htmlgraph 0.26.24__py3-none-any.whl → 0.27.0__py3-none-any.whl

htmlgraph/sdk.py

@@ -1,3430 +0,0 @@
-"""
-HtmlGraph SDK - AI-Friendly Interface
-
-Provides a fluent, ergonomic API for AI agents with:
-- Auto-discovery of .htmlgraph directory
-- Method chaining for all operations
-- Context managers for auto-save
-- Batch operations
-- Minimal boilerplate
-
-Example:
-    from htmlgraph import SDK
-
-    # Auto-discovers .htmlgraph directory
-    sdk = SDK(agent="claude")
-
-    # Fluent feature creation
-    feature = sdk.features.create(
-        title="User Authentication",
-        track="auth"
-    ).add_steps([
-        "Create login endpoint",
-        "Add JWT middleware",
-        "Write tests"
-    ]).set_priority("high").save()
-
-    # Work on a feature
-    with sdk.features.get("feature-001") as feature:
-        feature.start()
-        feature.complete_step(0)
-        # Auto-saves on exit
-
-    # Query
-    todos = sdk.features.where(status="todo", priority="high")
-
-    # Batch operations
-    sdk.features.mark_done(["feat-001", "feat-002", "feat-003"])
-"""
-
-from __future__ import annotations
-
-import os
-from pathlib import Path
-from typing import Any
-
-from htmlgraph.agent_detection import detect_agent_name
-from htmlgraph.agents import AgentInterface
-from htmlgraph.analytics import Analytics, CrossSessionAnalytics, DependencyAnalytics
-from htmlgraph.collections import (
-    BaseCollection,
-    BugCollection,
-    ChoreCollection,
-    EpicCollection,
-    FeatureCollection,
-    PhaseCollection,
-    SpikeCollection,
-    TaskDelegationCollection,
-    TodoCollection,
-)
-from htmlgraph.collections.insight import InsightCollection
-from htmlgraph.collections.metric import MetricCollection
-from htmlgraph.collections.pattern import PatternCollection
-from htmlgraph.collections.session import SessionCollection
-from htmlgraph.context_analytics import ContextAnalytics
-from htmlgraph.db.schema import HtmlGraphDB
-from htmlgraph.graph import HtmlGraph
-from htmlgraph.models import Node, Step
-from htmlgraph.session_manager import SessionManager
-from htmlgraph.session_warning import check_and_show_warning
-from htmlgraph.system_prompts import SystemPromptManager
-from htmlgraph.track_builder import TrackCollection
-from htmlgraph.types import (
-    ActiveWorkItem,
-    BottleneckDict,
-    SessionStartInfo,
-)
-
-
-class SDK:
-    """
-    Main SDK interface for AI agents.
-
-    Auto-discovers .htmlgraph directory and provides fluent API for all collections.
-
-    Available Collections:
-        - features: Feature work items with builder support
-        - bugs: Bug reports
-        - chores: Maintenance and chore tasks
-        - spikes: Investigation and research spikes
-        - epics: Large bodies of work
-        - phases: Project phases
-        - sessions: Agent sessions
-        - tracks: Work tracks
-        - agents: Agent information
-        - todos: Persistent task tracking (mirrors TodoWrite API)
-        - patterns: Workflow patterns (optimal/anti-pattern)
-        - insights: Session health insights
-        - metrics: Aggregated time-series metrics
-
-    System Prompt Management:
-        sdk.system_prompts - Manage system prompts
-            .get_active() - Get active prompt (project override OR plugin default)
-            .get_default() - Get plugin default system prompt
-            .get_project() - Get project-level override if exists
-            .create(template) - Create project-level override
-            .validate() - Validate prompt token count
-            .delete() - Delete project override (fall back to default)
-            .get_stats() - Get prompt statistics
-
-    Analytics & Decision Support:
-        sdk.dep_analytics - Dependency analysis
-            .find_bottlenecks(top_n=5) - Find blocking tasks
-            .get_parallel_work(max_agents=5) - Find parallelizable work
-            .recommend_next_tasks(agent_count=1) - Smart recommendations
-            .assess_dependency_risk() - Check for circular deps
-            .impact_analysis(node_id) - See what task unlocks
-
-        sdk.analytics - Work analytics
-            .get_work_type_distribution() - Breakdown by type
-            .get_spike_to_feature_ratio() - Investigation vs implementation
-            .get_maintenance_burden() - Chore vs feature ratio
-
-        sdk.context - Context tracking
-            .get_context_usage() - Session context metrics
-            .get_context_efficiency() - Efficiency score
-
-    Discovery & Help:
-        sdk.help() - Get structured help for all operations
-        sdk.help('analytics') - Get analytics-specific help
-        sdk.help('features') - Get feature collection help
-
-    Error Handling Patterns
-    =======================
-
-    SDK methods follow consistent error handling patterns by operation type:
-
-    LOOKUP OPERATIONS (Return None):
-        Single-item lookups return None when not found.
-        Always check the result before using.
-
-        >>> feature = sdk.features.get("nonexistent")
-        >>> if feature:
-        ...     print(feature.title)
-        ... else:
-        ...     print("Not found")
-
-    QUERY OPERATIONS (Return Empty List):
-        Queries return empty list when no matches or on error.
-        Safe to iterate without checking.
-
-        >>> results = sdk.features.where(status="impossible")
-        >>> for r in results:  # Empty iteration is safe
-        ...     print(r.title)
-
-    EDIT OPERATIONS (Raise Exception):
-        Edit operations raise NodeNotFoundError when target missing.
-        Use try/except to handle gracefully.
-
-        >>> from htmlgraph.exceptions import NodeNotFoundError
-        >>> try:
-        ...     with sdk.features.edit("nonexistent") as f:
-        ...         f.status = "done"
-        ... except NodeNotFoundError:
-        ...     print("Feature not found")
-
-    CREATE OPERATIONS (Raise on Validation):
-        Create operations raise ValidationError on invalid input.
-
-        >>> try:
-        ...     sdk.features.create("")  # Empty title
-        ... except ValidationError:
-        ...     print("Title required")
-
-    BATCH OPERATIONS (Return Results Dict):
-        Batch operations return dict with success_count, failed_ids, and warnings.
-        Provides detailed feedback for partial failures.
-
-        >>> result = sdk.features.mark_done(["feat-1", "missing", "feat-2"])
-        >>> print(f"Completed {result['success_count']} of 3")
-        >>> if result['failed_ids']:
-        ...     print(f"Failed: {result['failed_ids']}")
-        ...     print(f"Reasons: {result['warnings']}")
-
-    Pattern Summary:
-        | Operation Type | Error Behavior      | Example Method        |
-        |----------------|--------------------|-----------------------|
-        | Lookup         | Return None        | .get(id)              |
-        | Query          | Return []          | .where(), .all()      |
-        | Edit           | Raise Exception    | .edit(id)             |
-        | Create         | Raise on Invalid   | .create(title)        |
-        | Batch          | Return Results Dict| .mark_done([ids])     |
-        | Delete         | Return Bool        | .delete(id)           |
-
-    Available Exceptions:
-        - NodeNotFoundError: Node with ID not found
-        - ValidationError: Invalid input parameters
-        - ClaimConflictError: Node already claimed by another agent
-
-    Example:
-        sdk = SDK(agent="claude")
-
-        # Work with features (has builder support)
-        feature = sdk.features.create("User Auth")
-            .set_priority("high")
-            .add_steps(["Login", "Logout"])
-            .save()
-
-        # Work with bugs
-        high_bugs = sdk.bugs.where(status="todo", priority="high")
-        with sdk.bugs.edit("bug-001") as bug:
-            bug.status = "in-progress"
-
-        # Work with any collection
-        all_spikes = sdk.spikes.all()
-        sdk.chores.mark_done(["chore-001", "chore-002"])
-        sdk.epics.assign(["epic-001"], agent="claude")
-    """
-
-    def __init__(
-        self,
-        directory: Path | str | None = None,
-        agent: str | None = None,
-        parent_session: str | None = None,
-        db_path: str | None = None,
-    ):
-        """
-        Initialize SDK.
-
-        Args:
-            directory: Path to .htmlgraph directory (auto-discovered if not provided)
-            agent: REQUIRED - Agent identifier for operations.
-                Used to attribute work items (features, spikes, bugs, etc) to the agent.
-                Examples: agent='explorer', agent='coder', agent='tester'
-                Critical for: Work attribution, result retrieval, orchestrator tracking
-                Falls back to: CLAUDE_AGENT_NAME env var, then detect_agent_name()
-                Raises ValueError if not provided and cannot be detected
-            parent_session: Parent session ID to log activities to (for nested contexts)
-            db_path: Path to SQLite database file (optional, defaults to ~/.htmlgraph/htmlgraph.db)
-        """
-        if directory is None:
-            directory = self._discover_htmlgraph()
-
-        if agent is None:
-            # Try environment variable fallback
-            agent = os.getenv("CLAUDE_AGENT_NAME")
-
-        if agent is None:
-            # Try automatic detection
-            detected = detect_agent_name()
-            if detected and detected != "cli":
-                # Only accept detected if it's not the default fallback
-                agent = detected
-            else:
-                # No valid agent found - fail fast with helpful error message
-                raise ValueError(
-                    "Agent identifier is required for work attribution. "
-                    "Pass agent='name' to SDK() initialization. "
-                    "Examples: SDK(agent='explorer'), SDK(agent='coder'), SDK(agent='tester')\n"
-                    "Alternatively, set CLAUDE_AGENT_NAME environment variable.\n"
-                    "Critical for: Work attribution, result retrieval, orchestrator tracking"
-                )
-
-        self._directory = Path(directory)
-        self._agent_id = agent
-        self._parent_session = parent_session or os.getenv("HTMLGRAPH_PARENT_SESSION")
-
-        # Initialize SQLite database (Phase 2)
-        self._db = HtmlGraphDB(
-            db_path or str(Path.home() / ".htmlgraph" / "htmlgraph.db")
-        )
-        self._db.connect()
-        self._db.create_tables()
-
-        # Initialize underlying HtmlGraphs first (for backward compatibility and sharing)
-        # These are shared with SessionManager to avoid double-loading features
-        self._graph = HtmlGraph(self._directory / "features")
-        self._bugs_graph = HtmlGraph(self._directory / "bugs")
-
-        # Initialize SessionManager with shared graph instances to avoid double-loading
-        self.session_manager = SessionManager(
-            self._directory,
-            features_graph=self._graph,
-            bugs_graph=self._bugs_graph,
-        )
-
-        # Agent interface (for backward compatibility)
-        self._agent_interface = AgentInterface(
-            self._directory / "features", agent_id=agent
-        )
-
-        # Collection interfaces - all work item types (all with builder support)
-        self.features = FeatureCollection(self)
-        self.bugs = BugCollection(self)
-        self.chores = ChoreCollection(self)
-        self.spikes = SpikeCollection(self)
-        self.epics = EpicCollection(self)
-        self.phases = PhaseCollection(self)
-
-        # Non-work collections
-        self.sessions: SessionCollection = SessionCollection(self)
-        self.tracks: TrackCollection = TrackCollection(
-            self
-        )  # Use specialized collection with builder support
-        self.agents: BaseCollection = BaseCollection(self, "agents", "agent")
-
-        # Learning collections (Active Learning Persistence)
-        self.patterns = PatternCollection(self)
-        self.insights = InsightCollection(self)
-        self.metrics = MetricCollection(self)
-
-        # Todo collection (persistent task tracking)
-        self.todos = TodoCollection(self)
-
-        # Task delegation collection (observability for spawned agents)
-        self.task_delegations = TaskDelegationCollection(self)
-
-        # Create learning directories if needed
-        (self._directory / "patterns").mkdir(exist_ok=True)
-        (self._directory / "insights").mkdir(exist_ok=True)
-        (self._directory / "metrics").mkdir(exist_ok=True)
-        (self._directory / "todos").mkdir(exist_ok=True)
-        (self._directory / "task-delegations").mkdir(exist_ok=True)
-
-        # Analytics interface (Phase 2: Work Type Analytics)
-        self.analytics = Analytics(self)
-
-        # Dependency analytics interface (Advanced graph analytics)
-        self.dep_analytics = DependencyAnalytics(self._graph)
-
-        # Cross-session analytics interface (Git commit-based analytics)
-        self.cross_session_analytics = CrossSessionAnalytics(self)
-
-        # Context analytics interface (Context usage tracking)
-        self.context = ContextAnalytics(self)
-
-        # Pattern learning interface (Phase 2: Behavior Pattern Learning)
-        from htmlgraph.analytics.pattern_learning import PatternLearner
-
-        self.pattern_learning = PatternLearner(self._directory)
-
-        # Lazy-loaded orchestrator for subagent management
-        self._orchestrator: Any = None
-
-        # System prompt manager (lazy-loaded)
-        self._system_prompts: SystemPromptManager | None = None
-
-        # Session warning system (workaround for Claude Code hook bug #10373)
-        # Shows orchestrator instructions on first SDK usage per session
-        self._session_warning = check_and_show_warning(
-            self._directory,
-            agent=self._agent_id,
-            session_id=None,  # Will be set by session manager if available
-        )
-
-    @staticmethod
-    def _discover_htmlgraph() -> Path:
-        """
-        Auto-discover .htmlgraph directory.
-
-        Searches current directory and parents.
-        """
-        current = Path.cwd()
-
-        # Check current directory
-        if (current / ".htmlgraph").exists():
-            return current / ".htmlgraph"
-
-        # Check parent directories
-        for parent in current.parents:
-            if (parent / ".htmlgraph").exists():
-                return parent / ".htmlgraph"
-
-        # Default to current directory
-        return current / ".htmlgraph"
-
-    @property
-    def agent(self) -> str | None:
-        """Get current agent ID."""
-        return self._agent_id
-
-    @property
-    def system_prompts(self) -> SystemPromptManager:
-        """
-        Access system prompt management.
-
-        Provides methods to:
-        - Get active prompt (project override OR plugin default)
-        - Create/delete project-level overrides
-        - Validate token counts
-        - Get prompt statistics
-
-        Lazy-loaded on first access.
-
-        Returns:
-            SystemPromptManager instance
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-
-            # Get active prompt
-            >>> prompt = sdk.system_prompts.get_active()
-
-            # Create project override
-            >>> sdk.system_prompts.create("## Custom prompt\\n...")
-
-            # Validate token count
-            >>> result = sdk.system_prompts.validate()
-            >>> print(result['message'])
-
-            # Get statistics
-            >>> stats = sdk.system_prompts.get_stats()
-            >>> print(f"Source: {stats['source']}")
-        """
-        if self._system_prompts is None:
-            self._system_prompts = SystemPromptManager(self._directory)
-        return self._system_prompts
-
-    def dismiss_session_warning(self) -> bool:
-        """
-        Dismiss the session warning after reading it.
-
-        IMPORTANT: Call this as your FIRST action after seeing the orchestrator
-        warning. This confirms you've read the instructions.
-
-        Returns:
-            True if warning was dismissed, False if already dismissed
-
-        Example:
-            sdk = SDK(agent="claude")
-            # Warning shown automatically...
-
-            # First action: dismiss to confirm you read it
-            sdk.dismiss_session_warning()
-
-            # Now proceed with orchestration
-            sdk.spawn_coder(feature_id="feat-123", ...)
-        """
-        if self._session_warning:
-            return self._session_warning.dismiss(
-                agent=self._agent_id,
-                session_id=None,
-            )
-        return False
-
-    def get_warning_status(self) -> dict[str, Any]:
-        """
-        Get current session warning status.
-
-        Returns:
-            Dict with dismissed status, timestamp, and show count
-        """
-        if self._session_warning:
-            return self._session_warning.get_status()
-        return {"dismissed": True, "show_count": 0}
-
-    # =========================================================================
-    # SQLite Database Integration (Phase 2)
-    # =========================================================================
-
-    def db(self) -> HtmlGraphDB:
-        """
-        Get the SQLite database instance.
-
-        Returns:
-            HtmlGraphDB instance for executing queries
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> db = sdk.db()
-            >>> events = db.get_session_events("sess-123")
-            >>> features = db.get_features_by_status("todo")
-        """
-        return self._db
-
-    def query(self, sql: str, params: tuple = ()) -> list[dict[str, Any]]:
-        """
-        Execute a raw SQL query on the SQLite database.
-
-        Args:
-            sql: SQL query string
-            params: Query parameters (for safe parameterized queries)
-
-        Returns:
-            List of result dictionaries
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> results = sdk.query(
-            ...     "SELECT * FROM features WHERE status = ? AND priority = ?",
-            ...     ("todo", "high")
-            ... )
-            >>> for row in results:
-            ...     print(row["title"])
-        """
-        if not self._db.connection:
-            self._db.connect()
-
-        cursor = self._db.connection.cursor()  # type: ignore[union-attr]
-        cursor.execute(sql, params)
-        rows = cursor.fetchall()
-        return [dict(row) for row in rows]
-
-    def execute_query_builder(
-        self, sql: str, params: tuple = ()
-    ) -> list[dict[str, Any]]:
-        """
-        Execute a query using the Queries builder.
-
-        Args:
-            sql: SQL query from Queries builder
-            params: Parameters from Queries builder
-
-        Returns:
-            List of result dictionaries
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> sql, params = Queries.get_features_by_status("todo", limit=5)
-            >>> results = sdk.execute_query_builder(sql, params)
-        """
-        return self.query(sql, params)
-
-    def export_to_html(
-        self,
-        output_dir: str | None = None,
-        include_features: bool = True,
-        include_sessions: bool = True,
-        include_events: bool = False,
-    ) -> dict[str, int]:
-        """
-        Export SQLite data to HTML files for backward compatibility.
-
-        Args:
-            output_dir: Directory to export to (defaults to .htmlgraph)
-            include_features: Export features
-            include_sessions: Export sessions
-            include_events: Export events (detailed, use with care)
-
-        Returns:
-            Dict with export counts: {"features": int, "sessions": int, "events": int}
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> result = sdk.export_to_html()
-            >>> print(f"Exported {result['features']} features")
-        """
-        if output_dir is None:
-            output_dir = str(self._directory)
-
-        output_path = Path(output_dir)
-        counts = {"features": 0, "sessions": 0, "events": 0}
-
-        if include_features:
-            # Export all features from SQLite to HTML
-            features_dir = output_path / "features"
-            features_dir.mkdir(parents=True, exist_ok=True)
-
-            try:
-                cursor = self._db.connection.cursor()  # type: ignore[union-attr]
-                cursor.execute("SELECT * FROM features")
-                rows = cursor.fetchall()
-
-                for row in rows:
-                    feature_dict = dict(row)
-                    feature_id = feature_dict["id"]
-                    # Write HTML file (simplified export)
-                    html_file = features_dir / f"{feature_id}.html"
-                    html_file.write_text(
-                        f"<h1>{feature_dict['title']}</h1>"
-                        f"<p>Status: {feature_dict['status']}</p>"
-                        f"<p>Type: {feature_dict['type']}</p>"
-                    )
-                    counts["features"] += 1
-            except Exception as e:
-                import logging
-
-                logging.error(f"Error exporting features: {e}")
-
-        if include_sessions:
-            # Export all sessions from SQLite to HTML
-            sessions_dir = output_path / "sessions"
-            sessions_dir.mkdir(parents=True, exist_ok=True)
-
-            try:
-                cursor = self._db.connection.cursor()  # type: ignore[union-attr]
-                cursor.execute("SELECT * FROM sessions")
-                rows = cursor.fetchall()
-
-                for row in rows:
-                    session_dict = dict(row)
-                    session_id = session_dict["session_id"]
-                    # Write HTML file (simplified export)
-                    html_file = sessions_dir / f"{session_id}.html"
-                    html_file.write_text(
-                        f"<h1>Session {session_id}</h1>"
-                        f"<p>Agent: {session_dict['agent_assigned']}</p>"
-                        f"<p>Status: {session_dict['status']}</p>"
-                    )
-                    counts["sessions"] += 1
-            except Exception as e:
-                import logging
-
-                logging.error(f"Error exporting sessions: {e}")
-
-        return counts
-
-    def _log_event(
-        self,
-        event_type: str,
-        tool_name: str | None = None,
-        input_summary: str | None = None,
-        output_summary: str | None = None,
-        context: dict[str, Any] | None = None,
-        cost_tokens: int = 0,
-    ) -> bool:
-        """
-        Log an event to the SQLite database with parent-child linking.
-
-        Internal method used by collections to track operations.
-        Automatically creates a session if one doesn't exist.
-        Reads parent event ID from HTMLGRAPH_PARENT_ACTIVITY env var for hierarchical tracking.
-
-        Args:
-            event_type: Type of event (tool_call, completion, error, etc.)
-            tool_name: Tool that was called
-            input_summary: Summary of input
-            output_summary: Summary of output
-            context: Additional context metadata
-            cost_tokens: Token cost estimate
-
-        Returns:
-            True if logged successfully, False otherwise
-
-        Example (internal use):
-            >>> sdk._log_event(
-            ...     event_type="tool_call",
-            ...     tool_name="Edit",
-            ...     input_summary="Edit file.py",
-            ...     cost_tokens=100
-            ... )
-        """
-        from uuid import uuid4
-
-        event_id = f"evt-{uuid4().hex[:12]}"
-        session_id = self._parent_session or "cli-session"
-
-        # Read parent event ID from environment variable for hierarchical linking
-        parent_event_id = os.getenv("HTMLGRAPH_PARENT_ACTIVITY")
-
-        # Ensure session exists before logging event
-        try:
-            self._ensure_session_exists(session_id, parent_event_id=parent_event_id)
-        except Exception as e:
-            import logging
-
-            logging.debug(f"Failed to ensure session exists: {e}")
-            # Continue anyway - session creation failure shouldn't block event logging
-
-        return self._db.insert_event(
-            event_id=event_id,
-            agent_id=self._agent_id,
-            event_type=event_type,
-            session_id=session_id,
-            tool_name=tool_name,
-            input_summary=input_summary,
-            output_summary=output_summary,
-            context=context,
-            parent_event_id=parent_event_id,
-            cost_tokens=cost_tokens,
-        )
-
-    def _ensure_session_exists(
-        self, session_id: str, parent_event_id: str | None = None
-    ) -> None:
-        """
-        Create a session record if it doesn't exist.
-
-        Args:
-            session_id: Session ID to ensure exists
-            parent_event_id: Event that spawned this session (optional)
-        """
-        if not self._db.connection:
-            self._db.connect()
-
-        cursor = self._db.connection.cursor()  # type: ignore[union-attr]
-        cursor.execute(
-            "SELECT COUNT(*) FROM sessions WHERE session_id = ?", (session_id,)
-        )
-        exists = cursor.fetchone()[0] > 0
-
-        if not exists:
-            # Create session record
-            self._db.insert_session(
-                session_id=session_id,
-                agent_assigned=self._agent_id,
-                is_subagent=self._parent_session is not None,
-                parent_session_id=self._parent_session,
-                parent_event_id=parent_event_id,
-            )
-
-    def reload(self) -> None:
-        """Reload all data from disk."""
-        self._graph.reload()
-        self._agent_interface.reload()
-        # SessionManager reloads implicitly on access via its converters/graphs
-
-    def summary(self, max_items: int = 10) -> str:
-        """
-        Get project summary.
-
-        Returns:
-            Compact overview for AI agent orientation
-        """
-        return self._agent_interface.get_summary(max_items)
-
-    def my_work(self) -> dict[str, Any]:
-        """
-        Get current agent's workload.
-
-        Returns:
-            Dict with in_progress, completed counts
-        """
-        if not self._agent_id:
-            raise ValueError("No agent ID set")
-        return self._agent_interface.get_workload(self._agent_id)
-
-    def next_task(
-        self, priority: str | None = None, auto_claim: bool = True
-    ) -> Node | None:
-        """
-        Get next available task for this agent.
-
-        Args:
-            priority: Optional priority filter
-            auto_claim: Automatically claim the task
-
-        Returns:
-            Next available Node or None
-        """
-        return self._agent_interface.get_next_task(
-            agent_id=self._agent_id,
-            priority=priority,
-            node_type="feature",
-            auto_claim=auto_claim,
-        )
-
-    def set_session_handoff(
-        self,
-        handoff_notes: str | None = None,
-        recommended_next: str | None = None,
-        blockers: list[str] | None = None,
-        session_id: str | None = None,
-    ) -> Any:
-        """
-        Set handoff context on a session.
-
-        Args:
-            handoff_notes: Notes for next session/agent
-            recommended_next: Suggested next steps
-            blockers: List of blockers
-            session_id: Specific session ID (defaults to active session)
-
-        Returns:
-            Updated Session or None if not found
-        """
-        if not session_id:
-            if self._agent_id:
-                active = self.session_manager.get_active_session_for_agent(
-                    self._agent_id
-                )
-            else:
-                active = self.session_manager.get_active_session()
-            if not active:
-                return None
-            session_id = active.id
-
-        return self.session_manager.set_session_handoff(
-            session_id=session_id,
-            handoff_notes=handoff_notes,
-            recommended_next=recommended_next,
-            blockers=blockers,
-        )
-
-    def continue_from_last(
-        self,
-        agent: str | None = None,
-        auto_create_session: bool = True,
-    ) -> tuple[Any, Any]:
-        """
-        Continue work from the last completed session.
-
-        Loads context from previous session including handoff notes,
-        recommended files, blockers, and recent commits.
-
-        Args:
-            agent: Filter by agent (None = current SDK agent)
-            auto_create_session: Create new session if True
-
-        Returns:
-            Tuple of (new_session, resume_info) or (None, None)
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> session, resume = sdk.continue_from_last()
-            >>> if resume:
-            ...     print(resume.summary)
-            ...     print(resume.next_focus)
-            ...     for file in resume.recommended_files:
-            ...         print(f"  - {file}")
-        """
-        if not agent:
-            agent = self._agent_id
-
-        return self.session_manager.continue_from_last(
-            agent=agent,
-            auto_create_session=auto_create_session,
-        )
-
-    def end_session_with_handoff(
-        self,
-        session_id: str | None = None,
-        summary: str | None = None,
-        next_focus: str | None = None,
-        blockers: list[str] | None = None,
-        keep_context: list[str] | None = None,
-        auto_recommend_context: bool = True,
-    ) -> Any:
-        """
-        End session with handoff information for next session.
-
-        Args:
-            session_id: Session to end (None = active session)
-            summary: What was accomplished
-            next_focus: What should be done next
-            blockers: List of blockers
-            keep_context: List of files to keep context for
-            auto_recommend_context: Auto-recommend files from git
-
-        Returns:
-            Updated Session or None
-
-        Example:
-            >>> sdk.end_session_with_handoff(
-            ...     summary="Completed OAuth integration",
-            ...     next_focus="Implement JWT token refresh",
-            ...     blockers=["Waiting for security review"],
-            ...     keep_context=["src/auth/oauth.py"]
-            ... )
-        """
-        if not session_id:
-            if self._agent_id:
-                active = self.session_manager.get_active_session_for_agent(
-                    self._agent_id
-                )
-            else:
-                active = self.session_manager.get_active_session()
-            if not active:
-                return None
-            session_id = active.id
-
-        return self.session_manager.end_session_with_handoff(
-            session_id=session_id,
-            summary=summary,
-            next_focus=next_focus,
-            blockers=blockers,
-            keep_context=keep_context,
-            auto_recommend_context=auto_recommend_context,
-        )
-
-    def start_session(
-        self,
-        session_id: str | None = None,
-        title: str | None = None,
-        agent: str | None = None,
-    ) -> Any:
-        """
-        Start a new session.
-
-        Args:
-            session_id: Optional session ID
-            title: Optional session title
-            agent: Optional agent override (defaults to SDK agent)
-
-        Returns:
-            New Session instance
-        """
-        return self.session_manager.start_session(
-            session_id=session_id,
-            agent=agent or self._agent_id or "cli",
-            title=title,
-            parent_session_id=self._parent_session,
-        )
-
-    def end_session(
-        self,
-        session_id: str,
-        handoff_notes: str | None = None,
-        recommended_next: str | None = None,
-        blockers: list[str] | None = None,
-    ) -> Any:
-        """
-        End a session.
-
-        Args:
-            session_id: Session ID to end
-            handoff_notes: Optional handoff notes
-            recommended_next: Optional recommendations
-            blockers: Optional blockers
-
-        Returns:
-            Ended Session instance
-        """
-        return self.session_manager.end_session(
-            session_id=session_id,
-            handoff_notes=handoff_notes,
-            recommended_next=recommended_next,
-            blockers=blockers,
-        )
-
-    def get_status(self) -> dict[str, Any]:
-        """
-        Get project status.
-
-        Returns:
-            Dict with status metrics (WIP, counts, etc.)
-        """
-        return self.session_manager.get_status()
-
-    def dedupe_sessions(
-        self,
-        max_events: int = 1,
-        move_dir_name: str = "_orphans",
-        dry_run: bool = False,
-        stale_extra_active: bool = True,
-    ) -> dict[str, int]:
-        """
-        Move low-signal sessions (e.g. SessionStart-only) out of the main sessions dir.
-
-        Args:
-            max_events: Maximum events threshold (sessions with <= this many events are moved)
-            move_dir_name: Directory name to move orphaned sessions to
-            dry_run: If True, only report what would be done without actually moving files
-            stale_extra_active: If True, also mark extra active sessions as stale
-
-        Returns:
-            Dict with counts: {"scanned": int, "moved": int, "missing": int, "staled_active": int, "kept_active": int}
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> result = sdk.dedupe_sessions(max_events=1, dry_run=False)
-            >>> print(f"Scanned: {result['scanned']}, Moved: {result['moved']}")
-        """
-        return self.session_manager.dedupe_orphan_sessions(
-            max_events=max_events,
-            move_dir_name=move_dir_name,
-            dry_run=dry_run,
-            stale_extra_active=stale_extra_active,
-        )
-
-    def track_activity(
-        self,
-        tool: str,
-        summary: str,
-        file_paths: list[str] | None = None,
-        success: bool = True,
-        feature_id: str | None = None,
-        session_id: str | None = None,
-        parent_activity_id: str | None = None,
-        payload: dict[str, Any] | None = None,
-    ) -> Any:
-        """
-        Track an activity in the current or specified session.
-
-        Args:
-            tool: Tool name (Edit, Bash, Read, etc.)
-            summary: Human-readable summary of the activity
-            file_paths: Files involved in this activity
-            success: Whether the tool call succeeded
-            feature_id: Explicit feature ID (skips attribution if provided)
-            session_id: Session ID (defaults to parent session if available, then active session)
-            parent_activity_id: ID of parent activity (e.g., Skill/Task invocation)
-            payload: Optional rich payload data
-
-        Returns:
-            Created ActivityEntry with attribution
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> entry = sdk.track_activity(
-            ...     tool="CustomTool",
-            ...     summary="Performed custom analysis",
-            ...     file_paths=["src/main.py"],
-            ...     success=True
-            ... )
-            >>> print(f"Tracked: [{entry.tool}] {entry.summary}")
-        """
-        # Determine target session: explicit parameter > parent_session > active > none
-        if not session_id:
-            # Priority 1: Parent session (explicitly provided or from env var)
-            if self._parent_session:
-                session_id = self._parent_session
-            else:
-                # Priority 2: Active session for this agent
-                active = self.session_manager.get_active_session(agent=self._agent_id)
-                if active:
-                    session_id = active.id
-                else:
-                    raise ValueError(
-                        "No active session. Start one with sdk.start_session()"
-                    )
-
-        # Get parent activity ID from environment if not provided
-        if not parent_activity_id:
-            parent_activity_id = os.getenv("HTMLGRAPH_PARENT_ACTIVITY")
-
-        return self.session_manager.track_activity(
-            session_id=session_id,
-            tool=tool,
-            summary=summary,
-            file_paths=file_paths,
-            success=success,
-            feature_id=feature_id,
-            parent_activity_id=parent_activity_id,
-            payload=payload,
-        )
-
-    # =========================================================================
-    # Strategic Planning & Analytics (Agent-Friendly Interface)
-    # =========================================================================
-
-    def find_bottlenecks(self, top_n: int = 5) -> list[BottleneckDict]:
-        """
-        Identify tasks blocking the most downstream work.
-
-        Note: Prefer using sdk.dep_analytics.find_bottlenecks() directly.
-        This method exists for backward compatibility.
-
-        Args:
-            top_n: Maximum number of bottlenecks to return
-
-        Returns:
-            List of bottleneck tasks with impact metrics
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> # Preferred approach
-            >>> bottlenecks = sdk.dep_analytics.find_bottlenecks(top_n=3)
-            >>> # Or via SDK (backward compatibility)
-            >>> bottlenecks = sdk.find_bottlenecks(top_n=3)
-            >>> for bn in bottlenecks:
-            ...     print(f"{bn['title']} blocks {bn['blocks_count']} tasks")
-        """
-        bottlenecks = self.dep_analytics.find_bottlenecks(top_n=top_n)
-
-        # Convert to agent-friendly dict format for backward compatibility
-        return [
-            {
-                "id": bn.id,
-                "title": bn.title,
-                "status": bn.status,
-                "priority": bn.priority,
-                "blocks_count": bn.transitive_blocking,
-                "impact_score": bn.weighted_impact,
-                "blocked_tasks": bn.blocked_nodes[:5],
-            }
-            for bn in bottlenecks
-        ]
-
-    def get_parallel_work(self, max_agents: int = 5) -> dict[str, Any]:
-        """
-        Find tasks that can be worked on simultaneously.
-
-        Note: Prefer using sdk.dep_analytics.find_parallelizable_work() directly.
-        This method exists for backward compatibility.
-
-        Args:
-            max_agents: Maximum number of parallel agents to plan for
-
-        Returns:
-            Dict with parallelization opportunities
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> # Preferred approach
-            >>> report = sdk.dep_analytics.find_parallelizable_work(status="todo")
-            >>> # Or via SDK (backward compatibility)
-            >>> parallel = sdk.get_parallel_work(max_agents=3)
-            >>> print(f"Can work on {parallel['max_parallelism']} tasks at once")
-            >>> print(f"Ready now: {parallel['ready_now']}")
-        """
-        report = self.dep_analytics.find_parallelizable_work(status="todo")
-
-        ready_now = (
-            report.dependency_levels[0].nodes if report.dependency_levels else []
-        )
-
-        return {
-            "max_parallelism": report.max_parallelism,
-            "ready_now": ready_now[:max_agents],
-            "total_ready": len(ready_now),
-            "level_count": len(report.dependency_levels),
-            "next_level": report.dependency_levels[1].nodes
-            if len(report.dependency_levels) > 1
-            else [],
-        }
-
-    def recommend_next_work(self, agent_count: int = 1) -> list[dict[str, Any]]:
-        """
-        Get smart recommendations for what to work on next.
-
-        Note: Prefer using sdk.dep_analytics.recommend_next_tasks() directly.
-        This method exists for backward compatibility.
-
-        Considers priority, dependencies, and transitive impact.
-
-        Args:
-            agent_count: Number of agents/tasks to recommend
-
-        Returns:
-            List of recommended tasks with reasoning
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> # Preferred approach
-            >>> recs = sdk.dep_analytics.recommend_next_tasks(agent_count=3)
-            >>> # Or via SDK (backward compatibility)
-            >>> recs = sdk.recommend_next_work(agent_count=3)
-            >>> for rec in recs:
-            ...     print(f"{rec['title']} (score: {rec['score']})")
-            ...     print(f"  Reasons: {rec['reasons']}")
-        """
-        recommendations = self.dep_analytics.recommend_next_tasks(
-            agent_count=agent_count, lookahead=5
-        )
-
-        return [
-            {
-                "id": rec.id,
-                "title": rec.title,
-                "priority": rec.priority,
-                "score": rec.score,
-                "reasons": rec.reasons,
-                "estimated_hours": rec.estimated_effort,
-                "unlocks_count": len(rec.unlocks),
-                "unlocks": rec.unlocks[:3],
-            }
-            for rec in recommendations.recommendations
-        ]
-
-    def assess_risks(self) -> dict[str, Any]:
-        """
-        Assess dependency-related risks in the project.
-
-        Note: Prefer using sdk.dep_analytics.assess_dependency_risk() directly.
-        This method exists for backward compatibility.
-
-        Identifies single points of failure, circular dependencies,
-        and orphaned tasks.
-
-        Returns:
-            Dict with risk assessment results
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> # Preferred approach
-            >>> risk = sdk.dep_analytics.assess_dependency_risk()
-            >>> # Or via SDK (backward compatibility)
-            >>> risks = sdk.assess_risks()
-            >>> if risks['high_risk_count'] > 0:
-            ...     print(f"Warning: {risks['high_risk_count']} high-risk tasks")
-        """
-        risk = self.dep_analytics.assess_dependency_risk()
-
-        return {
-            "high_risk_count": len(risk.high_risk),
-            "high_risk_tasks": [
-                {
-                    "id": node.id,
-                    "title": node.title,
-                    "risk_score": node.risk_score,
-                    "risk_factors": [f.description for f in node.risk_factors],
-                }
-                for node in risk.high_risk
-            ],
-            "circular_dependencies": risk.circular_dependencies,
-            "orphaned_count": len(risk.orphaned_nodes),
-            "orphaned_tasks": risk.orphaned_nodes[:5],
-            "recommendations": risk.recommendations,
-        }
-
-    def analyze_impact(self, node_id: str) -> dict[str, Any]:
-        """
-        Analyze the impact of completing a specific task.
-
-        Note: Prefer using sdk.dep_analytics.impact_analysis() directly.
-        This method exists for backward compatibility.
-
-        Args:
-            node_id: Task to analyze
-
-        Returns:
-            Dict with impact analysis
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> # Preferred approach
-            >>> impact = sdk.dep_analytics.impact_analysis("feature-001")
-            >>> # Or via SDK (backward compatibility)
-            >>> impact = sdk.analyze_impact("feature-001")
-            >>> print(f"Completing this unlocks {impact['unlocks_count']} tasks")
-        """
-        impact = self.dep_analytics.impact_analysis(node_id)
-
-        return {
-            "node_id": node_id,
-            "direct_dependents": impact.direct_dependents,
-            "total_impact": impact.transitive_dependents,
-            "completion_impact": impact.completion_impact,
-            "unlocks_count": len(impact.affected_nodes),
-            "affected_tasks": impact.affected_nodes[:10],
-        }
-
-    def get_work_queue(
-        self, agent_id: str | None = None, limit: int = 10, min_score: float = 0.0
-    ) -> list[dict[str, Any]]:
-        """
-        Get prioritized work queue showing recommended work, active work, and dependencies.
-
-        This method provides a comprehensive view of:
-        1. Recommended next work (using smart analytics)
-        2. Active work by all agents
-        3. Blocked items and what's blocking them
-        4. Priority-based scoring
-
-        Args:
-            agent_id: Agent to get queue for (defaults to SDK agent)
-            limit: Maximum number of items to return (default: 10)
-            min_score: Minimum score threshold (default: 0.0)
-
-        Returns:
-            List of work queue items with scoring and metadata:
-                - task_id: Work item ID
-                - title: Work item title
-                - status: Current status
-                - priority: Priority level
-                - score: Routing score
-                - complexity: Complexity level (if set)
-                - effort: Estimated effort (if set)
-                - blocks_count: Number of tasks this blocks (if any)
-                - blocked_by: List of blocking task IDs (if blocked)
-                - agent_assigned: Current assignee (if any)
-                - type: Work item type (feature, bug, spike, etc.)
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> queue = sdk.get_work_queue(limit=5)
-            >>> for item in queue:
-            ...     print(f"{item['score']:.1f} - {item['title']}")
-            ...     if item.get('blocked_by'):
-            ...         print(f"  ⚠️  Blocked by: {', '.join(item['blocked_by'])}")
-        """
-        from htmlgraph.routing import AgentCapabilityRegistry, CapabilityMatcher
-
-        agent = agent_id or self._agent_id or "cli"
-
-        # Get all work item types
-        all_work = []
-        for collection_name in ["features", "bugs", "spikes", "chores", "epics"]:
-            collection = getattr(self, collection_name, None)
-            if collection:
-                # Get todo and blocked items
-                for item in collection.where(status="todo"):
-                    all_work.append(item)
-                for item in collection.where(status="blocked"):
-                    all_work.append(item)
-
-        if not all_work:
-            return []
-
-        # Get recommendations from analytics (uses strategic scoring)
-        recommendations = self.recommend_next_work(agent_count=limit * 2)
-        rec_scores = {rec["id"]: rec["score"] for rec in recommendations}
-
-        # Build routing registry
-        registry = AgentCapabilityRegistry()
-
-        # Register current agent
-        registry.register_agent(agent, capabilities=[], wip_limit=5)
-
-        # Get current WIP count for agent
-        wip_count = len(self.features.where(status="in-progress", agent_assigned=agent))
-        registry.set_wip(agent, wip_count)
-
-        # Score each work item
-        queue_items = []
-        for item in all_work:
-            # Use strategic score if available, otherwise use routing score
-            if item.id in rec_scores:
-                score = rec_scores[item.id]
-            else:
-                # Fallback to routing score
-                agent_profile = registry.get_agent(agent)
-                if agent_profile:
-                    score = CapabilityMatcher.score_agent_task_fit(agent_profile, item)
-                else:
-                    score = 0.0
-
-            # Apply minimum score filter
-            if score < min_score:
-                continue
-
-            # Build queue item
-            queue_item = {
-                "task_id": item.id,
-                "title": item.title,
-                "status": item.status,
-                "priority": item.priority,
-                "score": score,
-                "type": item.type,
-                "complexity": getattr(item, "complexity", None),
-                "effort": getattr(item, "estimated_effort", None),
-                "agent_assigned": getattr(item, "agent_assigned", None),
-                "blocks_count": 0,
-                "blocked_by": [],
-            }
-
-            # Add dependency information
-            if hasattr(item, "edges"):
-                # Check if this item blocks others
-                blocks = item.edges.get("blocks", [])
-                queue_item["blocks_count"] = len(blocks)
-
-                # Check if this item is blocked
-                blocked_by = item.edges.get("blocked_by", [])
-                queue_item["blocked_by"] = blocked_by
-
-            queue_items.append(queue_item)
-
-        # Sort by score (descending)
-        queue_items.sort(key=lambda x: x["score"], reverse=True)
-
-        # Limit results
-        return queue_items[:limit]
-
-    def work_next(
-        self,
-        agent_id: str | None = None,
-        auto_claim: bool = False,
-        min_score: float = 0.0,
-    ) -> Node | None:
-        """
-        Get the next best task for an agent using smart routing.
-
-        Uses both strategic analytics and capability-based routing to find
-        the optimal next task.
-
-        Args:
-            agent_id: Agent to get task for (defaults to SDK agent)
-            auto_claim: Automatically claim the task (default: False)
-            min_score: Minimum score threshold (default: 0.0)
-
-        Returns:
-            Next best Node or None if no suitable task found
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> task = sdk.work_next(auto_claim=True)
-            >>> if task:
-            ...     print(f"Working on: {task.title}")
-            ...     # Task is automatically claimed and assigned
-        """
-        agent = agent_id or self._agent_id or "cli"
-
-        # Get work queue - get more items since we filter for actionable (todo) only
-        queue = self.get_work_queue(agent_id=agent, limit=20, min_score=min_score)
-
-        if not queue:
-            return None
-
-        # Find the first actionable (todo) task - blocked tasks are not actionable
-        top_item = None
-        for item in queue:
-            if item["status"] == "todo":
-                top_item = item
-                break
-
-        if top_item is None:
-            return None
-
-        # Fetch the actual node
-        task = None
-        for collection_name in ["features", "bugs", "spikes", "chores", "epics"]:
-            collection = getattr(self, collection_name, None)
-            if collection:
-                try:
-                    task = collection.get(top_item["task_id"])
-                    if task:
-                        break
-                except (ValueError, FileNotFoundError):
-                    continue
-
-        if not task:
-            return None
-
-        # Auto-claim if requested
-        if auto_claim and task.status == "todo" and collection is not None:
-            # Claim the task
-            # collection.edit returns context manager or None
-            task_editor: Any = collection.edit(task.id)
-            if task_editor is not None:
-                # collection.edit returns context manager
-                with task_editor as t:
-                    t.status = "in-progress"
-                    t.agent_assigned = agent
-
-        result: Node | None = task
-        return result
-
-    # =========================================================================
-    # Planning Workflow Integration
-    # =========================================================================
-
-    def start_planning_spike(
-        self,
-        title: str,
-        context: str = "",
-        timebox_hours: float = 4.0,
-        auto_start: bool = True,
-    ) -> Node:
-        """
-        Create a planning spike to research and design before implementation.
-
-        This is for timeboxed investigation before creating a full track.
-
-        Args:
-            title: Spike title (e.g., "Plan User Authentication System")
-            context: Background information
-            timebox_hours: Time limit for spike (default: 4 hours)
-            auto_start: Automatically start the spike (default: True)
-
-        Returns:
-            Created spike Node
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> spike = sdk.start_planning_spike(
-            ...     "Plan Real-time Notifications",
-            ...     context="Users need live updates. Research options.",
-            ...     timebox_hours=3.0
-            ... )
-        """
-        from htmlgraph.ids import generate_id
-        from htmlgraph.models import Spike, SpikeType
-
-        # Create spike directly (SpikeBuilder doesn't exist yet)
-        spike_id = generate_id(node_type="spike", title=title)
-        spike = Spike(
-            id=spike_id,
-            title=title,
-            type="spike",
-            status="in-progress" if auto_start and self._agent_id else "todo",
-            spike_type=SpikeType.ARCHITECTURAL,
-            timebox_hours=int(timebox_hours),
-            agent_assigned=self._agent_id if auto_start and self._agent_id else None,
-            steps=[
-                Step(description="Research existing solutions and patterns"),
-                Step(description="Define requirements and constraints"),
-                Step(description="Design high-level architecture"),
-                Step(description="Identify dependencies and risks"),
-                Step(description="Create implementation plan"),
-            ],
-            content=f"<p>{context}</p>" if context else "",
-            edges={},
-            properties={},
-        )
-
-        self._graph.add(spike)
-        return spike
-
-    def create_track_from_plan(
-        self,
-        title: str,
-        description: str,
-        spike_id: str | None = None,
-        priority: str = "high",
-        requirements: list[str | tuple[str, str]] | None = None,
-        phases: list[tuple[str, list[str]]] | None = None,
-    ) -> dict[str, Any]:
-        """
-        Create a track with spec and plan from planning results.
-
-        Args:
-            title: Track title
-            description: Track description
-            spike_id: Optional spike ID that led to this track
-            priority: Track priority (default: "high")
-            requirements: List of requirements (strings or (req, priority) tuples)
-            phases: List of (phase_name, tasks) tuples for the plan
-
-        Returns:
-            Dict with track, spec, and plan details
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> track_info = sdk.create_track_from_plan(
-            ...     title="User Authentication System",
-            ...     description="OAuth 2.0 with JWT tokens",
-            ...     requirements=[
-            ...         ("OAuth 2.0 integration", "must-have"),
-            ...         ("JWT token management", "must-have"),
-            ...         "Password reset flow"
-            ...     ],
-            ...     phases=[
-            ...         ("Phase 1: OAuth", ["Setup providers (2h)", "Callback (2h)"]),
-            ...         ("Phase 2: JWT", ["Token signing (2h)", "Refresh (1.5h)"])
-            ...     ]
-            ... )
-        """
-
-        builder = (
-            self.tracks.builder()
-            .title(title)
-            .description(description)
-            .priority(priority)
-        )
-
-        # Add reference to planning spike if provided
-        if spike_id:
-            # Access internal data for track builder
-            data: dict[str, Any] = builder._data  # type: ignore[attr-defined]
-            data["properties"]["planning_spike"] = spike_id
-
-        # Add spec if requirements provided
-        if requirements:
-            # Convert simple strings to (requirement, "must-have") tuples
-            req_list = []
-            for req in requirements:
-                if isinstance(req, str):
-                    req_list.append((req, "must-have"))
-                else:
-                    req_list.append(req)
-
-            builder.with_spec(
-                overview=description,
-                context=f"Track created from planning spike: {spike_id}"
-                if spike_id
-                else "",
-                requirements=req_list,
-                acceptance_criteria=[],
-            )
-
-        # Add plan if phases provided
-        if phases:
-            builder.with_plan_phases(phases)
-
-        track = builder.create()
-
-        return {
-            "track_id": track.id,
-            "title": track.title,
-            "has_spec": bool(requirements),
-            "has_plan": bool(phases),
-            "spike_id": spike_id,
-            "priority": priority,
-        }
-
-    def smart_plan(
-        self,
-        description: str,
-        create_spike: bool = True,
-        timebox_hours: float = 4.0,
-        research_completed: bool = False,
-        research_findings: dict[str, Any] | None = None,
-    ) -> dict[str, Any]:
-        """
-        Smart planning workflow: analyzes project context and creates spike or track.
-
-        This is the main entry point for planning new work. It:
-        1. Checks current project state
-        2. Provides context from strategic analytics
-        3. Creates a planning spike or track as appropriate
-
-        **IMPORTANT: Research Phase Required**
-        For complex features, you should complete research BEFORE planning:
-        1. Use /htmlgraph:research or WebSearch to gather best practices
-        2. Document findings (libraries, patterns, anti-patterns)
-        3. Pass research_completed=True and research_findings to this method
-        4. This ensures planning is informed by industry best practices
-
-        Research-first workflow:
-            1. /htmlgraph:research "{topic}" → Gather external knowledge
-            2. sdk.smart_plan(..., research_completed=True) → Plan with context
-            3. Complete spike steps → Design solution
-            4. Create track from plan → Structure implementation
-
-        Args:
-            description: What you want to plan (e.g., "User authentication system")
-            create_spike: Create a spike for research (default: True)
-            timebox_hours: If creating spike, time limit (default: 4 hours)
-            research_completed: Whether research was performed (default: False)
-            research_findings: Structured research findings (optional)
-
-        Returns:
-            Dict with planning context and created spike/track info
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> # WITH research (recommended for complex work)
-            >>> research = {
-            ...     "topic": "OAuth 2.0 best practices",
-            ...     "sources_count": 5,
-            ...     "recommended_library": "authlib",
-            ...     "key_insights": ["Use PKCE", "Implement token rotation"]
-            ... }
-            >>> plan = sdk.smart_plan(
-            ...     "User authentication system",
-            ...     create_spike=True,
-            ...     research_completed=True,
-            ...     research_findings=research
-            ... )
-            >>> print(f"Created: {plan['spike_id']}")
-            >>> print(f"Research informed: {plan['research_informed']}")
-        """
-        # Get project context from strategic analytics
-        bottlenecks = self.find_bottlenecks(top_n=3)
-        risks = self.assess_risks()
-        parallel = self.get_parallel_work(max_agents=5)
-
-        context = {
-            "bottlenecks_count": len(bottlenecks),
-            "high_risk_count": risks["high_risk_count"],
-            "parallel_capacity": parallel["max_parallelism"],
-            "description": description,
-        }
-
-        # Build context string with research info
-        context_str = f"Project context:\n- {len(bottlenecks)} bottlenecks\n- {risks['high_risk_count']} high-risk items\n- {parallel['max_parallelism']} parallel capacity"
-
-        if research_completed and research_findings:
-            context_str += f"\n\nResearch completed:\n- Topic: {research_findings.get('topic', description)}"
-            if "sources_count" in research_findings:
-                context_str += f"\n- Sources: {research_findings['sources_count']}"
-            if "recommended_library" in research_findings:
-                context_str += (
-                    f"\n- Recommended: {research_findings['recommended_library']}"
-                )
-
-        # Validation: warn if complex work planned without research
-        is_complex = any(
-            [
-                "auth" in description.lower(),
-                "security" in description.lower(),
-                "real-time" in description.lower(),
-                "websocket" in description.lower(),
-                "oauth" in description.lower(),
-                "performance" in description.lower(),
-                "integration" in description.lower(),
-            ]
-        )
-
-        warnings = []
-        if is_complex and not research_completed:
-            warnings.append(
-                "⚠️  Complex feature detected without research. "
-                "Consider using /htmlgraph:research first to gather best practices."
-            )
-
-        if create_spike:
-            spike = self.start_planning_spike(
-                title=f"Plan: {description}",
-                context=context_str,
-                timebox_hours=timebox_hours,
-            )
-
-            # Store research metadata in spike properties if provided
-            if research_completed and research_findings:
-                spike.properties["research_completed"] = True
-                spike.properties["research_findings"] = research_findings
-                self._graph.update(spike)
-
-            result = {
-                "type": "spike",
-                "spike_id": spike.id,
-                "title": spike.title,
-                "status": spike.status,
-                "timebox_hours": timebox_hours,
-                "project_context": context,
-                "research_informed": research_completed,
-                "next_steps": [
-                    "Research and design the solution"
-                    if not research_completed
-                    else "Design solution using research findings",
-                    "Complete spike steps",
-                    "Use SDK.create_track_from_plan() to create track",
-                ],
-            }
-
-            if warnings:
-                result["warnings"] = warnings
-
-            return result
-        else:
-            # Direct track creation (for when you already know what to do)
-            track_info = self.create_track_from_plan(
-                title=description, description=f"Planned with context: {context}"
-            )
-
-            result = {
-                "type": "track",
-                **track_info,
-                "project_context": context,
-                "research_informed": research_completed,
-                "next_steps": [
-                    "Create features from track plan",
-                    "Link features to track",
-                    "Start implementation",
-                ],
-            }
-
-            if warnings:
-                result["warnings"] = warnings
-
-            return result
-
-    def plan_parallel_work(
-        self,
-        max_agents: int = 5,
-        shared_files: list[str] | None = None,
-    ) -> dict[str, Any]:
-        """
-        Plan and prepare parallel work execution.
-
-        This integrates with smart_plan to enable parallel agent dispatch.
-        Uses the 6-phase ParallelWorkflow:
-        1. Pre-flight analysis (dependencies, risks)
-        2. Context preparation (shared file caching)
-        3. Prompt generation (for Task tool)
-
-        Args:
-            max_agents: Maximum parallel agents (default: 5)
-            shared_files: Files to pre-cache for all agents
-
-        Returns:
-            Dict with parallel execution plan:
-                - can_parallelize: Whether parallelization is recommended
-                - analysis: Pre-flight analysis results
-                - prompts: Ready-to-use Task tool prompts
-                - recommendations: Optimization suggestions
-
-        Example:
-            >>> sdk = SDK(agent="orchestrator")
-            >>> plan = sdk.plan_parallel_work(max_agents=3)
-            >>> if plan["can_parallelize"]:
-            ...     # Use prompts with Task tool
-            ...     for p in plan["prompts"]:
-            ...         Task(prompt=p["prompt"], description=p["description"])
-        """
-        from htmlgraph.parallel import ParallelWorkflow
-
-        workflow = ParallelWorkflow(self)
-
-        # Phase 1: Pre-flight analysis
-        analysis = workflow.analyze(max_agents=max_agents)
-
-        result = {
-            "can_parallelize": analysis.can_parallelize,
-            "max_parallelism": analysis.max_parallelism,
-            "ready_tasks": analysis.ready_tasks,
-            "blocked_tasks": analysis.blocked_tasks,
-            "speedup_factor": analysis.speedup_factor,
-            "recommendation": analysis.recommendation,
-            "warnings": analysis.warnings,
-            "prompts": [],
-        }
-
-        if not analysis.can_parallelize:
-            result["reason"] = analysis.recommendation
-            return result
-
-        # Phase 2 & 3: Prepare tasks and generate prompts
-        tasks = workflow.prepare_tasks(
-            analysis.ready_tasks[:max_agents],
-            shared_files=shared_files,
-        )
-        prompts = workflow.generate_prompts(tasks)
-
-        result["prompts"] = prompts
-        result["task_count"] = len(prompts)
-
-        # Add efficiency guidelines
-        result["guidelines"] = {
-            "dispatch": "Send ALL Task calls in a SINGLE message for true parallelism",
-            "patterns": [
-                "Grep → Read (search before reading)",
-                "Read → Edit → Bash (read, modify, test)",
-                "Glob → Read (find files first)",
-            ],
-            "avoid": [
-                "Sequential Task calls (loses parallelism)",
-                "Read → Read → Read (cache instead)",
-                "Edit → Edit → Edit (batch edits)",
-            ],
-        }
-
-        return result
-
-    def aggregate_parallel_results(
-        self,
-        agent_ids: list[str],
-    ) -> dict[str, Any]:
-        """
-        Aggregate results from parallel agent execution.
-
-        Call this after parallel agents complete to:
-        - Collect health metrics
-        - Detect anti-patterns
-        - Identify conflicts
-        - Generate recommendations
-
-        Args:
-            agent_ids: List of agent/transcript IDs to analyze
-
-        Returns:
-            Dict with aggregated results and validation
-
-        Example:
-            >>> # After parallel work completes
-            >>> results = sdk.aggregate_parallel_results([
-            ...     "agent-abc123",
-            ...     "agent-def456",
-            ...     "agent-ghi789",
-            ... ])
-            >>> print(f"Health: {results['avg_health_score']:.0%}")
-            >>> print(f"Conflicts: {results['conflicts']}")
-        """
-        from htmlgraph.parallel import ParallelWorkflow
-
-        workflow = ParallelWorkflow(self)
-
-        # Phase 5: Aggregate
-        aggregate = workflow.aggregate(agent_ids)
-
-        # Phase 6: Validate
-        validation = workflow.validate(aggregate)
-
-        return {
-            "total_agents": aggregate.total_agents,
-            "successful": aggregate.successful,
-            "failed": aggregate.failed,
-            "total_duration_seconds": aggregate.total_duration_seconds,
-            "parallel_speedup": aggregate.parallel_speedup,
-            "avg_health_score": aggregate.avg_health_score,
-            "total_anti_patterns": aggregate.total_anti_patterns,
-            "files_modified": aggregate.files_modified,
-            "conflicts": aggregate.conflicts,
-            "recommendations": aggregate.recommendations,
-            "validation": validation,
-            "all_passed": all(validation.values()),
-        }
-
-    # =========================================================================
-    # Subagent Orchestration
-    # =========================================================================
-
-    @property
-    def orchestrator(self) -> Any:
-        """
-        Get the subagent orchestrator for spawning explorer/coder agents.
-
-        Lazy-loaded on first access.
-
-        Returns:
-            SubagentOrchestrator instance
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> explorer = sdk.orchestrator.spawn_explorer(
-            ...     task="Find all API endpoints",
-            ...     scope="src/"
-            ... )
-        """
-        if self._orchestrator is None:
-            from htmlgraph.orchestrator import SubagentOrchestrator
-
-            self._orchestrator = SubagentOrchestrator(self)  # type: ignore[assignment]
-        return self._orchestrator
-
-    def spawn_explorer(
-        self,
-        task: str,
-        scope: str | None = None,
-        patterns: list[str] | None = None,
-        questions: list[str] | None = None,
-    ) -> dict[str, Any]:
-        """
-        Spawn an explorer subagent for codebase discovery.
-
-        Explorer agents are optimized for finding files, searching patterns,
-        and mapping code without modifying anything.
-
-        Args:
-            task: What to explore/discover
-            scope: Directory scope (e.g., "src/")
-            patterns: Glob patterns to focus on
-            questions: Specific questions to answer
-
-        Returns:
-            Dict with prompt ready for Task tool
-
-        Note:
-            Returns dict with 'prompt', 'description', 'subagent_type' keys.
-            Returns empty dict if spawning fails.
-
-        Example:
-            >>> prompt = sdk.spawn_explorer(
-            ...     task="Find all database models",
-            ...     scope="src/models/",
-            ...     questions=["What ORM is used?"]
-            ... )
-            >>> # Execute with Task tool
-            >>> Task(prompt=prompt["prompt"], description=prompt["description"])
-
-        See also:
-            spawn_coder: Spawn implementation agent with feature context
-            orchestrate: Full exploration + implementation workflow
-        """
-        subagent_prompt = self.orchestrator.spawn_explorer(
-            task=task,
-            scope=scope,
-            patterns=patterns,
-            questions=questions,
-        )
-        result: dict[str, Any] = subagent_prompt.to_task_kwargs()
-        return result
-
-    def spawn_coder(
-        self,
-        feature_id: str,
-        context: str | None = None,
-        files_to_modify: list[str] | None = None,
-        test_command: str | None = None,
-    ) -> dict[str, Any]:
-        """
-        Spawn a coder subagent for implementing changes.
-
-        Coder agents are optimized for reading, modifying, and testing code.
-
-        Args:
-            feature_id: Feature being implemented
-            context: Results from explorer (string summary)
-            files_to_modify: Specific files to change
-            test_command: Command to verify changes
-
-        Returns:
-            Dict with prompt ready for Task tool
-
-        Note:
-            Returns dict with 'prompt', 'description', 'subagent_type' keys.
-            Requires valid feature_id. Returns empty dict if feature not found.
-
-        Example:
-            >>> prompt = sdk.spawn_coder(
-            ...     feature_id="feat-add-auth",
-            ...     context=explorer_results,
-            ...     test_command="uv run pytest tests/auth/"
-            ... )
-            >>> Task(prompt=prompt["prompt"], description=prompt["description"])
-
-        See also:
-            spawn_explorer: Explore codebase before implementation
-            orchestrate: Full exploration + implementation workflow
-        """
-        subagent_prompt = self.orchestrator.spawn_coder(
-            feature_id=feature_id,
-            context=context,
-            files_to_modify=files_to_modify,
-            test_command=test_command,
-        )
-        result: dict[str, Any] = subagent_prompt.to_task_kwargs()
-        return result
-
-    def orchestrate(
-        self,
-        feature_id: str,
-        exploration_scope: str | None = None,
-        test_command: str | None = None,
-    ) -> dict[str, Any]:
-        """
-        Orchestrate full feature implementation with explorer and coder.
-
-        Generates prompts for a two-phase workflow:
-        1. Explorer discovers relevant code and patterns
-        2. Coder implements the feature based on explorer findings
-
-        Args:
-            feature_id: Feature to implement
-            exploration_scope: Directory to explore
-            test_command: Test command for verification
-
-        Returns:
-            Dict with explorer and coder prompts
-
-        Example:
-            >>> prompts = sdk.orchestrate(
-            ...     "feat-add-caching",
-            ...     exploration_scope="src/cache/",
-            ...     test_command="uv run pytest tests/cache/"
-            ... )
-            >>> # Phase 1: Run explorer
-            >>> Task(prompt=prompts["explorer"]["prompt"], ...)
-            >>> # Phase 2: Run coder with explorer results
-            >>> Task(prompt=prompts["coder"]["prompt"], ...)
-
-        See also:
-            spawn_explorer: Just the exploration phase
-            spawn_coder: Just the implementation phase
-        """
-        prompts = self.orchestrator.orchestrate_feature(
-            feature_id=feature_id,
-            exploration_scope=exploration_scope,
-            test_command=test_command,
-        )
-        return {
-            "explorer": prompts["explorer"].to_task_kwargs(),
-            "coder": prompts["coder"].to_task_kwargs(),
-            "workflow": [
-                "1. Execute explorer Task and collect results",
-                "2. Parse explorer results for files and patterns",
-                "3. Execute coder Task with explorer context",
-                "4. Verify coder results and update feature status",
-            ],
-        }
-
-    # =========================================================================
-    # Session Management Optimization
-    # =========================================================================
-
-    def get_session_start_info(
-        self,
-        include_git_log: bool = True,
-        git_log_count: int = 5,
-        analytics_top_n: int = 3,
-        analytics_max_agents: int = 3,
-    ) -> SessionStartInfo:
-        """
-        Get comprehensive session start information in a single call.
-
-        Consolidates all information needed for session start into one method,
-        reducing context usage from 6+ tool calls to 1.
-
-        Args:
-            include_git_log: Include recent git commits (default: True)
-            git_log_count: Number of recent commits to include (default: 5)
-            analytics_top_n: Number of bottlenecks/recommendations (default: 3)
-            analytics_max_agents: Max agents for parallel work analysis (default: 3)
-
-        Returns:
-            Dict with comprehensive session start context:
-                - status: Project status (nodes, collections, WIP)
-                - active_work: Current active work item (if any)
-                - features: List of features with status
-                - sessions: Recent sessions
-                - git_log: Recent commits (if include_git_log=True)
-                - analytics: Strategic insights (bottlenecks, recommendations, parallel)
-
-        Note:
-            Returns empty dict {} if session context unavailable.
-            Always check for expected keys before accessing.
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> info = sdk.get_session_start_info()
-            >>> print(f"Project: {info['status']['total_nodes']} nodes")
-            >>> print(f"WIP: {info['status']['in_progress_count']}")
-            >>> if info.get('active_work'):
-            ...     print(f"Active: {info['active_work']['title']}")
-            >>> for bn in info['analytics']['bottlenecks']:
-            ...     print(f"Bottleneck: {bn['title']}")
-        """
-        import subprocess
-
-        result = {}
-
-        # 1. Project status
-        result["status"] = self.get_status()
-
-        # 2. Active work item (validation status) - always include, even if None
-        result["active_work"] = self.get_active_work_item()  # type: ignore[assignment]
-
-        # 3. Features list (simplified)
-        features_list: list[dict[str, object]] = []
-        for feature in self.features.all():
-            features_list.append(
-                {
-                    "id": feature.id,
-                    "title": feature.title,
-                    "status": feature.status,
-                    "priority": feature.priority,
-                    "steps_total": len(feature.steps),
-                    "steps_completed": sum(1 for s in feature.steps if s.completed),
-                }
-            )
-        result["features"] = features_list  # type: ignore[assignment]
-
-        # 4. Sessions list (recent 20)
-        sessions_list: list[dict[str, Any]] = []
-        for session in self.sessions.all()[:20]:
-            sessions_list.append(
-                {
-                    "id": session.id,
-                    "status": session.status,
-                    "agent": session.properties.get("agent", "unknown"),
-                    "event_count": session.properties.get("event_count", 0),
-                    "started": session.created.isoformat()
-                    if hasattr(session, "created")
-                    else None,
-                }
-            )
-        result["sessions"] = sessions_list  # type: ignore[assignment]
-
-        # 5. Git log (if requested)
-        if include_git_log:
-            try:
-                git_result = subprocess.run(
-                    ["git", "log", "--oneline", f"-{git_log_count}"],
-                    capture_output=True,
-                    text=True,
-                    check=True,
-                    cwd=self._directory.parent,
-                )
-                git_lines: list[str] = git_result.stdout.strip().split("\n")
-                result["git_log"] = git_lines  # type: ignore[assignment]
-            except (subprocess.CalledProcessError, FileNotFoundError):
-                empty_list: list[str] = []
-                result["git_log"] = empty_list  # type: ignore[assignment]
-
-        # 6. Strategic analytics
-        result["analytics"] = {
-            "bottlenecks": self.find_bottlenecks(top_n=analytics_top_n),
-            "recommendations": self.recommend_next_work(agent_count=analytics_top_n),
-            "parallel": self.get_parallel_work(max_agents=analytics_max_agents),
-        }
-
-        return result  # type: ignore[return-value]
-
-    def get_active_work_item(
-        self,
-        agent: str | None = None,
-        filter_by_agent: bool = False,
-        work_types: list[str] | None = None,
-    ) -> ActiveWorkItem | None:
-        """
-        Get the currently active work item (in-progress status).
-
-        This is used by the PreToolUse validation hook to check if code changes
-        have an active work item for attribution.
-
-        Args:
-            agent: Agent ID for filtering (optional)
-            filter_by_agent: If True, filter by agent. If False (default), return any active work item
-            work_types: Work item types to check (defaults to all: features, bugs, spikes, chores, epics)
-
-        Returns:
-            Dict with work item details or None if no active work item found:
-                - id: Work item ID
-                - title: Work item title
-                - type: Work item type (feature, bug, spike, chore, epic)
-                - status: Should be "in-progress"
-                - agent: Assigned agent
-                - steps_total: Total steps
-                - steps_completed: Completed steps
-                - auto_generated: (spikes only) True if auto-generated spike
-                - spike_subtype: (spikes only) "session-init" or "transition"
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> # Get any active work item
-            >>> active = sdk.get_active_work_item()
-            >>> if active:
-            ...     print(f"Working on: {active['title']}")
-            ...
-            >>> # Get only this agent's active work item
-            >>> active = sdk.get_active_work_item(filter_by_agent=True)
-        """
-        # Default to all work item types
-        if work_types is None:
-            work_types = ["features", "bugs", "spikes", "chores", "epics"]
-
-        # Search across all work item types
-        # Separate real work items from auto-generated spikes
-        real_work_items = []
-        auto_spikes = []
-
-        for work_type in work_types:
-            collection = getattr(self, work_type, None)
-            if collection is None:
-                continue
-
-            # Query for in-progress items
-            in_progress = collection.where(status="in-progress")
-
-            for item in in_progress:
-                # Filter by agent if requested
-                if filter_by_agent:
-                    agent_id = agent or self._agent_id
-                    if agent_id and hasattr(item, "agent_assigned"):
-                        if item.agent_assigned != agent_id:
-                            continue
-
-                item_dict = {
-                    "id": item.id,
-                    "title": item.title,
-                    "type": item.type,
-                    "status": item.status,
-                    "agent": getattr(item, "agent_assigned", None),
-                    "steps_total": len(item.steps) if hasattr(item, "steps") else 0,
-                    "steps_completed": sum(1 for s in item.steps if s.completed)
-                    if hasattr(item, "steps")
-                    else 0,
-                }
-
-                # Add spike-specific fields for auto-spike detection
-                if item.type == "spike":
-                    item_dict["auto_generated"] = getattr(item, "auto_generated", False)
-                    item_dict["spike_subtype"] = getattr(item, "spike_subtype", None)
-
-                    # Separate auto-spikes from real work
-                    # Auto-spikes are temporary tracking items (session-init, transition, conversation-init)
-                    is_auto_spike = item_dict["auto_generated"] and item_dict[
-                        "spike_subtype"
-                    ] in ("session-init", "transition", "conversation-init")
-
-                    if is_auto_spike:
-                        auto_spikes.append(item_dict)
-                    else:
-                        # Real user-created spike
-                        real_work_items.append(item_dict)
-                else:
-                    # Features, bugs, chores, epics are always real work
-                    real_work_items.append(item_dict)
-
-        # Prioritize real work items over auto-spikes
-        # Auto-spikes should only show if there's NO other active work item
-        if real_work_items:
-            return real_work_items[0]  # type: ignore[return-value]
-
-        if auto_spikes:
-            return auto_spikes[0]  # type: ignore[return-value]
-
-        return None
-
-    # =========================================================================
-    # Operations Layer - Server, Hooks, Events, Analytics
-    # =========================================================================
-
-    def start_server(
-        self,
-        port: int = 8080,
-        host: str = "localhost",
-        watch: bool = True,
-        auto_port: bool = False,
-    ) -> Any:
-        """
-        Start HtmlGraph server for browsing graph via web UI.
-
-        Args:
-            port: Port to listen on (default: 8080)
-            host: Host to bind to (default: "localhost")
-            watch: Enable file watching for auto-reload (default: True)
-            auto_port: Automatically find available port if specified port is in use (default: False)
-
-        Returns:
-            ServerStartResult with handle, warnings, and config used
-
-        Raises:
-            PortInUseError: If port is in use and auto_port=False
-            ServerStartError: If server fails to start
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> result = sdk.start_server(port=8080, watch=True)
-            >>> print(f"Server running at {result.handle.url}")
-            >>> # Open browser to http://localhost:8080
-            >>>
-            >>> # Stop server when done
-            >>> sdk.stop_server(result.handle)
-
-        See also:
-            stop_server: Stop running server
-            get_server_status: Check if server is running
-        """
-        from htmlgraph.operations import server
-
-        return server.start_server(
-            port=port,
-            graph_dir=self._directory,
-            static_dir=self._directory.parent,  # Project root for index.html
-            host=host,
-            watch=watch,
-            auto_port=auto_port,
-        )
-
-    def stop_server(self, handle: Any) -> None:
-        """
-        Stop a running HtmlGraph server.
-
-        Args:
-            handle: ServerHandle returned from start_server()
-
-        Raises:
-            ServerStartError: If shutdown fails
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> result = sdk.start_server()
-            >>> # Work with server...
-            >>> sdk.stop_server(result.handle)
-        """
-        from htmlgraph.operations import server
-
-        server.stop_server(handle)
-
-    def get_server_status(self, handle: Any | None = None) -> Any:
-        """
-        Check server status.
-
-        Args:
-            handle: Optional ServerHandle to check
-
-        Returns:
-            ServerStatus indicating whether server is running
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> result = sdk.start_server()
-            >>> status = sdk.get_server_status(result.handle)
-            >>> print(f"Running: {status.running}")
-        """
-        from htmlgraph.operations import server
-
-        return server.get_server_status(handle)
-
-    def install_hooks(self, use_copy: bool = False) -> Any:
-        """
-        Install Git hooks for automatic tracking.
-
-        Installs hooks that automatically track sessions, activities, and features
-        as you work.
-
-        Args:
-            use_copy: Force copy instead of symlink (default: False)
-
-        Returns:
-            HookInstallResult with installation details
-
-        Raises:
-            HookInstallError: If installation fails
-            HookConfigError: If configuration is invalid
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> result = sdk.install_hooks()
-            >>> print(f"Installed: {result.installed}")
-            >>> print(f"Skipped: {result.skipped}")
-            >>> if result.warnings:
-            ...     print(f"Warnings: {result.warnings}")
-
-        See also:
-            list_hooks: List installed hooks
-            validate_hook_config: Validate hook configuration
-        """
-        from htmlgraph.operations import hooks
-
-        return hooks.install_hooks(
-            project_dir=self._directory.parent,
-            use_copy=use_copy,
-        )
-
-    def list_hooks(self) -> Any:
-        """
-        List Git hooks status (enabled/disabled/missing).
-
-        Returns:
-            HookListResult with enabled, disabled, and missing hooks
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> result = sdk.list_hooks()
-            >>> print(f"Enabled: {result.enabled}")
-            >>> print(f"Disabled: {result.disabled}")
-            >>> print(f"Missing: {result.missing}")
-        """
-        from htmlgraph.operations import hooks
-
-        return hooks.list_hooks(project_dir=self._directory.parent)
-
-    def validate_hook_config(self) -> Any:
-        """
-        Validate hook configuration.
-
-        Returns:
-            HookValidationResult with validation status
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> result = sdk.validate_hook_config()
-            >>> if not result.valid:
-            ...     print(f"Errors: {result.errors}")
-            >>> if result.warnings:
-            ...     print(f"Warnings: {result.warnings}")
-        """
-        from htmlgraph.operations import hooks
-
-        return hooks.validate_hook_config(project_dir=self._directory.parent)
-
-    def export_sessions(self, overwrite: bool = False) -> Any:
-        """
-        Export legacy session HTML logs to JSONL events.
-
-        Converts HTML session files to JSONL format for efficient querying.
-
-        Args:
-            overwrite: Whether to overwrite existing JSONL files (default: False)
-
-        Returns:
-            EventExportResult with counts of written, skipped, failed files
-
-        Raises:
-            EventOperationError: If export fails
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> result = sdk.export_sessions()
-            >>> print(f"Exported {result.written} sessions")
-            >>> print(f"Skipped {result.skipped} (already exist)")
-            >>> if result.failed > 0:
-            ...     print(f"Failed {result.failed} sessions")
-
-        See also:
-            rebuild_event_index: Rebuild SQLite index from JSONL
-            query_events: Query exported events
-        """
-        from htmlgraph.operations import events
-
-        return events.export_sessions(
-            graph_dir=self._directory,
-            overwrite=overwrite,
-        )
-
-    def rebuild_event_index(self) -> Any:
-        """
-        Rebuild SQLite analytics index from JSONL events.
-
-        Creates an optimized SQLite index for fast analytics queries.
-
-        Returns:
-            EventRebuildResult with db_path and counts of inserted/skipped events
-
-        Raises:
-            EventOperationError: If rebuild fails
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> result = sdk.rebuild_event_index()
-            >>> print(f"Rebuilt index: {result.db_path}")
-            >>> print(f"Inserted {result.inserted} events")
-            >>> print(f"Skipped {result.skipped} (duplicates)")
-
-        See also:
-            export_sessions: Export HTML sessions to JSONL first
-        """
-        from htmlgraph.operations import events
-
-        return events.rebuild_index(graph_dir=self._directory)
-
-    def query_events(
-        self,
-        session_id: str | None = None,
-        tool: str | None = None,
-        feature_id: str | None = None,
-        since: str | None = None,
-        limit: int | None = None,
-    ) -> Any:
-        """
-        Query events from JSONL logs with optional filters.
-
-        Args:
-            session_id: Filter by session ID (None = all sessions)
-            tool: Filter by tool name (e.g., 'Bash', 'Edit')
-            feature_id: Filter by attributed feature ID
-            since: Only events after this timestamp (ISO string)
-            limit: Maximum number of events to return
-
-        Returns:
-            EventQueryResult with matching events and total count
-
-        Raises:
-            EventOperationError: If query fails
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> # Get all events for a session
-            >>> result = sdk.query_events(session_id="sess-123")
-            >>> print(f"Found {result.total} events")
-            >>>
-            >>> # Get recent Bash events
-            >>> result = sdk.query_events(
-            ...     tool="Bash",
-            ...     since="2025-01-01T00:00:00Z",
-            ...     limit=10
-            ... )
-            >>> for event in result.events:
-            ...     print(f"{event['timestamp']}: {event['summary']}")
-
-        See also:
-            export_sessions: Export sessions to JSONL first
-            get_event_stats: Get event statistics
-        """
-        from htmlgraph.operations import events
-
-        return events.query_events(
-            graph_dir=self._directory,
-            session_id=session_id,
-            tool=tool,
-            feature_id=feature_id,
-            since=since,
-            limit=limit,
-        )
-
-    def get_event_stats(self) -> Any:
-        """
-        Get statistics about events in the system.
-
-        Returns:
-            EventStats with counts of total events, sessions, and files
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> stats = sdk.get_event_stats()
-            >>> print(f"Total events: {stats.total_events}")
-            >>> print(f"Sessions: {stats.session_count}")
-            >>> print(f"JSONL files: {stats.file_count}")
-        """
-        from htmlgraph.operations import events
-
-        return events.get_event_stats(graph_dir=self._directory)
-
-    def analyze_session(self, session_id: str) -> Any:
-        """
-        Compute detailed analytics for a single session.
-
-        Analyzes work distribution, spike-to-feature ratio, maintenance burden,
-        transition metrics, and more.
-
-        Args:
-            session_id: ID of the session to analyze
-
-        Returns:
-            AnalyticsSessionResult with session metrics and warnings
-
-        Raises:
-            AnalyticsOperationError: If session cannot be analyzed
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> result = sdk.analyze_session("sess-123")
-            >>> print(f"Primary work type: {result.metrics['primary_work_type']}")
-            >>> print(f"Total events: {result.metrics['total_events']}")
-            >>> print(f"Work distribution: {result.metrics['work_distribution']}")
-            >>> if result.warnings:
-            ...     print(f"Warnings: {result.warnings}")
-
-        See also:
-            analyze_project: Analyze entire project
-        """
-        from htmlgraph.operations import analytics
-
-        return analytics.analyze_session(
-            graph_dir=self._directory,
-            session_id=session_id,
-        )
-
-    def analyze_project(self) -> Any:
-        """
-        Compute project-wide analytics.
-
-        Analyzes all sessions, work distribution, spike-to-feature ratios,
-        maintenance burden, and session types across the entire project.
-
-        Returns:
-            AnalyticsProjectResult with project metrics and warnings
-
-        Raises:
-            AnalyticsOperationError: If project cannot be analyzed
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> result = sdk.analyze_project()
-            >>> print(f"Total sessions: {result.metrics['total_sessions']}")
-            >>> print(f"Work distribution: {result.metrics['work_distribution']}")
-            >>> print(f"Spike-to-feature ratio: {result.metrics['spike_to_feature_ratio']}")
-            >>> print(f"Session types: {result.metrics['session_types']}")
-            >>> for session in result.metrics['recent_sessions']:
-            ...     print(f"  {session['session_id']}: {session['primary_work_type']}")
-
-        See also:
-            analyze_session: Analyze a single session
-            get_work_recommendations: Get work recommendations
-        """
-        from htmlgraph.operations import analytics
-
-        return analytics.analyze_project(graph_dir=self._directory)
-
-    def get_work_recommendations(self) -> Any:
-        """
-        Get smart work recommendations based on project state.
-
-        Uses dependency analytics to recommend next tasks based on priority,
-        dependencies, and impact.
-
-        Returns:
-            RecommendationsResult with recommendations, reasoning, and warnings
-
-        Raises:
-            AnalyticsOperationError: If recommendations cannot be generated
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> result = sdk.get_work_recommendations()
-            >>> for rec in result.recommendations:
-            ...     print(f"{rec['title']} (score: {rec['score']})")
-            ...     print(f"  Reasons: {rec['reasons']}")
-            ...     print(f"  Unlocks: {rec['unlocks']}")
-            >>> print(f"Reasoning: {result.reasoning}")
-
-        See also:
-            recommend_next_work: Legacy method (backward compatibility)
-            get_work_queue: Get prioritized work queue
-        """
-        from htmlgraph.operations import analytics
-
-        return analytics.get_recommendations(graph_dir=self._directory)
-
-    # =========================================================================
-    # Task Attribution - Subagent Work Tracking
-    # =========================================================================
-
-    def get_task_attribution(self, task_id: str) -> dict[str, Any]:
-        """
-        Get attribution - which subagent did what work in this task?
-
-        Queries the database to find all events associated with a Claude Code task,
-        showing which subagent executed each tool call.
-
-        Args:
-            task_id: Claude Code's internal task ID (available from Task() response)
-
-        Returns:
-            Dictionary with task_id, by_subagent mapping, and total_events count
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> result = sdk.get_task_attribution("task-abc123-xyz789")
-            >>> for subagent, events in result['by_subagent'].items():
-            ...     print(f"{subagent}:")
-            ...     for event in events:
-            ...         print(f"  - {event['tool']}: {event['summary']}")
-            >>> print(f"Total events: {result['total_events']}")
-
-        See also:
-            get_subagent_work: Get all work grouped by subagent in a session
-        """
-        from htmlgraph.config import get_database_path
-        from htmlgraph.db.schema import HtmlGraphDB
-
-        try:
-            db = HtmlGraphDB(str(get_database_path()))
-            events = db.get_events_for_task(task_id)
-
-            # Group by subagent_type
-            by_subagent: dict[str, list[dict[str, Any]]] = {}
-            for event in events:
-                agent = event.get("subagent_type", "orchestrator")
-                if agent not in by_subagent:
-                    by_subagent[agent] = []
-                by_subagent[agent].append(
-                    {
-                        "tool": event.get("tool_name"),
-                        "summary": event.get("input_summary"),
-                        "timestamp": event.get("created_at"),
-                        "event_id": event.get("event_id"),
-                        "success": not event.get("is_error", False),
-                    }
-                )
-
-            return {
-                "task_id": task_id,
-                "by_subagent": by_subagent,
-                "total_events": len(events),
-            }
-        except Exception as e:
-            return {
-                "task_id": task_id,
-                "by_subagent": {},
-                "total_events": 0,
-                "error": str(e),
-            }
-
-    def get_subagent_work(self, session_id: str) -> dict[str, list[dict[str, Any]]]:
-        """
-        Get all work grouped by which subagent did it in a session.
-
-        Shows which subagent (researcher, general-purpose, etc.) executed each
-        tool call within a session.
-
-        Args:
-            session_id: Session ID to analyze
-
-        Returns:
-            Dictionary mapping subagent_type to list of events they executed.
-            Each event includes: tool_name, input_summary, output_summary, created_at, event_id
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> work = sdk.get_subagent_work("sess-123")
-            >>> for subagent, events in work.items():
-            ...     print(f"{subagent} ({len(events)} events):")
-            ...     for event in events:
-            ...         print(f"  - {event['tool_name']}: {event['input_summary']}")
-
-        See also:
-            get_task_attribution: Get work for a specific Claude Code task
-            analyze_session: Get session metrics and analytics
-        """
-        from htmlgraph.config import get_database_path
-        from htmlgraph.db.schema import HtmlGraphDB
-
-        try:
-            db = HtmlGraphDB(str(get_database_path()))
-            return db.get_subagent_work(session_id)
-        except Exception:
-            return {}
-
-    # =========================================================================
-    # Help & Documentation
-    # =========================================================================
-
-    def help(self, topic: str | None = None) -> str:
-        """
-        Get help on SDK usage.
-
-        Args:
-            topic: Optional topic (e.g., 'features', 'sessions', 'analytics', 'orchestration')
-
-        Returns:
-            Formatted help text
-
-        Example:
-            >>> sdk = SDK(agent="claude")
-            >>> print(sdk.help())  # List all topics
-            >>> print(sdk.help('features'))  # Feature collection help
-            >>> print(sdk.help('analytics'))  # Analytics help
-
-        See also:
-            Python's built-in help(sdk) for full API documentation
-            sdk.features, sdk.bugs, sdk.spikes for work item managers
-        """
-        if topic is None:
-            return self._help_index()
-        return self._help_topic(topic)
-
-    def _help_index(self) -> str:
-        """Return overview of all available methods/collections."""
-        return """HtmlGraph SDK - Quick Reference
-
-COLLECTIONS (Work Items):
-  sdk.features     - Feature work items with builder support
-  sdk.bugs         - Bug reports
-  sdk.spikes       - Investigation and research spikes
-  sdk.chores       - Maintenance and chore tasks
-  sdk.epics        - Large bodies of work
-  sdk.phases       - Project phases
-
-COLLECTIONS (Non-Work):
-  sdk.sessions     - Agent sessions
-  sdk.tracks       - Work tracks with builder support
-  sdk.agents       - Agent information
-
-LEARNING (Active Learning):
-  sdk.patterns     - Workflow patterns (optimal/anti-pattern)
-  sdk.insights     - Session health insights
-  sdk.metrics      - Aggregated time-series metrics
-
-CORE METHODS:
-  sdk.summary()           - Get project summary
-  sdk.my_work()           - Get current agent's workload
-  sdk.next_task()         - Get next available task
-  sdk.reload()            - Reload all data from disk
-
-SESSION MANAGEMENT:
-  sdk.start_session()     - Start a new session
-  sdk.end_session()       - End a session
-  sdk.track_activity()    - Track activity in session
-  sdk.dedupe_sessions()   - Clean up low-signal sessions
-  sdk.get_status()        - Get project status
-
-STRATEGIC ANALYTICS:
-  sdk.find_bottlenecks()     - Identify blocking tasks
-  sdk.recommend_next_work()  - Get smart recommendations
-  sdk.get_parallel_work()    - Find parallelizable work
-  sdk.assess_risks()         - Assess dependency risks
-  sdk.analyze_impact()       - Analyze task impact
-
-WORK QUEUE:
-  sdk.get_work_queue()    - Get prioritized work queue
-  sdk.work_next()         - Get next best task (smart routing)
-
-PLANNING WORKFLOW:
-  sdk.smart_plan()              - Smart planning with research
-  sdk.start_planning_spike()    - Create planning spike
-  sdk.create_track_from_plan()  - Create track from plan
-  sdk.plan_parallel_work()      - Plan parallel execution
-  sdk.aggregate_parallel_results() - Aggregate parallel results
-
-ORCHESTRATION:
-  sdk.spawn_explorer()    - Spawn explorer subagent
-  sdk.spawn_coder()       - Spawn coder subagent
-  sdk.orchestrate()       - Orchestrate feature implementation
-
-SESSION OPTIMIZATION:
-  sdk.get_session_start_info() - Get comprehensive session start info
-  sdk.get_active_work_item()   - Get currently active work item
-
-ANALYTICS INTERFACES:
-  sdk.analytics        - Work type analytics
-  sdk.dep_analytics    - Dependency analytics
-  sdk.context          - Context analytics
-
-OPERATIONS (Server, Hooks, Events):
-  sdk.start_server()         - Start web server for graph browsing
-  sdk.stop_server()          - Stop running server
-  sdk.install_hooks()        - Install Git hooks for tracking
-  sdk.list_hooks()           - List Git hooks status
-  sdk.export_sessions()      - Export HTML sessions to JSONL
-  sdk.rebuild_event_index()  - Rebuild SQLite index from events
-  sdk.query_events()         - Query JSONL event logs
-  sdk.get_event_stats()      - Get event statistics
-  sdk.analyze_session()      - Analyze single session metrics
-  sdk.analyze_project()      - Analyze project-wide metrics
-  sdk.get_work_recommendations() - Get work recommendations
-
-ERROR HANDLING:
-  Lookup (.get)      - Returns None if not found
-  Query (.where)     - Returns empty list on no matches
-  Edit (.edit)       - Raises NodeNotFoundError if missing
-  Batch (.mark_done) - Returns dict with success_count, failed_ids, warnings
-
-For detailed help on a topic:
-  sdk.help('features')      - Feature collection methods
-  sdk.help('analytics')     - Analytics methods
-  sdk.help('sessions')      - Session management
-  sdk.help('orchestration') - Subagent orchestration
-  sdk.help('planning')      - Planning workflow
-  sdk.help('operations')    - Server, hooks, events operations
-"""
-
-    def __dir__(self) -> list[str]:
-        """Return attributes with most useful ones first for discoverability."""
-        priority = [
-            # Work item managers
-            "features",
-            "bugs",
-            "spikes",
-            "chores",
-            "epics",
-            "phases",
-            # Non-work collections
-            "tracks",
-            "sessions",
-            "agents",
-            # Learning collections
-            "patterns",
-            "insights",
-            "metrics",
-            # Orchestration
-            "spawn_explorer",
-            "spawn_coder",
-            "orchestrate",
-            # Session management
-            "get_session_start_info",
-            "start_session",
-            "end_session",
-            # Strategic analytics
-            "find_bottlenecks",
-            "recommend_next_work",
-            "get_parallel_work",
-            # Work queue
-            "get_work_queue",
-            "work_next",
-            # Operations
-            "start_server",
-            "install_hooks",
-            "export_sessions",
-            "analyze_project",
-            # Help
-            "help",
-        ]
-        # Get all attributes
-        all_attrs = object.__dir__(self)
-        # Separate into priority, regular, and dunder attributes
-        regular = [a for a in all_attrs if not a.startswith("_") and a not in priority]
-        dunder = [a for a in all_attrs if a.startswith("_")]
-        # Return priority items first, then regular, then dunder
-        return priority + regular + dunder
-
-    def _help_topic(self, topic: str) -> str:
-        """Return specific help for topic."""
-        topic = topic.lower()
-
-        if topic in ["feature", "features"]:
-            return """FEATURES COLLECTION
-
-Create and manage feature work items with builder support.
-
-COMMON METHODS:
-  sdk.features.create(title)     - Create new feature (returns builder)
-  sdk.features.get(id)           - Get feature by ID
-  sdk.features.all()             - Get all features
-  sdk.features.where(**filters)  - Query features
-  sdk.features.edit(id)          - Edit feature (context manager)
-  sdk.features.mark_done(ids)    - Mark features as done
-  sdk.features.assign(ids, agent) - Assign features to agent
-
-BUILDER PATTERN:
-  feature = (sdk.features.create("User Auth")
-    .set_priority("high")
-    .add_steps(["Login", "Logout", "Reset password"])
-    .add_edge("blocked_by", "feat-database")
-    .save())
-
-QUERIES:
-  high_priority = sdk.features.where(status="todo", priority="high")
-  my_features = sdk.features.where(agent_assigned="claude")
-  blocked = sdk.features.where(status="blocked")
-
-CONTEXT MANAGER:
-  with sdk.features.edit("feat-001") as f:
-      f.status = "in-progress"
-      f.complete_step(0)
-      # Auto-saves on exit
-
-BATCH OPERATIONS:
-  result = sdk.features.mark_done(["feat-001", "feat-002"])
-  print(f"Completed {result['success_count']} features")
-  if result['failed_ids']:
-      print(f"Failed: {result['failed_ids']}")
-
-COMMON MISTAKES:
-  ❌ sdk.features.mark_complete([ids])  → ✅ sdk.features.mark_done([ids])
-  ❌ sdk.feature.create(...)            → ✅ sdk.features.create(...)
-  ❌ claim(id, agent_id=...)            → ✅ claim(id, agent=...)
-  ❌ builder.status = "done"            → ✅ builder.save(); then edit()
-
-See also: sdk.help('bugs'), sdk.help('spikes'), sdk.help('chores')
-"""
-
-        elif topic in ["bug", "bugs"]:
-            return """BUGS COLLECTION
-
-Create and manage bug reports.
-
-COMMON METHODS:
-  sdk.bugs.create(title)         - Create new bug (returns builder)
-  sdk.bugs.get(id)               - Get bug by ID
-  sdk.bugs.all()                 - Get all bugs
-  sdk.bugs.where(**filters)      - Query bugs
-  sdk.bugs.edit(id)              - Edit bug (context manager)
-
-BUILDER PATTERN:
-  bug = (sdk.bugs.create("Login fails on Safari")
-    .set_priority("critical")
-    .add_steps(["Reproduce", "Fix", "Test"])
-    .save())
-
-QUERIES:
-  critical = sdk.bugs.where(priority="critical", status="todo")
-  my_bugs = sdk.bugs.where(agent_assigned="claude")
-
-See also: sdk.help('features'), sdk.help('spikes')
-"""
-
-        elif topic in ["spike", "spikes"]:
-            return """SPIKES COLLECTION
-
-Create and manage investigation/research spikes.
-
-COMMON METHODS:
-  sdk.spikes.create(title)       - Create new spike (returns builder)
-  sdk.spikes.get(id)             - Get spike by ID
-  sdk.spikes.all()               - Get all spikes
-  sdk.spikes.where(**filters)    - Query spikes
-
-BUILDER PATTERN:
-  spike = (sdk.spikes.create("Research OAuth providers")
-    .set_priority("high")
-    .add_steps(["Research", "Document findings"])
-    .save())
-
-PLANNING SPIKES:
-  spike = sdk.start_planning_spike(
-      "Plan User Auth",
-      context="Users need login",
-      timebox_hours=4.0
-  )
-
-See also: sdk.help('planning'), sdk.help('features')
-"""
-
-        elif topic in ["chore", "chores"]:
-            return """CHORES COLLECTION
-
-Create and manage maintenance and chore tasks.
-
-COMMON METHODS:
-  sdk.chores.create(title)       - Create new chore (returns builder)
-  sdk.chores.get(id)             - Get chore by ID
-  sdk.chores.all()               - Get all chores
-  sdk.chores.where(**filters)    - Query chores
-
-BUILDER PATTERN:
-  chore = (sdk.chores.create("Update dependencies")
-    .set_priority("medium")
-    .add_steps(["Run uv update", "Test", "Commit"])
-    .save())
-
-See also: sdk.help('features'), sdk.help('bugs')
-"""
-
-        elif topic in ["epic", "epics"]:
-            return """EPICS COLLECTION
-
-Create and manage large bodies of work.
-
-COMMON METHODS:
-  sdk.epics.create(title)        - Create new epic (returns builder)
-  sdk.epics.get(id)              - Get epic by ID
-  sdk.epics.all()                - Get all epics
-  sdk.epics.where(**filters)     - Query epics
-
-BUILDER PATTERN:
-  epic = (sdk.epics.create("Authentication System")
-    .set_priority("critical")
-    .add_steps(["Design", "Implement", "Test", "Deploy"])
-    .save())
-
-See also: sdk.help('features'), sdk.help('tracks')
-"""
-
-        elif topic in ["track", "tracks"]:
-            return """TRACKS COLLECTION
-
-Create and manage work tracks with builder support.
-
-COMMON METHODS:
-  sdk.tracks.create(title)       - Create new track (returns builder)
-  sdk.tracks.builder()           - Get track builder
-  sdk.tracks.get(id)             - Get track by ID
-  sdk.tracks.all()               - Get all tracks
-  sdk.tracks.where(**filters)    - Query tracks
-
-BUILDER PATTERN:
-  track = (sdk.tracks.builder()
-    .title("User Authentication")
-    .description("OAuth 2.0 system")
-    .priority("high")
-    .with_spec(
-        overview="OAuth integration",
-        requirements=[("OAuth 2.0", "must-have")],
-        acceptance_criteria=["Login works"]
-    )
-    .with_plan_phases([
-        ("Phase 1", ["Setup (2h)", "Config (1h)"]),
-        ("Phase 2", ["Testing (2h)"])
-    ])
-    .create())
-
-FROM PLANNING:
-  track_info = sdk.create_track_from_plan(
-      title="User Auth",
-      description="OAuth system",
-      requirements=[("OAuth", "must-have")],
-      phases=[("Phase 1", ["Setup", "Config"])]
-  )
-
-See also: sdk.help('planning'), sdk.help('features')
-"""
-
-        elif topic in ["session", "sessions"]:
-            return """SESSION MANAGEMENT
-
-Create and manage agent sessions.
-
-SESSION METHODS:
-  sdk.start_session(title=...)   - Start new session
-  sdk.end_session(id)            - End session
-  sdk.track_activity(...)        - Track activity in session
-  sdk.dedupe_sessions(...)       - Clean up low-signal sessions
-  sdk.get_status()               - Get project status
-
-SESSION COLLECTION:
-  sdk.sessions.get(id)           - Get session by ID
-  sdk.sessions.all()             - Get all sessions
-  sdk.sessions.where(**filters)  - Query sessions
-
-TYPICAL WORKFLOW:
-  # Session start hook handles this automatically
-  session = sdk.start_session(title="Fix login bug")
-
-  # Track activities (handled by hooks)
-  sdk.track_activity(
-      tool="Edit",
-      summary="Fixed auth logic",
-      file_paths=["src/auth.py"],
-      success=True
-  )
-
-  # End session
-  sdk.end_session(
-      session.id,
-      handoff_notes="Login bug fixed, needs testing"
-  )
-
-CLEANUP:
-  # Remove orphaned sessions (<=1 event)
-  result = sdk.dedupe_sessions(max_events=1, dry_run=False)
-
-See also: sdk.help('analytics')
-"""
-
-        elif topic in ["analytic", "analytics", "strategic"]:
-            return """STRATEGIC ANALYTICS
-
-Find bottlenecks, recommend work, and assess risks.
-
-DEPENDENCY ANALYTICS:
-  bottlenecks = sdk.find_bottlenecks(top_n=5)
-  # Returns tasks blocking the most work
-
-  parallel = sdk.get_parallel_work(max_agents=5)
-  # Returns tasks that can run simultaneously
-
-  recs = sdk.recommend_next_work(agent_count=3)
-  # Returns smart recommendations with scoring
-
-  risks = sdk.assess_risks()
-  # Returns high-risk tasks and circular deps
-
-  impact = sdk.analyze_impact("feat-001")
-  # Returns what unlocks if you complete this task
-
-DIRECT ACCESS (preferred):
-  sdk.dep_analytics.find_bottlenecks(top_n=5)
-  sdk.dep_analytics.recommend_next_tasks(agent_count=3)
-  sdk.dep_analytics.find_parallelizable_work(status="todo")
-  sdk.dep_analytics.assess_dependency_risk()
-  sdk.dep_analytics.impact_analysis("feat-001")
-
-WORK TYPE ANALYTICS:
-  sdk.analytics.get_wip_by_type()
-  sdk.analytics.get_completion_rates()
-  sdk.analytics.get_agent_workload()
-
-CONTEXT ANALYTICS:
-  sdk.context.track_usage(...)
-  sdk.context.get_usage_report()
-
-See also: sdk.help('planning'), sdk.help('work_queue')
-"""
-
-        elif topic in ["queue", "work_queue", "routing"]:
-            return """WORK QUEUE & ROUTING
-
-Get prioritized work using smart routing.
-
-WORK QUEUE:
-  queue = sdk.get_work_queue(limit=10, min_score=0.0)
-  # Returns prioritized list with scores
-
-  for item in queue:
-      print(f"{item['score']:.1f} - {item['title']}")
-      if item.get('blocked_by'):
-          print(f"  Blocked by: {item['blocked_by']}")
-
-SMART ROUTING:
-  task = sdk.work_next(auto_claim=True, min_score=0.5)
-  # Returns next best task using analytics + capabilities
-
-  if task:
-      print(f"Working on: {task.title}")
-      # Task is auto-claimed and assigned
-
-SIMPLE NEXT TASK:
-  task = sdk.next_task(priority="high", auto_claim=True)
-  # Simpler version without smart routing
-
-See also: sdk.help('analytics')
-"""
-
-        elif topic in ["plan", "planning", "workflow"]:
-            return """PLANNING WORKFLOW
-
-Research, plan, and create tracks for new work.
-
-SMART PLANNING:
-  plan = sdk.smart_plan(
-      "User authentication system",
-      create_spike=True,
-      timebox_hours=4.0,
-      research_completed=True,  # IMPORTANT: Do research first!
-      research_findings={
-          "topic": "OAuth 2.0 best practices",
-          "recommended_library": "authlib",
-          "key_insights": ["Use PKCE", "Token rotation"]
-      }
-  )
-
-PLANNING SPIKE:
-  spike = sdk.start_planning_spike(
-      "Plan Real-time Notifications",
-      context="Users need live updates",
-      timebox_hours=3.0
-  )
-
-CREATE TRACK FROM PLAN:
-  track_info = sdk.create_track_from_plan(
-      title="User Authentication",
-      description="OAuth 2.0 with JWT",
-      requirements=[
-          ("OAuth 2.0 integration", "must-have"),
-          ("JWT token management", "must-have")
-      ],
-      phases=[
-          ("Phase 1: OAuth", ["Setup (2h)", "Callback (2h)"]),
-          ("Phase 2: JWT", ["Token signing (2h)"])
-      ]
-  )
-
-PARALLEL PLANNING:
-  plan = sdk.plan_parallel_work(max_agents=3)
-  if plan["can_parallelize"]:
-      for p in plan["prompts"]:
-          Task(prompt=p["prompt"])
-
-  # After parallel work completes
-  results = sdk.aggregate_parallel_results([
-      "agent-1", "agent-2", "agent-3"
-  ])
-
-See also: sdk.help('tracks'), sdk.help('spikes')
-"""
-
-        elif topic in ["orchestration", "orchestrate", "subagent", "subagents"]:
-            return """SUBAGENT ORCHESTRATION
-
-Spawn explorer and coder subagents for complex work.
-
-EXPLORER (Discovery):
-  prompt = sdk.spawn_explorer(
-      task="Find all API endpoints",
-      scope="src/api/",
-      patterns=["*.py"],
-      questions=["What framework is used?"]
-  )
-  # Execute with Task tool
-  Task(prompt=prompt["prompt"], description=prompt["description"])
-
-CODER (Implementation):
-  prompt = sdk.spawn_coder(
-      feature_id="feat-add-auth",
-      context=explorer_results,
-      files_to_modify=["src/auth.py"],
-      test_command="uv run pytest tests/auth/"
-  )
-  Task(prompt=prompt["prompt"], description=prompt["description"])
-
-FULL ORCHESTRATION:
-  prompts = sdk.orchestrate(
-      "feat-add-caching",
-      exploration_scope="src/cache/",
-      test_command="uv run pytest tests/cache/"
-  )
-
-  # Phase 1: Explorer
-  Task(prompt=prompts["explorer"]["prompt"])
-
-  # Phase 2: Coder (with explorer results)
-  Task(prompt=prompts["coder"]["prompt"])
-
-WORKFLOW:
-  1. Explorer discovers code patterns and files
-  2. Coder implements changes using explorer findings
-  3. Both agents auto-track in sessions
-  4. Feature gets updated with progress
-
-See also: sdk.help('planning')
-"""
-
-        elif topic in ["optimization", "session_start", "active_work"]:
-            return """SESSION OPTIMIZATION
-
-Reduce context usage with optimized methods.
-
-SESSION START INFO:
-  info = sdk.get_session_start_info(
-      include_git_log=True,
-      git_log_count=5,
-      analytics_top_n=3
-  )
-
-  # Single call returns:
-  # - status: Project status
-  # - active_work: Current work item
-  # - features: All features
-  # - sessions: Recent sessions
-  # - git_log: Recent commits
-  # - analytics: Bottlenecks, recommendations, parallel
-
-ACTIVE WORK ITEM:
-  active = sdk.get_active_work_item()
-  if active:
-      print(f"Working on: {active['title']}")
-      print(f"Progress: {active['steps_completed']}/{active['steps_total']}")
-
-  # Filter by agent
-  active = sdk.get_active_work_item(filter_by_agent=True)
-
-BENEFITS:
-  - 6+ tool calls → 1 method call
-  - Reduced token usage
-  - Faster session initialization
-  - All context in one place
-
-See also: sdk.help('sessions')
-"""
-
-        elif topic in ["operation", "operations", "server", "hooks", "events"]:
-            return """OPERATIONS - Server, Hooks, Events
-
-Infrastructure operations for running HtmlGraph.
-
-SERVER OPERATIONS:
-  # Start server for web UI
-  result = sdk.start_server(port=8080, watch=True)
-  print(f"Server at {result.handle.url}")
-
-  # Stop server
-  sdk.stop_server(result.handle)
-
-  # Check status
-  status = sdk.get_server_status(result.handle)
-
-HOOK OPERATIONS:
-  # Install Git hooks for automatic tracking
-  result = sdk.install_hooks()
-  print(f"Installed: {result.installed}")
-
-  # List hook status
-  result = sdk.list_hooks()
-  print(f"Enabled: {result.enabled}")
-  print(f"Missing: {result.missing}")
-
-  # Validate configuration
-  result = sdk.validate_hook_config()
-  if not result.valid:
-      print(f"Errors: {result.errors}")
-
-EVENT OPERATIONS:
-  # Export HTML sessions to JSONL
-  result = sdk.export_sessions()
-  print(f"Exported {result.written} sessions")
-
-  # Rebuild SQLite index
-  result = sdk.rebuild_event_index()
-  print(f"Inserted {result.inserted} events")
-
-  # Query events
-  result = sdk.query_events(
-      session_id="sess-123",
-      tool="Bash",
-      limit=10
-  )
-  for event in result.events:
-      print(f"{event['timestamp']}: {event['summary']}")
-
-  # Get statistics
-  stats = sdk.get_event_stats()
-  print(f"Total events: {stats.total_events}")
-
-ANALYTICS OPERATIONS:
-  # Analyze session
-  result = sdk.analyze_session("sess-123")
-  print(f"Primary work: {result.metrics['primary_work_type']}")
-
-  # Analyze project
-  result = sdk.analyze_project()
-  print(f"Total sessions: {result.metrics['total_sessions']}")
-  print(f"Work distribution: {result.metrics['work_distribution']}")
-
-  # Get recommendations
-  result = sdk.get_work_recommendations()
-  for rec in result.recommendations:
-      print(f"{rec['title']} (score: {rec['score']})")
-
-See also: sdk.help('analytics'), sdk.help('sessions')
-"""
-
-        else:
-            return f"""Unknown topic: '{topic}'
-
-Available topics:
-  - features, bugs, spikes, chores, epics (work collections)
-  - tracks, sessions, agents (non-work collections)
-  - analytics, strategic (dependency and work analytics)
-  - work_queue, routing (smart task routing)
-  - planning, workflow (planning and track creation)
-  - orchestration, subagents (explorer/coder spawning)
-  - optimization, session_start (context optimization)
-
-Try: sdk.help() for full overview
-"""