crackerjack 0.29.0__py3-none-any.whl → 0.31.4__py3-none-any.whl

crackerjack/agents/coordinator.py

@@ -0,0 +1,512 @@
+import asyncio
+import logging
+import operator
+import typing as t
+from collections import defaultdict
+
+from crackerjack.services.debug import get_ai_agent_debugger
+
+from .base import (
+    AgentContext,
+    FixResult,
+    Issue,
+    IssueType,
+    Priority,
+    SubAgent,
+    agent_registry,
+)
+from .tracker import get_agent_tracker
+
+
+class AgentCoordinator:
+    def __init__(self, context: AgentContext) -> None:
+        self.context = context
+        self.agents: list[SubAgent] = []
+        self.logger = logging.getLogger(__name__)
+        self._issue_cache: dict[str, FixResult] = {}
+        self._collaboration_threshold = 0.7
+        self.tracker = get_agent_tracker()
+        self.debugger = get_ai_agent_debugger()
+        self.proactive_mode = True  # Enabled by default
+
+    def initialize_agents(self) -> None:
+        self.agents = agent_registry.create_all(self.context)
+        agent_types = [a.name for a in self.agents]
+        self.logger.info(f"Initialized {len(self.agents)} agents: {agent_types}")
+
+        self.tracker.register_agents(agent_types)
+        self.tracker.set_coordinator_status("active")
+
+        self.debugger.log_agent_activity(
+            agent_name="coordinator",
+            activity="agents_initialized",
+            metadata={"agent_count": len(self.agents), "agent_types": agent_types},
+        )
+
+    async def handle_issues(self, issues: list[Issue]) -> FixResult:
+        if not self.agents:
+            self.initialize_agents()
+
+        if not issues:
+            return FixResult(success=True, confidence=1.0)
+
+        self.logger.info(f"Handling {len(issues)} issues")
+
+        issues_by_type = self._group_issues_by_type(issues)
+
+        overall_result = FixResult(success=True, confidence=1.0)
+
+        for issue_type, type_issues in issues_by_type.items():
+            type_result = await self._handle_issues_by_type(issue_type, type_issues)
+            overall_result = overall_result.merge_with(type_result)
+
+        return overall_result
+
+    async def handle_single_issue(self, issue: Issue) -> FixResult:
+        if not self.agents:
+            self.initialize_agents()
+
+        cache_key = issue.context_key
+        if cache_key in self._issue_cache:
+            cached_result = self._issue_cache[cache_key]
+            self.logger.info(f"Using cached result for {cache_key}")
+            self.tracker.track_cache_hit()
+            return cached_result
+
+        self.tracker.track_cache_miss()
+
+        agent_scores = await self._evaluate_agents_for_issue(issue)
+
+        if not agent_scores:
+            self.logger.warning(f"No agents can handle issue: {issue.message}")
+            return FixResult(
+                success=False,
+                confidence=0.0,
+                remaining_issues=[f"No agent available for: {issue.message}"],
+                recommendations=["Manual intervention required"],
+            )
+
+        best_agent, best_score = agent_scores[0]
+
+        if best_score >= self._collaboration_threshold:
+            result = await self._handle_with_single_agent(best_agent, issue)
+        else:
+            result = await self._handle_with_collaboration(agent_scores[:3], issue)
+
+        self._issue_cache[cache_key] = result
+
+        return result
+
+    async def _handle_issues_by_type(
+        self,
+        issue_type: IssueType,
+        issues: list[Issue],
+    ) -> FixResult:
+        self.logger.info(f"Handling {len(issues)} {issue_type.value} issues")
+
+        specialist_agents = [
+            agent for agent in self.agents if issue_type in agent.get_supported_types()
+        ]
+
+        if not specialist_agents:
+            self.logger.warning(f"No specialist agents for {issue_type.value}")
+            return FixResult(
+                success=False,
+                confidence=0.0,
+                remaining_issues=[f"No agents for {issue_type.value} issues"],
+            )
+
+        tasks: list[t.Coroutine[t.Any, t.Any, FixResult]] = []
+        for issue in issues:
+            best_specialist = await self._find_best_specialist(specialist_agents, issue)
+            if best_specialist:
+                task = self._handle_with_single_agent(best_specialist, issue)
+                tasks.append(task)
+
+        if not tasks:
+            return FixResult(success=False, confidence=0.0)
+
+        results = await asyncio.gather(*tasks, return_exceptions=True)
+
+        combined_result = FixResult(success=True, confidence=1.0)
+        for result in results:
+            if isinstance(result, FixResult):
+                combined_result = combined_result.merge_with(result)
+            else:
+                self.logger.error(f"Agent task failed: {result}")
+                combined_result.success = False
+                combined_result.remaining_issues.append(f"Agent failed: {result}")
+
+        return combined_result
+
+    async def _evaluate_agents_for_issue(
+        self,
+        issue: Issue,
+    ) -> list[tuple[SubAgent, float]]:
+        evaluations: list[tuple[SubAgent, float]] = []
+
+        for agent in self.agents:
+            try:
+                confidence = await agent.can_handle(issue)
+                if confidence > 0.0:
+                    evaluations.append((agent, confidence))
+            except Exception as e:
+                self.logger.exception(f"Error evaluating {agent.name} for issue: {e}")
+
+        evaluations.sort(key=operator.itemgetter(1), reverse=True)
+        return evaluations
+
+    async def _find_best_specialist(
+        self,
+        specialists: list[SubAgent],
+        issue: Issue,
+    ) -> SubAgent | None:
+        best_agent = None
+        best_score = 0.0
+
+        for agent in specialists:
+            try:
+                score = await agent.can_handle(issue)
+                if score > best_score:
+                    best_score = score
+                    best_agent = agent
+            except Exception as e:
+                self.logger.exception(f"Error evaluating specialist {agent.name}: {e}")
+
+        return best_agent
+
+    async def _handle_with_single_agent(
+        self,
+        agent: SubAgent,
+        issue: Issue,
+    ) -> FixResult:
+        self.logger.info(f"Handling issue with {agent.name}: {issue.message[:100]}")
+
+        confidence = await agent.can_handle(issue)
+        self.tracker.track_agent_processing(agent.name, issue, confidence)
+
+        self.debugger.log_agent_activity(
+            agent_name=agent.name,
+            activity="processing_started",
+            issue_id=issue.id,
+            confidence=confidence,
+            metadata={"issue_type": issue.type.value, "severity": issue.severity.value},
+        )
+
+        try:
+            result = await agent.analyze_and_fix(issue)
+            if result.success:
+                self.logger.info(f"{agent.name} successfully fixed issue")
+            else:
+                self.logger.warning(f"{agent.name} failed to fix issue")
+
+            self.tracker.track_agent_complete(agent.name, result)
+
+            self.debugger.log_agent_activity(
+                agent_name=agent.name,
+                activity="processing_completed",
+                issue_id=issue.id,
+                confidence=result.confidence,
+                result={
+                    "success": result.success,
+                    "remaining_issues": len(result.remaining_issues),
+                },
+                metadata={"fix_applied": result.success},
+            )
+
+            return result
+        except Exception as e:
+            self.logger.exception(f"{agent.name} threw exception: {e}")
+            error_result = FixResult(
+                success=False,
+                confidence=0.0,
+                remaining_issues=[f"{agent.name} failed with exception: {e}"],
+            )
+
+            self.tracker.track_agent_complete(agent.name, error_result)
+            return error_result
+
+    async def _handle_with_collaboration(
+        self,
+        agent_scores: list[tuple[SubAgent, float]],
+        issue: Issue,
+    ) -> FixResult:
+        self.logger.info(
+            f"Using collaborative approach for issue: {issue.message[:100]}",
+        )
+
+        results: list[FixResult] = []
+
+        for agent, _ in agent_scores:
+            try:
+                result = await agent.analyze_and_fix(issue)
+                results.append(result)
+
+                if result.success and result.confidence >= 0.8:
+                    self.logger.info(f"{agent.name} solved issue with high confidence")
+                    break
+
+            except Exception as e:
+                self.logger.exception(f"Collaborative agent {agent.name} failed: {e}")
+                results.append(
+                    FixResult(
+                        success=False,
+                        confidence=0.0,
+                        remaining_issues=[f"{agent.name} failed: {e}"],
+                    ),
+                )
+
+        if not results:
+            return FixResult(success=False, confidence=0.0)
+
+        combined_result = results[0]
+        for result in results[1:]:
+            combined_result = combined_result.merge_with(result)
+
+        return combined_result
+
+    def _group_issues_by_type(
+        self,
+        issues: list[Issue],
+    ) -> dict[IssueType, list[Issue]]:
+        grouped: defaultdict[IssueType, list[Issue]] = defaultdict(list)
+        for issue in issues:
+            grouped[issue.type].append(issue)
+        return dict(grouped)
+
+    def get_agent_capabilities(self) -> dict[str, dict[str, t.Any]]:
+        if not self.agents:
+            self.initialize_agents()
+
+        capabilities = {}
+        for agent in self.agents:
+            capabilities[agent.name] = {
+                "supported_types": [t.value for t in agent.get_supported_types()],
+                "class": agent.__class__.__name__,
+            }
+        return capabilities
+
+    def clear_cache(self) -> None:
+        self._issue_cache.clear()
+        self.logger.info("Cleared issue cache")
+
+    async def test_agent_connectivity(self) -> dict[str, bool]:
+        if not self.agents:
+            self.initialize_agents()
+
+        test_issue = Issue(
+            id="test",
+            type=IssueType.FORMATTING,
+            severity=Priority.LOW,
+            message="Test connectivity",
+        )
+
+        connectivity = {}
+        for agent in self.agents:
+            try:
+                confidence = await agent.can_handle(test_issue)
+                connectivity[agent.name] = confidence >= 0.0
+            except Exception as e:
+                self.logger.exception(f"Connectivity test failed for {agent.name}: {e}")
+                connectivity[agent.name] = False
+
+        return connectivity
+
+    async def handle_issues_proactively(self, issues: list[Issue]) -> FixResult:
+        """Handle issues with proactive planning phase."""
+        if not self.proactive_mode:
+            return await self.handle_issues(issues)
+
+        if not self.agents:
+            self.initialize_agents()
+
+        if not issues:
+            return FixResult(success=True, confidence=1.0)
+
+        self.logger.info(f"Handling {len(issues)} issues with proactive planning")
+
+        # Phase 1: Create architectural plan
+        architectural_plan = await self._create_architectural_plan(issues)
+
+        # Phase 2: Apply fixes following the plan
+        overall_result = await self._apply_fixes_with_plan(issues, architectural_plan)
+
+        # Phase 3: Validate against plan
+        validation_result = await self._validate_against_plan(
+            overall_result, architectural_plan
+        )
+
+        return validation_result
+
+    async def _create_architectural_plan(self, issues: list[Issue]) -> dict[str, t.Any]:
+        """Create architectural plan using ArchitectAgent."""
+        architect = self._get_architect_agent()
+
+        if not architect:
+            self.logger.warning("No ArchitectAgent available for planning")
+            return {"strategy": "reactive_fallback", "patterns": []}
+
+        # Analyze all issues together for comprehensive planning
+        complex_issues = [
+            issue
+            for issue in issues
+            if issue.type
+            in {IssueType.COMPLEXITY, IssueType.DRY_VIOLATION, IssueType.PERFORMANCE}
+        ]
+
+        if not complex_issues:
+            return {"strategy": "simple_fixes", "patterns": ["standard_patterns"]}
+
+        # Use the first complex issue for planning (represents the architectural challenge)
+        primary_issue = complex_issues[0]
+
+        try:
+            # Get plan from ArchitectAgent
+            plan = await architect.plan_before_action(primary_issue)
+
+            # Extend plan to cover all issues
+            plan["all_issues"] = [issue.id for issue in issues]
+            plan["issue_types"] = list({issue.type.value for issue in issues})
+
+            self.logger.info(
+                f"Created architectural plan: {plan.get('strategy', 'unknown')}"
+            )
+            return plan
+
+        except Exception as e:
+            self.logger.exception(f"Failed to create architectural plan: {e}")
+            return {"strategy": "reactive_fallback", "patterns": [], "error": str(e)}
+
+    async def _apply_fixes_with_plan(
+        self, issues: list[Issue], plan: dict[str, t.Any]
+    ) -> FixResult:
+        """Apply fixes following the architectural plan."""
+        strategy = plan.get("strategy", "reactive_fallback")
+
+        if strategy == "reactive_fallback":
+            # Fallback to standard processing
+            return await self.handle_issues(issues)
+
+        self.logger.info(f"Applying fixes with {strategy} strategy")
+
+        # Group issues by priority based on plan
+        prioritized_issues = self._prioritize_issues_by_plan(issues, plan)
+
+        overall_result = FixResult(success=True, confidence=1.0)
+
+        for issue_group in prioritized_issues:
+            group_result = await self._handle_issue_group_with_plan(issue_group, plan)
+            overall_result = overall_result.merge_with(group_result)
+
+            # If a critical group fails, consider the overall strategy
+            if not group_result.success and self._is_critical_group(issue_group, plan):
+                overall_result.success = False
+                overall_result.remaining_issues.append(
+                    f"Critical issue group failed: {[i.id for i in issue_group]}"
+                )
+
+        return overall_result
+
+    async def _validate_against_plan(
+        self, result: FixResult, plan: dict[str, t.Any]
+    ) -> FixResult:
+        """Validate the results against the architectural plan."""
+        validation_steps = plan.get("validation", [])
+
+        if not validation_steps:
+            return result
+
+        self.logger.info(f"Validating against plan: {validation_steps}")
+
+        # Add validation recommendations
+        result.recommendations.extend(
+            [
+                f"Validate with: {', '.join(validation_steps)}",
+                f"Applied strategy: {plan.get('strategy', 'unknown')}",
+                f"Used patterns: {', '.join(plan.get('patterns', []))}",
+            ]
+        )
+
+        return result
+
+    def _get_architect_agent(self) -> SubAgent | None:
+        """Get the ArchitectAgent from available agents."""
+        for agent in self.agents:
+            if agent.__class__.__name__ == "ArchitectAgent":
+                return agent
+        return None
+
+    def _prioritize_issues_by_plan(
+        self, issues: list[Issue], plan: dict[str, t.Any]
+    ) -> list[list[Issue]]:
+        """Prioritize issues based on the architectural plan."""
+        strategy = plan.get("strategy", "reactive_fallback")
+
+        if strategy == "external_specialist_guided":
+            # Handle complex issues first, then simple ones
+            complex_issues = [
+                issue
+                for issue in issues
+                if issue.type in {IssueType.COMPLEXITY, IssueType.DRY_VIOLATION}
+            ]
+            other_issues = [issue for issue in issues if issue not in complex_issues]
+            return [complex_issues, other_issues] if complex_issues else [other_issues]
+
+        # Default: group by type for parallel processing
+        groups = self._group_issues_by_type(issues)
+        return list(groups.values())
+
+    async def _handle_issue_group_with_plan(
+        self, issues: list[Issue], plan: dict[str, t.Any]
+    ) -> FixResult:
+        """Handle a group of issues with architectural guidance."""
+        if not issues:
+            return FixResult(success=True, confidence=1.0)
+
+        # Get the best agent for this group, preferring ArchitectAgent for complex issues
+        representative_issue = issues[0]
+
+        if self._should_use_architect_for_group(issues, plan):
+            architect = self._get_architect_agent()
+            if architect:
+                # Use architect for the whole group
+                group_result = FixResult(success=True, confidence=1.0)
+
+                for issue in issues:
+                    issue_result = await architect.analyze_and_fix(issue)
+                    group_result = group_result.merge_with(issue_result)
+
+                return group_result
+
+        # Use standard routing for simpler issues
+        return await self._handle_issues_by_type(representative_issue.type, issues)
+
+    def _should_use_architect_for_group(
+        self, issues: list[Issue], plan: dict[str, t.Any]
+    ) -> bool:
+        """Determine if ArchitectAgent should handle this group."""
+        strategy = plan.get("strategy", "")
+
+        # Always use architect for specialist-guided strategies
+        if strategy == "external_specialist_guided":
+            return True
+
+        # Use architect for complex or architectural issues
+        architectural_types = {
+            IssueType.COMPLEXITY,
+            IssueType.DRY_VIOLATION,
+            IssueType.PERFORMANCE,
+        }
+
+        return any(issue.type in architectural_types for issue in issues)
+
+    def _is_critical_group(self, issues: list[Issue], plan: dict[str, t.Any]) -> bool:
+        """Determine if this issue group is critical to the plan."""
+        # Complex and DRY violations are typically critical
+        critical_types = {IssueType.COMPLEXITY, IssueType.DRY_VIOLATION}
+        return any(issue.type in critical_types for issue in issues)
+
+    def set_proactive_mode(self, enabled: bool) -> None:
+        """Enable or disable proactive planning mode."""
+        self.proactive_mode = enabled
+        self.logger.info(f"Proactive mode {'enabled' if enabled else 'disabled'}")