moai-adk 0.35.1__py3-none-any.whl

moai_adk/core/analysis/session_analyzer.py

@@ -0,0 +1,400 @@
+"""
+MoAI-ADK Session Analyzer
+
+Analyzes Claude Code session logs to generate data-driven improvement suggestions
+
+This module provides the SessionAnalyzer class for analyzing Claude Code session logs
+and generating improvement suggestions based on usage patterns.
+"""
+
+import json
+from collections import defaultdict
+from datetime import datetime, timedelta
+from pathlib import Path
+from typing import Any, Dict, Optional, cast
+
+
+class SessionAnalyzer:
+    """Claude Code session log analyzer"""
+
+    def __init__(self, days_back: int = 7, verbose: bool = False):
+        """
+        Initialize SessionAnalyzer
+
+        Args:
+            days_back: Number of days to analyze (default: 7)
+            verbose: Enable verbose output (default: False)
+        """
+        self.claude_projects = Path.home() / ".claude" / "projects"
+        self.days_back = days_back
+        self.verbose = verbose
+
+        self.patterns = {
+            "total_sessions": 0,
+            "total_events": 0,
+            "tool_usage": defaultdict(int),
+            "tool_failures": defaultdict(int),
+            "error_patterns": defaultdict(int),
+            "permission_requests": defaultdict(int),
+            "hook_failures": defaultdict(int),
+            "command_frequency": defaultdict(int),
+            "average_session_length": 0,
+            "success_rate": 0.0,
+            "failed_sessions": 0,
+        }
+
+        self.sessions_data: list[Dict[str, Any]] = []
+
+    def parse_sessions(self) -> Dict[str, Any]:
+        """
+        Parse all session logs from the last N days
+
+        Returns:
+            Dictionary containing analysis patterns and metrics
+        """
+        if not self.claude_projects.exists():
+            if self.verbose:
+                print(f"⚠️ Claude projects directory not found: {self.claude_projects}")
+            return self.patterns
+
+        cutoff_date = datetime.now() - timedelta(days=self.days_back)
+
+        # Look for both session-*.json and UUID.jsonl files
+        session_files: list[Path] = []
+        session_files.extend(self.claude_projects.glob("*/session-*.json"))
+        session_files.extend(self.claude_projects.glob("*/*.jsonl"))
+
+        if self.verbose:
+            print(f"Found {len(session_files)} session files")
+
+        for session_file in session_files:
+            # Check file modification time
+            if datetime.fromtimestamp(session_file.stat().st_mtime) < cutoff_date:
+                continue
+
+            try:
+                # Handle both JSON and JSONL formats
+                if session_file.suffix == ".jsonl":
+                    # JSONL format: read line by line
+                    sessions = []
+                    with open(session_file, encoding="utf-8") as f:
+                        for line_num, line in enumerate(f, 1):
+                            line = line.strip()
+                            if line:
+                                try:
+                                    session = json.loads(line)
+                                    sessions.append(session)
+                                except json.JSONDecodeError as e:
+                                    if self.verbose:
+                                        print(f"⚠️ Error reading line {line_num} in {session_file}: {e}")
+
+                    # Analyze each session from the JSONL file
+                    for session in sessions:
+                        self._analyze_session(session)
+                        self.sessions_data.append(session)
+                else:
+                    # JSON format: single session per file
+                    with open(session_file, encoding="utf-8") as f:
+                        session = json.load(f)
+                        self._analyze_session(session)
+                        self.sessions_data.append(session)
+            except (json.JSONDecodeError, IOError) as e:
+                if self.verbose:
+                    print(f"⚠️ Error reading {session_file}: {e}")
+
+        self.patterns["total_sessions"] = len(self.sessions_data)
+        return self.patterns
+
+    def _analyze_session(self, session: Dict[str, Any]):
+        """
+        Analyze individual session
+
+        Args:
+            session: Session data dictionary from Claude Code
+        """
+        # Handle session summary format (current JSONL format)
+        if session.get("type") == "summary":
+            # Count session types by summary content
+            summary = cast(str, session.get("summary", "")).lower()
+
+            # Simple analysis of session summaries
+            if any(keyword in summary for keyword in ["error", "fail", "issue", "problem"]):
+                self.patterns["failed_sessions"] = cast(int, self.patterns["failed_sessions"]) + 1
+                tool_failures = cast(defaultdict[str, int], self.patterns["tool_failures"])
+                tool_failures["session_error_in_summary"] += 1
+
+            # Extract potential tool usage from summary
+            tool_keywords = [
+                "test",
+                "build",
+                "deploy",
+                "analyze",
+                "create",
+                "update",
+                "fix",
+                "check",
+            ]
+            tool_usage = cast(defaultdict[str, int], self.patterns["tool_usage"])
+            for keyword in tool_keywords:
+                if keyword in summary:
+                    tool_usage[f"summary_{keyword}"] += 1
+
+            # Track session summaries as events
+            self.patterns["total_events"] = cast(int, self.patterns["total_events"]) + 1
+            return
+
+        # Handle detailed event format (legacy session-*.json format)
+        events = cast(list[Dict[str, Any]], session.get("events", []))
+        self.patterns["total_events"] = cast(int, self.patterns["total_events"]) + len(events)
+
+        has_error = False
+
+        for event in events:
+            event_type = event.get("type", "unknown")
+
+            # Extract tool usage patterns
+            if event_type == "tool_call":
+                tool_name = cast(str, event.get("toolName", "unknown")).split("(")[0]
+                tool_usage = cast(defaultdict[str, int], self.patterns["tool_usage"])
+                tool_usage[tool_name] += 1
+
+            # Tool error patterns
+            elif event_type == "tool_error":
+                error_msg = cast(str, event.get("error", "unknown error"))
+                tool_failures = cast(defaultdict[str, int], self.patterns["tool_failures"])
+                tool_failures[error_msg[:50]] += 1  # First 50 characters
+                has_error = True
+
+            # Permission requests
+            elif event_type == "permission_request":
+                perm_type = cast(str, event.get("permission_type", "unknown"))
+                perm_requests = cast(defaultdict[str, int], self.patterns["permission_requests"])
+                perm_requests[perm_type] += 1
+
+            # Hook failures
+            elif event_type == "hook_failure":
+                hook_name = cast(str, event.get("hook_name", "unknown"))
+                hook_failures = cast(defaultdict[str, int], self.patterns["hook_failures"])
+                hook_failures[hook_name] += 1
+                has_error = True
+
+            # Command usage
+            if "command" in event:
+                cmd_str = cast(str, event.get("command", "")).split()
+                if cmd_str:
+                    cmd = cmd_str[0]
+                    cmd_freq = cast(defaultdict[str, int], self.patterns["command_frequency"])
+                    cmd_freq[cmd] += 1
+
+        if has_error:
+            self.patterns["failed_sessions"] = cast(int, self.patterns["failed_sessions"]) + 1
+
+    def generate_report(self) -> str:
+        """
+        Generate markdown report
+
+        Returns:
+            Formatted markdown report string
+        """
+        timestamp = datetime.now().isoformat()
+        total_sessions = cast(int, self.patterns["total_sessions"])
+        failed_sessions = cast(int, self.patterns["failed_sessions"])
+        total_events = cast(int, self.patterns["total_events"])
+        success_rate = ((total_sessions - failed_sessions) / total_sessions * 100) if total_sessions > 0 else 0
+
+        report = f"""# MoAI-ADK Session Meta-Analysis Report
+
+**Generated at**: {timestamp}
+**Analysis period**: Last {self.days_back} days
+**Analysis scope**: `~/.claude/projects/`
+
+---
+
+## Overall Metrics
+
+| Metric | Value |
+|--------|-------|
+| **Total sessions** | {total_sessions} |
+| **Total events** | {total_events} |
+| **Successful sessions** | {total_sessions - failed_sessions} ({success_rate:.1f}%) |
+| **Failed sessions** | {failed_sessions} ({100 - success_rate:.1f}%) |
+| **Average session length** | {total_events / total_sessions if total_sessions > 0 else 0:.1f} |
+
+---
+
+## Tool Usage Patterns (Top 10)
+
+"""
+
+        # Top tool usage
+        tool_usage = cast(defaultdict[str, int], self.patterns["tool_usage"])
+        sorted_tools = sorted(tool_usage.items(), key=lambda x: x[1], reverse=True)
+
+        report += "| Tool | Usage Count |\n|------|----------|\n"
+        for tool, count in sorted_tools[:10]:
+            report += f"| `{tool}` | {count} |\n"
+
+        # Tool error patterns
+        report += "\n## Tool Error Patterns (Top 5)\n\n"
+
+        tool_failures = cast(defaultdict[str, int], self.patterns["tool_failures"])
+        if tool_failures:
+            sorted_errors = sorted(
+                tool_failures.items(),
+                key=lambda x: x[1],
+                reverse=True,
+            )
+            report += "| Error | Occurrence Count |\n|--------|----------|\n"
+            for error, count in sorted_errors[:5]:
+                report += f"| {error}... | {count} |\n"
+        else:
+            report += "✅ No tool errors\n"
+
+        # Hook failure analysis
+        report += "\n## Hook Failure Analysis\n\n"
+
+        hook_failures = cast(defaultdict[str, int], self.patterns["hook_failures"])
+        if hook_failures:
+            for hook, count in sorted(
+                hook_failures.items(),
+                key=lambda x: x[1],
+                reverse=True,
+            ):
+                report += f"- **{hook}**: {count} times\n"
+        else:
+            report += "✅ No hook failures\n"
+
+        # Permission request analysis
+        report += "\n## Permission Request Patterns\n\n"
+
+        perm_requests = cast(defaultdict[str, int], self.patterns["permission_requests"])
+        if perm_requests:
+            sorted_perms = sorted(
+                perm_requests.items(),
+                key=lambda x: x[1],
+                reverse=True,
+            )
+            report += "| Permission Type | Request Count |\n|---------|----------|\n"
+            for perm, count in sorted_perms:
+                report += f"| {perm} | {count} |\n"
+        else:
+            report += "✅ No permission requests\n"
+
+        # Improvement suggestions
+        report += "\n## Improvement Suggestions\n\n"
+        report += self._generate_suggestions()
+
+        return report
+
+    def _generate_suggestions(self) -> str:
+        """
+        Generate improvement suggestions based on patterns
+
+        Returns:
+            Formatted suggestions string
+        """
+        suggestions: list[str] = []
+
+        # High permission requests - review permission settings
+        perm_requests = cast(defaultdict[str, int], self.patterns["permission_requests"])
+        if perm_requests:
+            top_perm = max(
+                perm_requests.items(),
+                key=lambda x: x[1],
+            )
+            if top_perm[1] >= 5:
+                suggestions.append(
+                    f"Permission: **{top_perm[0]}** requested frequently ({top_perm[1]} times)\n"
+                    f"   - Review `permissions` in `.claude/settings.json`\n"
+                    f"   - Change `allow` to `ask` or add new Bash tool rules"
+                )
+
+        # Tool failure patterns - add fallback strategies
+        tool_failures = cast(defaultdict[str, int], self.patterns["tool_failures"])
+        if tool_failures:
+            top_error = max(
+                tool_failures.items(),
+                key=lambda x: x[1],
+            )
+            if top_error[1] >= 3:
+                suggestions.append(
+                    f"Tool error: **{top_error[0]}...** ({top_error[1]} times)\n"
+                    f"   - Add fallback strategy to CLAUDE.md\n"
+                    f"   - Example: 'If X error occurs, try Y'"
+                )
+
+        # Hook failures - review hook logic
+        hook_failures = cast(defaultdict[str, int], self.patterns["hook_failures"])
+        if hook_failures:
+            for hook, count in sorted(
+                hook_failures.items(),
+                key=lambda x: x[1],
+                reverse=True,
+            )[:3]:
+                if count >= 2:
+                    suggestions.append(
+                        f"Hook failure: **{hook}** ({count} times)\n"
+                        f"   - Debug `.claude/hooks/alfred/{hook}.py`\n"
+                        f"   - Check timeouts, permissions, and file paths"
+                    )
+
+        # Low success rate - general diagnosis
+        total_sessions = cast(int, self.patterns["total_sessions"])
+        failed_sessions = cast(int, self.patterns["failed_sessions"])
+        success_rate = ((total_sessions - failed_sessions) / total_sessions * 100) if total_sessions > 0 else 0
+
+        if success_rate < 80 and total_sessions >= 5:
+            suggestions.append(
+                f"Low success rate: **{success_rate:.1f}%**\n"
+                f"   - Review recent session logs in detail\n"
+                f"   - Re-evaluate rules and constraints in CLAUDE.md\n"
+                f"   - Verify context synchronization between Alfred and Sub-agents"
+            )
+
+        if not suggestions:
+            suggestions.append("✅ No major issues detected\n   - Current settings and rules are working well")
+
+        return "\n\n".join(suggestions)
+
+    def save_report(self, output_path: Optional[Path] = None, project_path: Optional[Path] = None) -> Path:
+        """
+        Save report to file
+
+        Args:
+            output_path: Custom output file path (optional)
+            project_path: Project root path (defaults to current working directory)
+
+        Returns:
+            Path to the saved report file
+        """
+        if output_path is None:
+            if project_path is None:
+                project_path = Path.cwd()
+
+            output_dir = project_path / ".moai" / "reports"
+            output_dir.mkdir(parents=True, exist_ok=True)
+            output_path = output_dir / f"daily-{datetime.now().strftime('%Y-%m-%d')}.md"
+
+        report = self.generate_report()
+        output_path.write_text(report, encoding="utf-8")
+
+        if self.verbose:
+            print(f"📄 Report saved: {output_path}")
+
+        return output_path
+
+    def get_metrics(self) -> Dict[str, Any]:
+        """
+        Get analysis metrics as dictionary
+
+        Returns:
+            Dictionary containing analysis metrics
+        """
+        total_sessions = cast(int, self.patterns["total_sessions"])
+        if total_sessions > 0:
+            failed_sessions = cast(int, self.patterns["failed_sessions"])
+            total_events = cast(int, self.patterns["total_events"])
+            self.patterns["success_rate"] = (total_sessions - failed_sessions) / total_sessions * 100
+            self.patterns["average_session_length"] = total_events / total_sessions
+
+        return self.patterns.copy()