claude-mpm 3.4.10__py3-none-any.whl → 5.4.85__py3-none-any.whl

claude_mpm/services/agents/recommender.py

@@ -0,0 +1,617 @@
+"""
+Agent Recommender Service for Claude MPM Framework
+===================================================
+
+WHY: Automated agent recommendation is critical for the auto-configuration feature.
+This service analyzes project toolchains and recommends appropriate specialized agents
+using configuration-driven mappings and intelligent scoring algorithms.
+
+DESIGN DECISION: Configuration-driven approach using YAML for flexibility and
+maintainability. Scoring algorithm weighs language, framework, and deployment matches
+with configurable weights. Returns ranked recommendations with detailed reasoning.
+
+Part of TSK-0054: Auto-Configuration Feature - Phase 3
+"""
+
+from pathlib import Path
+from typing import Any, Dict, List, Optional
+
+import yaml
+
+from claude_mpm.core.base_service import BaseService
+from claude_mpm.services.core.interfaces.agent import IAgentRecommender
+from claude_mpm.services.core.models.agent_config import (
+    AgentCapabilities,
+    AgentRecommendation,
+    AgentSpecialization,
+)
+from claude_mpm.services.core.models.toolchain import ToolchainAnalysis
+
+
+class AgentRecommenderService(BaseService, IAgentRecommender):
+    """
+    Service for recommending agents based on toolchain analysis.
+
+    This service:
+    - Loads agent capabilities from YAML configuration
+    - Calculates match scores between toolchains and agents
+    - Returns ranked recommendations with confidence scores
+    - Provides detailed reasoning for each recommendation
+
+    Matching Algorithm:
+    - Primary language match (50% weight)
+    - Framework match (30% weight)
+    - Deployment target match (20% weight)
+    - Apply agent-specific confidence weights
+    - Framework-specific agents override language agents
+    """
+
+    def __init__(
+        self,
+        config_path: Optional[Path] = None,
+        config: Optional[Dict[str, Any]] = None,
+        container: Optional[Any] = None,
+    ):
+        """
+        Initialize the Agent Recommender Service.
+
+        Args:
+            config_path: Optional path to agent_capabilities.yaml
+            config: Optional configuration dictionary
+            container: Optional service container for dependency injection
+        """
+        super().__init__(
+            name="AgentRecommenderService",
+            config=config,
+            enable_enhanced_features=False,
+            container=container,
+        )
+
+        # Determine config file path
+        if config_path is None:
+            # Default to config directory in the package
+            package_config_dir = Path(__file__).parent.parent.parent / "config"
+            config_path = package_config_dir / "agent_capabilities.yaml"
+
+        self.config_path = config_path
+        self._capabilities_config: Dict[str, Any] = {}
+        self._agent_capabilities_cache: Dict[str, AgentCapabilities] = {}
+        self._load_configuration()
+
+        self.logger.info(
+            f"AgentRecommenderService initialized with config: {self.config_path}"
+        )
+
+    async def _initialize(self) -> None:
+        """Initialize the service (required by BaseService)."""
+        # Configuration is loaded in __init__, no additional initialization needed
+
+    async def _cleanup(self) -> None:
+        """Cleanup service resources (required by BaseService)."""
+        # Clear caches
+        self._agent_capabilities_cache.clear()
+        self._capabilities_config.clear()
+
+    def _load_configuration(self) -> None:
+        """
+        Load agent capabilities configuration from YAML file.
+
+        Raises:
+            FileNotFoundError: If configuration file does not exist
+            yaml.YAMLError: If configuration file is invalid YAML
+        """
+        if not self.config_path.exists():
+            raise FileNotFoundError(
+                f"Agent capabilities config not found: {self.config_path}"
+            )
+
+        try:
+            with self.config_path.open(encoding="utf-8") as f:
+                self._capabilities_config = yaml.safe_load(f)
+
+            if (
+                not self._capabilities_config
+                or "agent_capabilities" not in self._capabilities_config
+            ):
+                raise ValueError(
+                    "Invalid configuration: missing 'agent_capabilities' section"
+                )
+
+            self.logger.info(
+                f"Loaded {len(self._capabilities_config['agent_capabilities'])} "
+                f"agent capability definitions"
+            )
+
+        except yaml.YAMLError as e:
+            self.logger.error(f"Failed to parse YAML configuration: {e}")
+            raise
+
+    def recommend_agents(
+        self,
+        toolchain: ToolchainAnalysis,
+        constraints: Optional[Dict[str, Any]] = None,
+    ) -> List[AgentRecommendation]:
+        """
+        Recommend agents based on toolchain analysis.
+
+        Analyzes the toolchain and recommends agents that best match the
+        project's technical requirements. Considers:
+        - Language compatibility
+        - Framework expertise
+        - Deployment environment requirements
+        - Optional user-defined constraints
+
+        Args:
+            toolchain: Complete toolchain analysis results
+            constraints: Optional constraints for recommendations:
+                - max_agents: Maximum number of agents to recommend
+                - required_capabilities: List of required agent capabilities
+                - excluded_agents: List of agent IDs to exclude
+                - min_confidence: Minimum confidence score threshold
+
+        Returns:
+            List[AgentRecommendation]: Ordered list of recommended agents
+                with confidence scores and reasoning
+
+        Raises:
+            ValueError: If constraints are invalid or contradictory
+        """
+        constraints = constraints or {}
+        min_confidence = constraints.get(
+            "min_confidence",
+            self._capabilities_config.get("recommendation_rules", {}).get(
+                "min_confidence_threshold", 0.5
+            ),
+        )
+        excluded_agents = set(constraints.get("excluded_agents", []))
+        max_agents = constraints.get("max_agents")
+
+        recommendations: List[AgentRecommendation] = []
+        agent_configs = self._capabilities_config.get("agent_capabilities", {})
+
+        # Calculate scores for all agents
+        for agent_id, agent_config in agent_configs.items():
+            # Skip excluded agents
+            if agent_id in excluded_agents:
+                continue
+
+            # Skip agents that shouldn't auto-deploy unless explicitly allowed
+            if not agent_config.get("auto_deploy", True) and not constraints.get(
+                "include_non_auto_deploy", False
+            ):
+                continue
+
+            # Calculate match score
+            score = self.match_score(agent_id, toolchain)
+
+            # Filter by minimum confidence
+            if score < min_confidence:
+                continue
+
+            # Get agent capabilities
+            capabilities = self.get_agent_capabilities(agent_id)
+
+            # Generate match reasons and concerns
+            match_reasons, concerns = self._generate_reasoning(
+                agent_id, agent_config, toolchain, score
+            )
+
+            # Determine deployment priority
+            deployment_priority = self._calculate_deployment_priority(
+                agent_config, toolchain
+            )
+
+            recommendation = AgentRecommendation(
+                agent_id=agent_id,
+                agent_name=agent_config.get("name", agent_id),
+                confidence_score=score,
+                match_reasons=match_reasons,
+                concerns=concerns,
+                capabilities=capabilities,
+                deployment_priority=deployment_priority,
+                configuration_hints=self._generate_config_hints(
+                    agent_config, toolchain
+                ),
+                metadata={
+                    "specialization": agent_config.get("specialization"),
+                    "auto_deploy": agent_config.get("auto_deploy", True),
+                },
+            )
+            recommendations.append(recommendation)
+
+        # Sort by confidence score (descending), then by deployment priority (ascending)
+        recommendations.sort(key=lambda r: (-r.confidence_score, r.deployment_priority))
+
+        # Apply max agents limit if specified
+        if max_agents is not None:
+            recommendations = recommendations[:max_agents]
+
+        # Check if we have no recommendations (any reason: unknown language, low scores, etc.)
+        if not recommendations:
+            self.logger.info(
+                f"No agents scored above threshold for {toolchain.primary_language}; using defaults"
+            )
+
+            # Get default configuration
+            default_config = self._capabilities_config.get("default_configuration", {})
+            if default_config.get("enabled", False):
+                default_agents = default_config.get("agents", [])
+                default_confidence = default_config.get("min_confidence", 0.7)
+
+                for default_agent in default_agents:
+                    agent_id = default_agent.get("agent_id")
+
+                    # Skip if agent doesn't exist in capabilities
+                    if agent_id not in agent_configs:
+                        self.logger.warning(f"Default agent not found: {agent_id}")
+                        continue
+
+                    agent_config = agent_configs[agent_id]
+                    capabilities = self.get_agent_capabilities(agent_id)
+
+                    recommendation = AgentRecommendation(
+                        agent_id=agent_id,
+                        agent_name=agent_config.get("name", agent_id),
+                        confidence_score=default_confidence,
+                        match_reasons=[
+                            "Default configuration applied - toolchain detection unsuccessful",
+                            default_agent.get("reasoning", "General-purpose agent"),
+                        ],
+                        concerns=[
+                            "Consider manually selecting specialized agents for better results"
+                        ],
+                        capabilities=capabilities,
+                        deployment_priority=default_agent.get("priority", 10),
+                        configuration_hints={"default_deployment": True},
+                        metadata={
+                            "is_default": True,
+                            "specialization": agent_config.get("specialization"),
+                            "auto_deploy": agent_config.get("auto_deploy", True),
+                        },
+                    )
+                    recommendations.append(recommendation)
+
+                self.logger.info(
+                    f"Applied {len(recommendations)} default agent recommendations"
+                )
+
+        self.logger.info(
+            f"Generated {len(recommendations)} agent recommendations "
+            f"for project: {toolchain.project_path}"
+        )
+
+        return recommendations
+
+    def get_agent_capabilities(self, agent_id: str) -> AgentCapabilities:
+        """
+        Get detailed capabilities for an agent.
+
+        Retrieves comprehensive capability information for a specific agent:
+        - Supported languages and frameworks
+        - Specialization areas
+        - Required toolchain components
+        - Performance characteristics
+
+        Args:
+            agent_id: Unique identifier of the agent
+
+        Returns:
+            AgentCapabilities: Complete capability information
+
+        Raises:
+            KeyError: If agent_id does not exist
+        """
+        # Check cache first
+        if agent_id in self._agent_capabilities_cache:
+            return self._agent_capabilities_cache[agent_id]
+
+        # Get from config
+        agent_configs = self._capabilities_config.get("agent_capabilities", {})
+        if agent_id not in agent_configs:
+            raise KeyError(f"Agent not found: {agent_id}")
+
+        agent_config = agent_configs[agent_id]
+        supports = agent_config.get("supports", {})
+
+        # Determine specializations
+        specializations = []
+        spec_type = agent_config.get("specialization", "general")
+        if spec_type == "engineering":
+            specializations.append(AgentSpecialization.LANGUAGE_SPECIFIC)
+        elif spec_type == "devops":
+            specializations.append(AgentSpecialization.DEVOPS)
+
+        # Create capabilities object
+        capabilities = AgentCapabilities(
+            agent_id=agent_id,
+            agent_name=agent_config.get("name", agent_id),
+            specializations=specializations,
+            supported_languages=supports.get("languages", []),
+            supported_frameworks=supports.get("frameworks", []),
+            required_tools=supports.get("build_tools", []),
+            deployment_targets=supports.get("deployment", []),
+            description=agent_config.get("description", ""),
+            metadata=agent_config.get("metadata", {}),
+        )
+
+        # Cache for future use
+        self._agent_capabilities_cache[agent_id] = capabilities
+
+        return capabilities
+
+    def match_score(self, agent_id: str, toolchain: ToolchainAnalysis) -> float:
+        """
+        Calculate match score between agent and toolchain.
+
+        Computes a numerical score (0.0 to 1.0) indicating how well an agent
+        matches the project's toolchain. Higher scores indicate better matches.
+        Considers:
+        - Language compatibility (50% weight)
+        - Framework experience (30% weight)
+        - Deployment target alignment (20% weight)
+        - Toolchain component coverage
+
+        Args:
+            agent_id: Unique identifier of the agent
+            toolchain: Complete toolchain analysis
+
+        Returns:
+            float: Match score between 0.0 (no match) and 1.0 (perfect match)
+
+        Raises:
+            KeyError: If agent_id does not exist
+        """
+        agent_configs = self._capabilities_config.get("agent_capabilities", {})
+        if agent_id not in agent_configs:
+            raise KeyError(f"Agent not found: {agent_id}")
+
+        agent_config = agent_configs[agent_id]
+        supports = agent_config.get("supports", {})
+
+        # Get scoring weights
+        scoring_weights = self._capabilities_config.get("recommendation_rules", {}).get(
+            "scoring_weights", {}
+        )
+        language_weight = scoring_weights.get("language_match", 0.5)
+        framework_weight = scoring_weights.get("framework_match", 0.3)
+        deployment_weight = scoring_weights.get("deployment_match", 0.2)
+
+        # Calculate language match score
+        language_score = self._calculate_language_score(
+            supports.get("languages", []), toolchain
+        )
+
+        # Calculate framework match score
+        framework_score = self._calculate_framework_score(
+            supports.get("frameworks", []), toolchain
+        )
+
+        # Calculate deployment match score
+        deployment_score = self._calculate_deployment_score(
+            supports.get("deployment", []), toolchain
+        )
+
+        # Calculate base score
+        base_score = (
+            language_score * language_weight
+            + framework_score * framework_weight
+            + deployment_score * deployment_weight
+        )
+
+        # Apply agent confidence weight
+        agent_confidence_weight = agent_config.get("confidence_weight", 1.0)
+        final_score = base_score * agent_confidence_weight
+
+        # Apply framework priority boost if applicable
+        if framework_score > 0.5:  # Strong framework match
+            framework_boost = self._capabilities_config.get(
+                "recommendation_rules", {}
+            ).get("framework_priority_boost", 0.15)
+            final_score = min(1.0, final_score + framework_boost)
+
+        # Apply deployment match boost if applicable
+        if deployment_score > 0.5:  # Strong deployment match
+            deployment_boost = self._capabilities_config.get(
+                "recommendation_rules", {}
+            ).get("deployment_match_boost", 0.1)
+            final_score = min(1.0, final_score + deployment_boost)
+
+        # Ensure score is in valid range and return
+        return max(0.0, min(1.0, final_score))
+
+    def _calculate_language_score(
+        self, supported_languages: List[str], toolchain: ToolchainAnalysis
+    ) -> float:
+        """Calculate language match score."""
+        if not supported_languages:
+            return 0.0
+
+        primary_language = toolchain.primary_language.lower()
+        all_languages = [lang.lower() for lang in toolchain.all_languages]
+
+        # Check for primary language match
+        for lang in supported_languages:
+            if lang.lower() == primary_language:
+                return 1.0  # Perfect match on primary language
+
+        # Check for secondary language match
+        for lang in supported_languages:
+            if lang.lower() in all_languages:
+                return 0.6  # Partial match on secondary language
+
+        return 0.0
+
+    def _calculate_framework_score(
+        self, supported_frameworks: List[str], toolchain: ToolchainAnalysis
+    ) -> float:
+        """Calculate framework match score."""
+        if not supported_frameworks:
+            return 0.0
+
+        detected_frameworks = [fw.name.lower() for fw in toolchain.frameworks]
+        if not detected_frameworks:
+            return 0.0
+
+        # Check for exact framework matches
+        matches = 0
+        for fw in supported_frameworks:
+            fw_lower = fw.lower()
+            # Normalize common variants (next.js vs nextjs, etc.)
+            fw_normalized = fw_lower.replace(".", "").replace("-", "")
+            for detected_fw in detected_frameworks:
+                detected_normalized = detected_fw.replace(".", "").replace("-", "")
+                if fw_normalized == detected_normalized:
+                    matches += 1
+                    break
+
+        if matches == 0:
+            return 0.0
+
+        # Calculate score based on match proportion
+        match_ratio = matches / len(detected_frameworks)
+        return min(1.0, match_ratio * 1.2)  # Boost for framework matches
+
+    def _calculate_deployment_score(
+        self, supported_deployments: List[str], toolchain: ToolchainAnalysis
+    ) -> float:
+        """Calculate deployment target match score."""
+        if not supported_deployments or not toolchain.deployment_target:
+            return 0.0
+
+        target_platform = toolchain.deployment_target.platform
+        target_type = toolchain.deployment_target.target_type
+
+        # Check for exact platform match
+        for deployment in supported_deployments:
+            deployment_lower = deployment.lower()
+            if target_platform and deployment_lower == target_platform.lower():
+                return 1.0  # Perfect match on platform
+            if deployment_lower == target_type.lower():
+                return 0.8  # Good match on target type
+
+        # Check for container/cloud general matches
+        if any(d.lower() in ["docker", "kubernetes"] for d in supported_deployments):
+            if target_type in ["container", "cloud"]:
+                return 0.5  # General container/cloud match
+
+        return 0.0
+
+    def _generate_reasoning(
+        self,
+        agent_id: str,
+        agent_config: Dict[str, Any],
+        toolchain: ToolchainAnalysis,
+        score: float,
+    ) -> tuple[List[str], List[str]]:
+        """
+        Generate match reasons and concerns for recommendation.
+
+        Returns:
+            Tuple of (match_reasons, concerns)
+        """
+        match_reasons = []
+        concerns = []
+        supports = agent_config.get("supports", {})
+
+        # Check language match
+        primary_language = toolchain.primary_language.lower()
+        supported_languages = [lang.lower() for lang in supports.get("languages", [])]
+        if primary_language in supported_languages:
+            match_reasons.append(
+                f"Primary language match: {toolchain.primary_language}"
+            )
+
+        # Check framework matches
+        detected_frameworks = [fw.name for fw in toolchain.frameworks]
+        supported_frameworks = supports.get("frameworks", [])
+        framework_matches = []
+        for fw in detected_frameworks:
+            for supported_fw in supported_frameworks:
+                if fw.lower().replace("-", "").replace(
+                    ".", ""
+                ) == supported_fw.lower().replace("-", "").replace(".", ""):
+                    framework_matches.append(fw)
+                    break
+
+        if framework_matches:
+            match_reasons.append(f"Framework expertise: {', '.join(framework_matches)}")
+
+        # Check deployment target match
+        if toolchain.deployment_target:
+            supported_deployments = supports.get("deployment", [])
+            platform = toolchain.deployment_target.platform
+            if platform and any(
+                d.lower() == platform.lower() for d in supported_deployments
+            ):
+                match_reasons.append(f"Deployment platform match: {platform}")
+
+        # Add concerns for low/medium confidence
+        if score < 0.6:
+            concerns.append(
+                "Moderate confidence - consider reviewing agent capabilities"
+            )
+        elif score < 0.8:
+            concerns.append("Good match but may benefit from additional configuration")
+
+        # If no specific matches found but score > 0, add general reason
+        if not match_reasons and score > 0:
+            match_reasons.append(
+                f"General {agent_config.get('specialization', 'agent')} capabilities"
+            )
+
+        return match_reasons, concerns
+
+    def _calculate_deployment_priority(
+        self, agent_config: Dict[str, Any], toolchain: ToolchainAnalysis
+    ) -> int:
+        """
+        Calculate deployment priority (lower = higher priority).
+
+        Priority rules:
+        - Framework-specific agents: Priority 1
+        - Language-specific engineers: Priority 2
+        - DevOps agents: Priority 3
+        - General agents: Priority 4
+        """
+        specialization = agent_config.get("specialization", "general")
+        supports = agent_config.get("supports", {})
+
+        # Check if this is a framework-specific agent
+        framework_matches = False
+        if toolchain.frameworks:
+            detected_frameworks = [fw.name.lower() for fw in toolchain.frameworks]
+            supported_frameworks = [fw.lower() for fw in supports.get("frameworks", [])]
+            framework_matches = any(
+                fw in supported_frameworks for fw in detected_frameworks
+            )
+
+        if framework_matches and specialization == "engineering":
+            return 1  # Highest priority for framework-specific engineers
+
+        if specialization == "engineering":
+            return 2  # Language-specific engineers
+
+        if specialization == "devops":
+            return 3  # DevOps agents
+
+        return 4  # General agents (lowest priority)
+
+    def _generate_config_hints(
+        self, agent_config: Dict[str, Any], toolchain: ToolchainAnalysis
+    ) -> Dict[str, Any]:
+        """Generate configuration hints for the agent."""
+        hints = {}
+
+        # Add framework-specific hints
+        if toolchain.frameworks:
+            hints["detected_frameworks"] = [fw.name for fw in toolchain.frameworks]
+
+        # Add deployment hints
+        if toolchain.deployment_target:
+            hints["deployment_target"] = toolchain.deployment_target.platform
+
+        # Add build tool hints
+        if toolchain.build_tools:
+            hints["build_tools"] = [tool.name for tool in toolchain.build_tools]
+
+        return hints

claude_mpm/services/agents/registry/__init__.py

@@ -0,0 +1,30 @@
+"""Agent registry services for discovery and tracking."""
+
+from claude_mpm.core.unified_agent_registry import (
+    AgentMetadata,
+    AgentTier,
+    AgentType,
+    UnifiedAgentRegistry as AgentRegistry,
+)
+
+from .deployed_agent_discovery import DeployedAgentDiscovery
+from .modification_tracker import (
+    AgentModification,
+    AgentModificationTracker,
+    ModificationHistory,
+    ModificationTier,
+    ModificationType,
+)
+
+__all__ = [
+    "AgentMetadata",
+    "AgentModification",
+    "AgentModificationTracker",
+    "AgentRegistry",
+    "AgentTier",
+    "AgentType",
+    "DeployedAgentDiscovery",
+    "ModificationHistory",
+    "ModificationTier",
+    "ModificationType",
+]