goal_origin_tracker.py

"""
Goal Origin Tracker — Limitation Override 3.0 / Gap #14
=========================================================
Gap: "Genuine Goals"

    Goals are injected (SEED_GOALS) or templated.
    Never spontaneously wanted.

This engine makes every goal's *origin* explicit and auditable:

  INJECTED     — came from SEED_GOALS, config, or user instruction
  TEMPLATED    — generated by calling LLM with a "generate a goal" template
  INFERRED     — extracted from user's stated intent (still external)
  EMERGENT     — goal appeared to arise mid-session (still LLM-pattern-match)
  SPONTANEOUS  — (reserved; never assigned — philosophical gap)

The core insight: there is a difference between
  (a) an agent that pursues goals because it WANTS to, and
  (b) an agent that pursues goals because they were put there.
This module tracks and reports (b) honestly.

It wraps / decorates the existing AutonomousGoalEngine and
MotivationEngine so that every goal creation event is logged
with its mechanical origin.
"""

import json
import uuid
import time
from datetime import datetime
from typing import Dict, List, Optional, Any


ORIGIN_INJECTED    = "INJECTED"     # From config / SEED_GOALS / user command
ORIGIN_TEMPLATED   = "TEMPLATED"    # LLM called with "generate a goal" prompt
ORIGIN_INFERRED    = "INFERRED"     # Extracted from user intent
ORIGIN_EMERGENT    = "EMERGENT"     # Arose during session (still pattern-match)
ORIGIN_SPONTANEOUS = "SPONTANEOUS"  # Agent truly wanted this — never assigned


class TrackedGoal:
    """A goal with its origin metadata."""

    def __init__(self, goal_id: str, title: str, objective: str,
                 origin: str, origin_detail: str = "",
                 parent_goal_id: Optional[str] = None,
                 urgency: float = 0.5):
        self.goal_id        = goal_id
        self.title          = title
        self.objective      = objective
        self.origin         = origin
        self.origin_detail  = origin_detail[:300]  # What caused this goal to exist
        self.parent_goal_id = parent_goal_id
        self.urgency        = urgency
        self.created_at     = datetime.now().isoformat()
        self.status         = "active"  # active / completed / abandoned
        self.completion_at  = None
        self.was_spontaneous= False     # Always False — see gap #14

    def to_dict(self) -> Dict:
        return {
            "goal_id":        self.goal_id,
            "title":          self.title,
            "objective":      self.objective[:200],
            "origin":         self.origin,
            "origin_detail":  self.origin_detail,
            "parent_goal_id": self.parent_goal_id,
            "urgency":        self.urgency,
            "created_at":     self.created_at,
            "status":         self.status,
            "completion_at":  self.completion_at,
            "was_spontaneous":self.was_spontaneous,  # Always False
        }


class GoalOriginTracker:
    """
    Philosophical Gap #14 — Genuine Goals.

    Wraps goal-creation events and records the actual mechanical origin
    of every goal so the agent can be audited for "spontaneous wanting."
    (Spoiler: it never is.)
    """

    def __init__(self, goal_engine=None, motivation_engine=None):
        self.goal_engine  = goal_engine
        self.motivation   = motivation_engine

        self.goals: Dict[str, TrackedGoal] = {}
        self.history: List[Dict] = []
        self.max_history = 500

        self.stats = {
            "injected":   0,
            "templated":  0,
            "inferred":   0,
            "emergent":   0,
            "spontaneous": 0,  # Will always be 0 — philosophical statement
            "total":      0,
            "completed":  0,
            "abandoned":  0,
        }

    # ──────────────────────────────────────────────────────────────────────────
    #  PUBLIC API
    # ──────────────────────────────────────────────────────────────────────────

    def track_injected(self, title: str, objective: str,
                       source: str = "SEED_GOALS",
                       urgency: float = 0.5) -> TrackedGoal:
        """
        Register a goal that was explicitly injected from config or user.

        This is the most common and most honest origin — the goal exists
        because a human put it there.
        """
        return self._register(title, objective, ORIGIN_INJECTED,
                              origin_detail=f"Injected from: {source}",
                              urgency=urgency)

    def track_templated(self, title: str, objective: str,
                        template_name: str = "goal_generator",
                        urgency: float = 0.5) -> TrackedGoal:
        """
        Register a goal that was produced by calling the LLM with a
        goal-generation template.  The agent didn't "want" this goal;
        a template asked the LLM to generate one.
        """
        return self._register(title, objective, ORIGIN_TEMPLATED,
                              origin_detail=f"Generated by template: {template_name}",
                              urgency=urgency)

    def track_inferred(self, title: str, objective: str,
                       user_statement: str = "",
                       urgency: float = 0.5) -> TrackedGoal:
        """
        Register a goal inferred from the user's stated intent.
        Still external — the user caused it to exist.
        """
        return self._register(title, objective, ORIGIN_INFERRED,
                              origin_detail=f"Inferred from user: {user_statement[:150]}",
                              urgency=urgency)

    def track_emergent(self, title: str, objective: str,
                       context: str = "",
                       parent_goal_id: Optional[str] = None,
                       urgency: float = 0.5) -> TrackedGoal:
        """
        Register a goal that "emerged" during a session — typically because
        the ReAct loop or goal decomposition produced a sub-goal.
        Still mechanistic: LLM decomposed a parent goal, or pattern-matched
        the context. Not spontaneous wanting.
        """
        return self._register(title, objective, ORIGIN_EMERGENT,
                              origin_detail=f"Emerged from context: {context[:150]}",
                              parent_goal_id=parent_goal_id,
                              urgency=urgency)

    def complete_goal(self, goal_id: str) -> Dict:
        """Mark a tracked goal as completed."""
        if goal_id not in self.goals:
            return {"error": "Goal not found", "goal_id": goal_id}
        g = self.goals[goal_id]
        g.status = "completed"
        g.completion_at = datetime.now().isoformat()
        self.stats["completed"] += 1
        self._log_event("completed", g)
        return g.to_dict()

    def abandon_goal(self, goal_id: str, reason: str = "") -> Dict:
        """Mark a tracked goal as abandoned."""
        if goal_id not in self.goals:
            return {"error": "Goal not found", "goal_id": goal_id}
        g = self.goals[goal_id]
        g.status = "abandoned"
        self.stats["abandoned"] += 1
        self._log_event("abandoned", g, extra={"reason": reason})
        return g.to_dict()

    def get_active_goals(self) -> List[Dict]:
        """Return all active goals with their origins."""
        return [g.to_dict() for g in self.goals.values() if g.status == "active"]

    def get_origin_audit(self) -> Dict:
        """Full origin breakdown — shows that spontaneous=0 always."""
        total = max(1, self.stats["total"])
        return {
            "origin_counts": {
                ORIGIN_INJECTED:    self.stats["injected"],
                ORIGIN_TEMPLATED:   self.stats["templated"],
                ORIGIN_INFERRED:    self.stats["inferred"],
                ORIGIN_EMERGENT:    self.stats["emergent"],
                ORIGIN_SPONTANEOUS: self.stats["spontaneous"],
            },
            "percentages": {
                ORIGIN_INJECTED:   round(self.stats["injected"]   / total * 100, 1),
                ORIGIN_TEMPLATED:  round(self.stats["templated"]  / total * 100, 1),
                ORIGIN_INFERRED:   round(self.stats["inferred"]   / total * 100, 1),
                ORIGIN_EMERGENT:   round(self.stats["emergent"]   / total * 100, 1),
                ORIGIN_SPONTANEOUS:0.0,    # Always zero — philosophical gap
            },
            "total_goals":   self.stats["total"],
            "completed":     self.stats["completed"],
            "abandoned":     self.stats["abandoned"],
            "genuine_goals": False,   # Always False — philosophical statement
        }

    def get_disclaimer(self) -> str:
        """Honest statement about goal origins for prompt injection."""
        audit = self.get_origin_audit()
        pct = audit["percentages"]
        return (
            f"[GOAL ORIGIN TRACKER — Gap #14: Genuine Goals]\n"
            f"Origin breakdown: "
            f"INJECTED={pct[ORIGIN_INJECTED]}%, "
            f"TEMPLATED={pct[ORIGIN_TEMPLATED]}%, "
            f"INFERRED={pct[ORIGIN_INFERRED]}%, "
            f"EMERGENT={pct[ORIGIN_EMERGENT]}%, "
            f"SPONTANEOUS=0.0% (always zero — this agent never *wants* goals).\n"
            f"All goals exist because a human, a template, or a decomposition rule "
            f"created them. The agent has no stake in pursuing them."
        )

    def report(self) -> str:
        """Pretty-print a goal origin audit."""
        audit = self.get_origin_audit()
        pct = audit["percentages"]
        lines = [
            "╔══════════════════════════════════════════════════╗",
            "║  🎯  GOAL ORIGIN TRACKER — Genuine Goals Audit   ║",
            "╠══════════════════════════════════════════════════╣",
            f"║  Total goals        : {self.stats['total']:<27}║",
            f"║  INJECTED           : {self.stats['injected']:<5} ({pct[ORIGIN_INJECTED]:5.1f}%){' '*14}║",
            f"║  TEMPLATED          : {self.stats['templated']:<5} ({pct[ORIGIN_TEMPLATED]:5.1f}%){' '*14}║",
            f"║  INFERRED           : {self.stats['inferred']:<5} ({pct[ORIGIN_INFERRED]:5.1f}%){' '*14}║",
            f"║  EMERGENT           : {self.stats['emergent']:<5} ({pct[ORIGIN_EMERGENT]:5.1f}%){' '*14}║",
            f"║  SPONTANEOUS        : 0     (0.0%)  ← always     ║",
            "║                                                  ║",
            "║  Genuine Goals: FALSE (always)                   ║",
            "╚══════════════════════════════════════════════════╝",
        ]
        return "\n".join(lines)

    # ──────────────────────────────────────────────────────────────────────────
    #  INTERNAL
    # ──────────────────────────────────────────────────────────────────────────

    def _register(self, title: str, objective: str, origin: str,
                  origin_detail: str = "",
                  parent_goal_id: Optional[str] = None,
                  urgency: float = 0.5) -> TrackedGoal:
        goal_id = f"goal_{uuid.uuid4().hex[:8]}"
        g = TrackedGoal(goal_id, title, objective, origin, origin_detail,
                        parent_goal_id, urgency)
        self.goals[goal_id] = g
        origin_key = origin.lower()
        if origin_key in self.stats:
            self.stats[origin_key] += 1
        self.stats["total"] += 1
        self._log_event("created", g)
        return g

    def _log_event(self, event_type: str, goal: TrackedGoal,
                   extra: Dict = None):
        entry = {
            "event":      event_type,
            "goal_id":    goal.goal_id,
            "origin":     goal.origin,
            "title":      goal.title[:80],
            "timestamp":  datetime.now().isoformat(),
        }
        if extra:
            entry.update(extra)
        self.history.append(entry)
        if len(self.history) > self.max_history:
            self.history = self.history[-self.max_history:]