claude-mpm 4.24.0__py3-none-any.whl → 5.0.9__py3-none-any.whl

claude_mpm/services/agents/agent_builder.py

@@ -8,12 +8,13 @@ This service provides comprehensive agent lifecycle management including:
 - Integration with deployment services
 """
 
-import json
 import re
 from datetime import datetime, timezone
 from pathlib import Path
 from typing import Any, ClassVar, Dict, List, Optional, Tuple
 
+import yaml
+
 from claude_mpm.core.exceptions import AgentDeploymentError
 from claude_mpm.core.logging_config import get_logger
 
@@ -285,10 +286,15 @@ class AgentBuilderService:
         if not self.templates_dir.exists():
             return templates
 
-        for template_file in self.templates_dir.glob("*.json"):
+        # Agent templates migrated to Markdown with YAML frontmatter (v4.26.0+)
+        for template_file in self.templates_dir.glob("*.md"):
             try:
-                with template_file.open() as f:
-                    config = json.load(f)
+                # Read markdown content and extract YAML frontmatter
+                content = template_file.read_text()
+                config = self._extract_yaml_frontmatter(content)
+                if not config:
+                    self.logger.warning(f"No YAML frontmatter in {template_file.name}")
+                    continue
 
                 # Use filename stem as ID if not specified in config
                 template_id = config.get("id") or template_file.stem
@@ -308,6 +314,30 @@ class AgentBuilderService:
 
         return templates
 
+    def _extract_yaml_frontmatter(self, content: str) -> Optional[Dict[str, Any]]:
+        """
+        Extract and parse YAML frontmatter from markdown content.
+
+        Args:
+            content: File content to parse
+
+        Returns:
+            Parsed YAML frontmatter as dict, or None if not found/invalid
+        """
+        if not content.strip().startswith("---"):
+            return None
+
+        # Split on --- delimiters
+        parts = content.split("---", 2)
+        if len(parts) < 3:
+            return None
+
+        try:
+            return yaml.safe_load(parts[1])
+        except yaml.YAMLError as e:
+            self.logger.warning(f"Failed to parse YAML frontmatter: {e}")
+            return None
+
     def _validate_agent_id(self, agent_id: str) -> None:
         """Validate agent ID format.
 
@@ -371,16 +401,22 @@ class AgentBuilderService:
         if template_id in self._template_cache:
             return self._template_cache[template_id].copy()
 
-        template_file = self.templates_dir / f"{template_id}.json"
+        # Agent templates migrated to Markdown with YAML frontmatter (v4.26.0+)
+        template_file = self.templates_dir / f"{template_id}.md"
 
         if not template_file.exists():
             raise AgentDeploymentError(f"Template '{template_id}' not found")
 
         try:
-            with template_file.open() as f:
-                config = json.load(f)
-                self._template_cache[template_id] = config
-                return config.copy()
+            # Read markdown content and extract YAML frontmatter
+            content = template_file.read_text()
+            config = self._extract_yaml_frontmatter(content)
+            if not config:
+                raise AgentDeploymentError(
+                    f"No YAML frontmatter in template '{template_id}'"
+                )
+            self._template_cache[template_id] = config
+            return config.copy()
         except Exception as e:
             raise AgentDeploymentError(
                 f"Failed to load template '{template_id}': {e}"

claude_mpm/services/agents/agent_preset_service.py

@@ -0,0 +1,238 @@
+"""Service for managing agent preset configurations.
+
+This service resolves preset names to concrete agent lists, validates preset
+availability against configured sources, and provides metadata for CLI display.
+
+Design Decision: Separate service layer for presets
+Rationale: Keeps CLI commands thin, enables easy testing, allows preset logic
+to be reused across multiple command interfaces.
+
+Trade-offs:
+- Abstraction: Added layer vs. direct preset access in CLI
+- Testability: Easy to mock service in CLI tests
+- Reusability: Preset logic available to web UI, API, etc.
+"""
+
+from typing import Any, Dict, List
+
+from ...config.agent_presets import (
+    PRESETS,
+    get_preset_agents,
+    get_preset_info,
+    get_preset_names,
+)
+from ...core.logging_config import get_logger
+from .git_source_manager import GitSourceManager
+
+logger = get_logger(__name__)
+
+
+class AgentPresetService:
+    """Service for resolving and validating agent presets.
+
+    This service bridges preset definitions and git source management,
+    providing a unified interface for preset-based agent deployment.
+
+    Attributes:
+        source_manager: GitSourceManager for agent resolution
+    """
+
+    def __init__(self, source_manager: GitSourceManager):
+        """Initialize with GitSourceManager for agent resolution.
+
+        Args:
+            source_manager: GitSourceManager instance for accessing agent sources
+        """
+        self.source_manager = source_manager
+
+    def list_presets(self) -> List[Dict[str, Any]]:
+        """List all available presets with metadata.
+
+        Returns:
+            List of preset info dicts with keys:
+            - name: Preset name (str)
+            - description: Human-readable description (str)
+            - agent_count: Number of agents in preset (int)
+            - use_cases: List of use case strings (List[str])
+
+        Example:
+            >>> service = AgentPresetService(source_manager)
+            >>> presets = service.list_presets()
+            >>> len(presets) >= 10
+            True
+            >>> all('name' in p for p in presets)
+            True
+        """
+        return [get_preset_info(name) for name in get_preset_names()]
+
+    def validate_preset(self, preset_name: str) -> bool:
+        """Check if preset name is valid.
+
+        Args:
+            preset_name: Name of preset to validate
+
+        Returns:
+            True if preset exists, False otherwise
+
+        Example:
+            >>> service.validate_preset('minimal')
+            True
+            >>> service.validate_preset('invalid-preset')
+            False
+        """
+        return preset_name in PRESETS
+
+    def get_preset_agents(self, preset_name: str) -> List[str]:
+        """Get agent list for preset.
+
+        Args:
+            preset_name: Name of preset
+
+        Returns:
+            List of agent IDs (e.g., ["universal/memory-manager", ...])
+
+        Raises:
+            ValueError: If preset name invalid
+
+        Example:
+            >>> agents = service.get_preset_agents('minimal')
+            >>> len(agents)
+            6
+        """
+        return get_preset_agents(preset_name)
+
+    def resolve_agents(
+        self, preset_name: str, validate_availability: bool = True
+    ) -> Dict[str, Any]:
+        """Resolve preset to actual agent definitions from configured sources.
+
+        This method cross-references preset agent IDs with available agents
+        from configured git sources, detecting missing agents and conflicts.
+
+        Args:
+            preset_name: Name of preset
+            validate_availability: Check if agents exist in sources
+
+        Returns:
+            Dict with keys:
+            - preset_info: Preset metadata (dict)
+            - agents: List of agent dicts with source attribution (List[dict])
+            - missing_agents: List of agent IDs not found (List[str])
+            - conflicts: List of agents found in multiple sources (List[dict])
+
+        Raises:
+            ValueError: If preset name invalid
+
+        Example:
+            >>> result = service.resolve_agents('minimal', validate_availability=True)
+            >>> 'preset_info' in result
+            True
+            >>> 'agents' in result
+            True
+            >>> 'missing_agents' in result
+            True
+
+        Design Decision: Graceful degradation for missing agents
+        Rationale: Allow deployment to proceed with warnings rather than
+        failing completely. Users may have subset of sources configured.
+        """
+        logger.debug(f"[DEBUG] resolve_agents START: preset_name={preset_name}")
+
+        if not self.validate_preset(preset_name):
+            raise ValueError(f"Unknown preset: {preset_name}")
+
+        logger.debug("[DEBUG] Validation passed")
+        preset_info = get_preset_info(preset_name)
+        logger.debug(f"[DEBUG] Got preset info: {preset_info.get('name')}")
+
+        agent_ids = self.get_preset_agents(preset_name)
+        logger.debug(f"[DEBUG] Got agent IDs: {len(agent_ids)} agents")
+
+        result = {
+            "preset_info": preset_info,
+            "agents": [],
+            "missing_agents": [],
+            "conflicts": [],
+        }
+
+        if not validate_availability:
+            # Return agent IDs without validation
+            logger.debug("[DEBUG] Skipping validation, returning agent IDs only")
+            result["agents"] = [{"agent_id": aid} for aid in agent_ids]
+            return result
+
+        # Get all cached agents from sources
+        logger.debug("[DEBUG] Calling source_manager.list_cached_agents()")
+        all_agents = self.source_manager.list_cached_agents()
+        logger.debug(f"[DEBUG] Got {len(all_agents)} agents from source manager")
+
+        # Build lookup: agent_id -> list of sources
+        # Each entry is (source_identifier, agent_data)
+        logger.debug("[DEBUG] Building agent_sources lookup table")
+        agent_sources: Dict[str, List[tuple]] = {}
+        for idx, agent in enumerate(all_agents):
+            if idx % 10 == 0:
+                logger.debug(f"[DEBUG] Processing agent {idx}/{len(all_agents)}")
+            # Handle both 'agent_id' and 'id' keys for compatibility
+            agent_id = agent.get("agent_id") or agent.get("id")
+            if not agent_id:
+                logger.warning(f"Agent missing agent_id field: {agent.keys()}")
+                continue
+
+            # Bug #5 fix: Handle both string and dict formats for 'source' field
+            # GitSourceManager sets source as string (repo_identifier)
+            # Some agents may have source as dict with 'identifier' key
+            source_value = agent.get("source", "unknown")
+            if isinstance(source_value, dict):
+                source = source_value.get("identifier", "unknown")
+            else:
+                source = source_value  # Already a string
+
+            if agent_id not in agent_sources:
+                agent_sources[agent_id] = []
+            agent_sources[agent_id].append((source, agent))
+
+        logger.debug(
+            f"[DEBUG] Built lookup table with {len(agent_sources)} unique agents"
+        )
+
+        # Resolve each agent in preset
+        logger.debug(f"[DEBUG] Resolving {len(agent_ids)} preset agents")
+        for idx, agent_id in enumerate(agent_ids):
+            logger.debug(
+                f"[DEBUG] Resolving agent {idx + 1}/{len(agent_ids)}: {agent_id}"
+            )
+            if agent_id not in agent_sources:
+                logger.debug(f"[DEBUG] Agent {agent_id} not found in sources (missing)")
+                result["missing_agents"].append(agent_id)
+                continue
+
+            sources = agent_sources[agent_id]
+            logger.debug(f"[DEBUG] Agent {agent_id} found in {len(sources)} source(s)")
+
+            # Conflict if agent found in multiple sources
+            if len(sources) > 1:
+                logger.debug(
+                    f"[DEBUG] Agent {agent_id} has conflict (multiple sources)"
+                )
+                result["conflicts"].append(
+                    {
+                        "agent_id": agent_id,
+                        "sources": [s[0] for s in sources],
+                    }
+                )
+
+            # Use first source (highest priority)
+            source_name, agent_data = sources[0]
+            result["agents"].append(
+                {
+                    "agent_id": agent_id,
+                    "source": source_name,
+                    "metadata": agent_data,
+                }
+            )
+
+        logger.debug(
+            f"[DEBUG] resolve_agents COMPLETE: {len(result['agents'])} resolved"
+        )
+        return result