
+                            "location": "framework",
+                            "path": str(agent_file),
+                            "version": agent_def.metadata.version,
+                            "type": agent_def.metadata.type.value,
+                            "specializations": agent_def.metadata.specializations
+                        }
+        
+        # Check project agents
+        if location in [None, "project"] and self.project_dir.exists():
+            for agent_file in self.project_dir.glob("*.md"):
+                agent_name = agent_file.stem
+                agent_def = self.read_agent(agent_name)
+                if agent_def:
+                    agents[agent_name] = {
+                        "location": "project",
+                        "path": str(agent_file),
+                        "version": agent_def.metadata.version,
+                        "type": agent_def.metadata.type.value,
+                        "specializations": agent_def.metadata.specializations
+                    }
+        
+        return agents
+    
+    def get_agent_api(self, name: str) -> Optional[Dict[str, Any]]:
+        """
+        Get agent data in API-friendly format.
+        
+        Args:
+            name: Agent name
+            
+        Returns:
+            Agent data dictionary or None
+        """
+        agent_def = self.read_agent(name)
+        if not agent_def:
+            return None
+        
+        return agent_def.to_dict()
+    
+    # Private helper methods
+    
+    def _find_agent_file(self, name: str) -> Optional[Path]:
+        """Find agent file in project or framework directories."""
+        # Check project first (higher precedence)
+        if self.project_dir.exists():
+            project_path = self.project_dir / f"{name}.md"
+            if project_path.exists():
+                return project_path
+        
+        # Check framework
+        framework_path = self.framework_dir / f"{name}.md"
+        if framework_path.exists():
+            return framework_path
+        
+        return None
+    
+    def _parse_agent_markdown(self, content: str, name: str, file_path: str) -> AgentDefinition:
+        """Parse markdown content into AgentDefinition."""
+        # Parse frontmatter
+        post = frontmatter.loads(content)
+        
+        # Extract metadata
+        metadata = AgentMetadata(
+            type=AgentType(post.metadata.get("type", "core")),
+            model_preference=post.metadata.get("model_preference", "claude-3-sonnet"),
+            version=post.metadata.get("version", "1.0.0"),
+            last_updated=post.metadata.get("last_updated"),
+            author=post.metadata.get("author"),
+            tags=post.metadata.get("tags", []),
+            specializations=post.metadata.get("specializations", [])
+        )
+        
+        # Extract version from content if not in frontmatter
+        if not post.metadata.get("version"):
+            version = self.version_manager.extract_version_from_markdown(content)
+            if version:
+                metadata.version = version
+        
+        # Parse sections
+        sections = self._extract_sections(post.content)
+        
+        # Extract title
+        title_match = re.search(r'^#\s+(.+)$', post.content, re.MULTILINE)
+        title = title_match.group(1) if title_match else name.replace('-', ' ').title()
+        
+        # Build definition
+        definition = AgentDefinition(
+            name=name,
+            title=title,
+            file_path=file_path,
+            metadata=metadata,
+            primary_role=sections.get("Primary Role", ""),
+            when_to_use=self._parse_when_to_use(sections.get("When to Use This Agent", "")),
+            capabilities=self._parse_list_content(sections.get("Core Capabilities", "")),
+            authority=self._parse_authority(sections.get("Authority & Permissions", "")),
+            workflows=self._parse_workflows(sections.get("Agent-Specific Workflows", "")),
+            escalation_triggers=self._parse_list_content(sections.get("Unique Escalation Triggers", "")),
+            kpis=self._parse_list_content(sections.get("Key Performance Indicators", "")),
+            dependencies=self._parse_list_content(sections.get("Critical Dependencies", "")),
+            tools_commands=sections.get("Specialized Tools/Commands", ""),
+            raw_content=content,
+            raw_sections=sections
+        )
+        
+        return definition
+    
+    def _extract_sections(self, content: str) -> Dict[str, str]:
+        """Extract sections from markdown content."""
+        sections = {}
+        current_section = None
+        current_content = []
+        
+        # Split into lines
+        lines = content.split('\n')
+        
+        for line in lines:
+            # Check if this is a section header
+            header_match = re.match(r'^##\s+(?:🎯|🔧|🔑|📋|🚨|📊|🔄|🛠️)?\s*(.+)$', line)
+            if header_match:
+                # Save previous section
+                if current_section:
+                    sections[current_section] = '\n'.join(current_content).strip()
+                
+                # Start new section
+                current_section = header_match.group(1).strip()
+                current_content = []
+            else:
+                # Add to current section
+                if current_section:
+                    current_content.append(line)
+        
+        # Save last section
+        if current_section:
+            sections[current_section] = '\n'.join(current_content).strip()
+        
+        return sections
+    
+    def _parse_list_content(self, content: str) -> List[str]:
+        """Parse bullet point or numbered list content."""
+        items = []
+        for line in content.split('\n'):
+            # Match bullet points or numbered items
+            match = re.match(r'^[-*•]\s+(.+)$|^\d+\.\s+(.+)$', line.strip())
+            if match:
+                item = match.group(1) or match.group(2)
+                items.append(item.strip())
+        return items
+    
+    def _parse_when_to_use(self, content: str) -> Dict[str, List[str]]:
+        """Parse When to Use section."""
+        result = {"select": [], "do_not_select": []}
+        current_mode = None
+        
+        for line in content.split('\n'):
+            if "Select this agent when:" in line or "**Select this agent when:**" in line:
+                current_mode = "select"
+            elif "Do NOT select for:" in line or "**Do NOT select for:**" in line:
+                current_mode = "do_not_select"
+            elif current_mode and line.strip().startswith('-'):
+                item = line.strip()[1:].strip()
+                result[current_mode].append(item)
+        
+        return result
+    
+    def _parse_authority(self, content: str) -> AgentPermissions:
+        """Parse Authority & Permissions section."""
+        permissions = AgentPermissions()
+        current_section = None
+        
+        for line in content.split('\n'):
+            if "Exclusive Write Access" in line:
+                current_section = "write"
+            elif "Forbidden Operations" in line:
+                current_section = "forbidden"
+            elif "Read Access" in line:
+                current_section = "read"
+            elif current_section and line.strip().startswith('-'):
+                item = line.strip()[1:].strip()
+                # Remove inline comments
+                item = re.sub(r'\s*#.*$', '', item).strip()
+                
+                if current_section == "write":
+                    permissions.exclusive_write_access.append(item)
+                elif current_section == "forbidden":
+                    permissions.forbidden_operations.append(item)
+                elif current_section == "read":
+                    permissions.read_access.append(item)
+        
+        return permissions
+    
+    def _parse_workflows(self, content: str) -> List[AgentWorkflow]:
+        """Parse workflows from YAML blocks."""
+        workflows = []
+        
+        # Find all YAML blocks
+        yaml_blocks = re.findall(r'```yaml\n(.*?)\n```', content, re.DOTALL)
+        
+        for block in yaml_blocks:
+            try:
+                data = yaml.safe_load(block)
+                if isinstance(data, dict) and all(k in data for k in ["trigger", "process", "output"]):
+                    # Extract workflow name from preceding heading if available
+                    name_match = re.search(r'###\s+(.+)\n```yaml\n' + re.escape(block), content)
+                    name = name_match.group(1) if name_match else "Unnamed Workflow"
+                    
+                    workflow = AgentWorkflow(
+                        name=name,
+                        trigger=data["trigger"],
+                        process=data["process"] if isinstance(data["process"], list) else [data["process"]],
+                        output=data["output"],
+                        raw_yaml=block
+                    )
+                    workflows.append(workflow)
+            except yaml.YAMLError:
+                logger.warning("Failed to parse YAML workflow block")
+        
+        return workflows
+    
+    def _definition_to_markdown(self, definition: AgentDefinition) -> str:
+        """Convert AgentDefinition back to markdown."""
+        # Start with frontmatter
+        frontmatter_data = {
+            "type": definition.metadata.type.value,
+            "model_preference": definition.metadata.model_preference,
+            "version": definition.metadata.version,
+            "last_updated": definition.metadata.last_updated,
+            "author": definition.metadata.author,
+            "tags": definition.metadata.tags,
+            "specializations": definition.metadata.specializations
+        }
+        
+        # Remove None values
+        frontmatter_data = {k: v for k, v in frontmatter_data.items() if v is not None}
+        
+        # Build content
+        content = []
+        content.append(f"# {definition.title}\n")
+        
+        # Primary Role
+        content.append("## 🎯 Primary Role")
+        content.append(definition.primary_role)
+        content.append("")
+        
+        # When to Use
+        content.append("## 🎯 When to Use This Agent")
+        content.append("")
+        content.append("**Select this agent when:**")
+        for item in definition.when_to_use.get("select", []):
+            content.append(f"- {item}")
+        content.append("")
+        content.append("**Do NOT select for:**")
+        for item in definition.when_to_use.get("do_not_select", []):
+            content.append(f"- {item}")
+        content.append("")
+        
+        # Capabilities
+        content.append("## 🔧 Core Capabilities")
+        for capability in definition.capabilities:
+            content.append(f"- {capability}")
+        content.append("")
+        
+        # Authority
+        content.append("## 🔑 Authority & Permissions")
+        content.append("")
+        content.append("### ✅ Exclusive Write Access")
+        for item in definition.authority.exclusive_write_access:
+            content.append(f"- {item}")
+        content.append("")
+        content.append("### ❌ Forbidden Operations")
+        for item in definition.authority.forbidden_operations:
+            content.append(f"- {item}")
+        content.append("")
+        
+        # Workflows
+        if definition.workflows:
+            content.append("## 📋 Agent-Specific Workflows")
+            content.append("")
+            for workflow in definition.workflows:
+                content.append(f"### {workflow.name}")
+                content.append("```yaml")
+                yaml_content = {
+                    "trigger": workflow.trigger,
+                    "process": workflow.process,
+                    "output": workflow.output
+                }
+                content.append(yaml.dump(yaml_content, default_flow_style=False).strip())
+                content.append("```")
+                content.append("")
+        
+        # Escalation
+        if definition.escalation_triggers:
+            content.append("## 🚨 Unique Escalation Triggers")
+            for trigger in definition.escalation_triggers:
+                content.append(f"- {trigger}")
+            content.append("")
+        
+        # KPIs
+        if definition.kpis:
+            content.append("## 📊 Key Performance Indicators")
+            for i, kpi in enumerate(definition.kpis, 1):
+                content.append(f"{i}. {kpi}")
+            content.append("")
+        
+        # Dependencies
+        if definition.dependencies:
+            content.append("## 🔄 Critical Dependencies")
+            for dep in definition.dependencies:
+                content.append(f"- {dep}")
+            content.append("")
+        
+        # Tools
+        if definition.tools_commands:
+            content.append("## 🛠️ Specialized Tools/Commands")
+            content.append(definition.tools_commands)
+            content.append("")
+        
+        # Footer
+        content.append("---")
+        content.append(f"**Agent Type**: {definition.metadata.type.value}")
+        content.append(f"**Model Preference**: {definition.metadata.model_preference}")
+        content.append(f"**Version**: {definition.metadata.version}")
+        if definition.metadata.last_updated:
+            content.append(f"**Last Updated**: {definition.metadata.last_updated.strftime('%Y-%m-%d %H:%M:%S')}")
+        
+        # Combine with frontmatter
+        post = frontmatter.Post('\n'.join(content), **frontmatter_data)
+        return frontmatter.dumps(post)
+    
+    def _clear_agent_cache(self, name: str):
+        """Clear cache for a specific agent."""
+        try:
+            cache_key = f"agent_prompt:{name}:md"
+            self.cache.invalidate(cache_key)
+        except Exception as e:
+            logger.warning(f"Failed to clear cache for agent '{name}': {e}")