ltcai 11.5.2 → 11.7.0

package/latticeai/services/automation_intelligence.py

@@ -1,614 +0,0 @@
-"""Question-driven everyday automation intelligence (v9.4.0).
-
-Users repeat themselves: the same "오늘 뭐 해야 하지?", "프로젝트 상태 정리해줘",
-"그 폴더에 새 문서 뭐 들어왔어?" questions come back every day. This service
-watches for those patterns and turns them into one-click, consent-first
-automation suggestions:
-
-* **question_patterns** — mines the user's own chat history for recurring
-  question intents. Clustering is deterministic and local (token-signature
-  similarity, no model call), so the evidence shown to the user is their own
-  literal questions.
-* **suggestions** — converts recurring patterns and active local knowledge
-  sources (folders the user connected to the Brain) into concrete automation
-  proposals: a scheduled answer for a recurring question, a digest for a busy
-  knowledge folder, or one of the starter recipes when the intent matches.
-* **build_suggestion_workflow** — turns an accepted suggestion into a
-  workflow definition using the same consent-first shape as the starter
-  recipes: installed as a disabled draft, review-queue gated, local-only,
-  no external actions.
-* **overview** — one payload for the automation surface: suggestions,
-  installed automations with their enable state, and pattern evidence.
-
-Every suggestion id is deterministic, so install calls are idempotent and the
-UI can safely re-request suggestions without duplicates.
-
-v9.8.0 quality layer (additive): every suggestion carries a deterministic
-``confidence`` score plus its ``confidence_factors`` evidence (repeat count,
-distinct phrasings, intent match, related Brain nodes / indexed files);
-suggestions below a minimum confidence are suppressed, duplicate suggestions
-targeting an already-suggested or already-installed starter recipe are
-deduplicated, and the response reports the suppression counters under
-``quality``.
-"""
-
-from __future__ import annotations
-
-import hashlib
-import logging
-import re
-from dataclasses import dataclass, field
-from typing import Any, Dict, List, Optional
-
-from latticeai.core.timeutil import now_iso as _now
-
-LOGGER = logging.getLogger(__name__)
-
-_MIN_PATTERN_COUNT = 2
-_MAX_HISTORY = 4000
-_SIGNATURE_SIMILARITY = 0.6
-
-# Suggestion quality gates (v9.8.0, additive). Suggestions below the minimum
-# confidence are suppressed (insufficient evidence); suggestions between the
-# minimum and the low-confidence threshold are shown but flagged so the UI
-# can render them less prominently.
-_MIN_SUGGESTION_CONFIDENCE = 0.35
-_LOW_CONFIDENCE_THRESHOLD = 0.5
-_KG_GROUNDING_LIMIT = 5
-#: Charged when the graph *was* consulted and found nothing related. v11.2.0
-#: fixes a dead policy: the floor above was unreachable for recurring questions
-#: (the cheapest possible question — asked twice, one phrasing, no recipe —
-#: scored 0.475), so the suppression branch could only ever be exercised by a
-#: test that moved the constant. Absence of grounding is real evidence, and
-#: charging for it is what makes the floor mean something again.
-_UNGROUNDED_PENALTY = 0.15
-
-
-def _question_confidence(
-    count: int,
-    examples: List[str],
-    recipe_id: Optional[str],
-    kg_related: Optional[int],
-) -> tuple:
-    """Deterministic confidence for a recurring-question suggestion.
-
-    Evidence factors: how often the question repeats, how many distinct
-    phrasings exist, whether it maps onto a known starter-recipe intent, and
-    what the Brain knows about it.
-
-    That last factor is three-valued on purpose. ``None`` means the graph could
-    not be consulted — unknown, so nothing is added or taken away. A positive
-    count is supporting evidence. **Zero** is not the same as unknown: the
-    graph was asked, and it knows nothing about this question, which is a
-    reason to be less sure rather than a reason to be neutral.
-    """
-    score = 0.3 + 0.5 * min(1.0, (int(count) - 1) / 4)
-    score += min(0.15, 0.05 * len(examples or []))
-    if recipe_id:
-        score += 0.15
-    if kg_related:
-        score += min(0.2, 0.05 * int(kg_related))
-    elif kg_related == 0:
-        score -= _UNGROUNDED_PENALTY
-    factors = {
-        "repeat_count": int(count),
-        "distinct_examples": len(examples or []),
-        "intent_match": bool(recipe_id),
-        "kg_related_nodes": kg_related,
-        "kg_grounded": None if kg_related is None else bool(kg_related),
-    }
-    return round(max(0.0, min(1.0, score)), 2), factors
-
-
-def _source_confidence(indexed: int, watch_enabled: bool) -> tuple:
-    """Deterministic confidence for a knowledge-source digest suggestion."""
-    score = 0.25 + 0.6 * min(1.0, int(indexed) / 25)
-    if watch_enabled:
-        score += 0.1
-    factors = {"indexed_files": int(indexed), "watch_enabled": bool(watch_enabled)}
-    return round(min(1.0, score), 2), factors
-
-_QUESTION_HINT_RE = re.compile(
-    r"(\?|어때|뭐야|뭐가|뭘까|알려줘|보여줘|정리해|요약해|정리 좀|요약 좀|해줘"
-    r"|what|how|why|when|where|status|summar|remind|list|show me|tell me)",
-    re.IGNORECASE,
-)
-
-_STOPWORDS = {
-    "the", "a", "an", "is", "are", "was", "to", "of", "in", "on", "for", "and",
-    "or", "me", "my", "you", "please", "can", "could", "would", "it", "this",
-    "that", "do", "does", "what", "how", "about",
-    "좀", "그", "이", "저", "것", "거", "게", "내", "제", "나", "너", "우리",
-    "해줘", "해주세요", "주세요", "합니다", "있어", "있나", "있는지", "어떻게",
-    "뭐야", "뭐가", "알려줘", "보여줘",
-}
-
-# Intent buckets map recurring question language onto the starter recipes.
-_INTENT_RULES = (
-    (
-        "digest",
-        re.compile(r"(오늘|하루|today|daily).{0,12}(정리|요약|digest|summary|기억|메모)|"
-                   r"(정리|요약).{0,8}(해줘|해 줘|부탁)|summar(y|ize)", re.IGNORECASE),
-        "daily-memory-digest",
-    ),
-    (
-        "project_review",
-        re.compile(r"(프로젝트|project|진행|progress|status|상태|주간|weekly|이번 주)", re.IGNORECASE),
-        "weekly-project-review",
-    ),
-    (
-        "follow_up",
-        re.compile(r"(리마인드|remind|챙겨|잊지|deadline|마감|follow.?up|나중에|까먹)", re.IGNORECASE),
-        "follow-up-radar",
-    ),
-)
-
-
-def _tokens(text: str) -> List[str]:
-    return [
-        token
-        for token in re.findall(r"[\w가-힣]+", str(text or "").lower())
-        if len(token) > 1 and token not in _STOPWORDS
-    ]
-
-
-def _signature(tokens: List[str]) -> frozenset:
-    return frozenset(tokens)
-
-
-def _similarity(left: frozenset, right: frozenset) -> float:
-    if not left or not right:
-        return 0.0
-    return len(left & right) / len(left | right)
-
-
-def _stable_id(prefix: str, seed: str) -> str:
-    return f"{prefix}-{hashlib.sha256(seed.encode('utf-8')).hexdigest()[:10]}"
-
-
-@dataclass
-class _Pattern:
-    representative: str
-    signature: frozenset
-    count: int = 1
-    last_asked: str = ""
-    intent: str = "recurring_question"
-    recipe_id: Optional[str] = None
-    examples: List[str] = field(default_factory=list)
-
-    def as_dict(self) -> Dict[str, Any]:
-        return {
-            "id": _stable_id("pat", " ".join(sorted(self.signature))),
-            "representative": self.representative[:160],
-            "count": self.count,
-            "last_asked": self.last_asked,
-            "intent": self.intent,
-            "recipe_id": self.recipe_id,
-            "examples": self.examples[:3],
-        }
-
-
-class AutomationIntelligenceService:
-    def __init__(
-        self,
-        *,
-        conversation_store: Any = None,
-        knowledge_graph: Any = None,
-        store: Any = None,
-        enable_graph: bool = True,
-    ) -> None:
-        self._conversations = conversation_store
-        self._kg = knowledge_graph
-        self._store = store
-        self._enable_graph = bool(enable_graph and knowledge_graph is not None)
-
-    # ── question pattern mining ──────────────────────────────────────────
-
-    def _user_questions(
-        self, *, user_email: Optional[str], workspace_id: Optional[str]
-    ) -> List[Dict[str, Any]]:
-        if self._conversations is None:
-            return []
-        try:
-            items = self._conversations.history(
-                user_email=user_email,
-                allowed_workspaces={workspace_id} if workspace_id is not None else None,
-                include_legacy_global=workspace_id is None,
-                limit=_MAX_HISTORY,
-            )
-        except Exception:
-            LOGGER.exception("automation intelligence history read failed")
-            return []
-        questions = []
-        for item in items:
-            if item.get("role") != "user":
-                continue
-            content = str(item.get("content") or "").strip()
-            if not content or len(content) < 6 or content.startswith("/"):
-                continue
-            if not _QUESTION_HINT_RE.search(content):
-                continue
-            questions.append({"content": content, "timestamp": str(item.get("timestamp") or "")})
-        return questions
-
-    def question_patterns(
-        self, *, user_email: Optional[str] = None, workspace_id: Optional[str] = None
-    ) -> Dict[str, Any]:
-        questions = self._user_questions(user_email=user_email, workspace_id=workspace_id)
-        patterns: List[_Pattern] = []
-        for question in questions:
-            tokens = _tokens(question["content"])
-            if len(tokens) < 2:
-                continue
-            signature = _signature(tokens)
-            match = None
-            for pattern in patterns:
-                if _similarity(pattern.signature, signature) >= _SIGNATURE_SIMILARITY:
-                    match = pattern
-                    break
-            if match is None:
-                patterns.append(
-                    _Pattern(
-                        representative=question["content"],
-                        signature=signature,
-                        last_asked=question["timestamp"],
-                        examples=[question["content"]],
-                    )
-                )
-                continue
-            match.count += 1
-            # Union keeps the cluster stable as phrasing drifts.
-            match.signature = match.signature | signature
-            if question["timestamp"] >= match.last_asked:
-                match.last_asked = question["timestamp"]
-                match.representative = question["content"]
-            if question["content"] not in match.examples:
-                match.examples.append(question["content"])
-
-        recurring = [p for p in patterns if p.count >= _MIN_PATTERN_COUNT]
-        for pattern in recurring:
-            for intent, rule, recipe_id in _INTENT_RULES:
-                if rule.search(pattern.representative):
-                    pattern.intent = intent
-                    pattern.recipe_id = recipe_id
-                    break
-
-        recurring.sort(key=lambda p: (p.count, p.last_asked), reverse=True)
-        return {
-            "questions_scanned": len(questions),
-            "patterns": [p.as_dict() for p in recurring[:20]],
-            "generated_at": _now(),
-        }
-
-    # ── knowledge sources ────────────────────────────────────────────────
-
-    def _knowledge_sources(self) -> List[Dict[str, Any]]:
-        if not self._enable_graph or not hasattr(self._kg, "local_sources"):
-            return []
-        try:
-            return list(self._kg.local_sources().get("sources") or [])
-        except Exception:
-            LOGGER.exception("automation intelligence source read failed")
-            return []
-
-    def _kg_related_count(
-        self, text: str, *, workspace_id: Optional[str]
-    ) -> Optional[int]:
-        """Count Brain nodes related to a recurring question (grounding evidence).
-
-        Returns ``None`` when the graph is unavailable so confidence scoring can
-        distinguish "grounding impossible" from "grounded with zero hits".
-        Scoping mirrors the conversation-history read: an explicit workspace is
-        strict, no workspace falls back to the unscoped/legacy view.
-        """
-        if not self._enable_graph or not hasattr(self._kg, "search"):
-            return None
-        try:
-            report = self._kg.search(
-                str(text or "")[:200],
-                limit=_KG_GROUNDING_LIMIT,
-                allowed_workspaces={workspace_id} if workspace_id is not None else None,
-                include_legacy_global=workspace_id is None,
-            )
-        except Exception:
-            LOGGER.exception("automation intelligence KG grounding failed")
-            return None
-        if not isinstance(report, dict):
-            return None
-        return len(report.get("matches") or [])
-
-    # ── suggestions ──────────────────────────────────────────────────────
-
-    def _installed_workflows(
-        self, *, workspace_id: Optional[str]
-    ) -> tuple:
-        """Map installed automation workflows by suggestion id and recipe id.
-
-        The recipe map lets a recurring-question suggestion that targets an
-        already-installed starter recipe surface as *installed* instead of
-        re-suggesting the same automation — the install API is idempotent on
-        exactly this provenance, so the read side must agree.
-        """
-        by_suggestion: Dict[str, Dict[str, Any]] = {}
-        by_recipe: Dict[str, Dict[str, Any]] = {}
-        if self._store is None:
-            return by_suggestion, by_recipe
-        try:
-            workflows = self._store.list_workflows(workspace_id=workspace_id).get("workflows") or []
-        except Exception:
-            LOGGER.exception("automation intelligence workflow read failed")
-            return by_suggestion, by_recipe
-        for workflow in workflows:
-            metadata = (workflow or {}).get("metadata") or {}
-            created_from = metadata.get("created_from")
-            if created_from == "automation_suggestion" and metadata.get("suggestion_id"):
-                by_suggestion[str(metadata["suggestion_id"])] = workflow
-            elif created_from == "brain_automation_recipe" and metadata.get("recipe_id"):
-                by_recipe[str(metadata["recipe_id"])] = workflow
-        return by_suggestion, by_recipe
-
-    def suggestions(
-        self, *, user_email: Optional[str] = None, workspace_id: Optional[str] = None
-    ) -> Dict[str, Any]:
-        pattern_report = self.question_patterns(user_email=user_email, workspace_id=workspace_id)
-        by_suggestion, by_recipe = self._installed_workflows(workspace_id=workspace_id)
-
-        items: List[Dict[str, Any]] = []
-        suppressed_low_confidence = 0
-        suppressed_duplicates = 0
-        seen_recipe_ids: set = set()
-
-        # Patterns arrive sorted by (count, last_asked) desc, so the first
-        # suggestion per recipe is the strongest — later ones targeting the
-        # same recipe would install the identical workflow and are suppressed.
-        for pattern in pattern_report["patterns"]:
-            recipe_id = pattern["recipe_id"]
-            if recipe_id and recipe_id in seen_recipe_ids:
-                suppressed_duplicates += 1
-                continue
-            kg_related = self._kg_related_count(
-                pattern["representative"], workspace_id=workspace_id
-            )
-            confidence, factors = _question_confidence(
-                pattern["count"], pattern["examples"], recipe_id, kg_related
-            )
-            if confidence < _MIN_SUGGESTION_CONFIDENCE:
-                # Asked a couple of times, phrased one way, matching no recipe,
-                # and the Brain knows nothing about it — not enough to propose
-                # automating. Counted, so the surface can say how many were held
-                # back rather than silently showing fewer.
-                suppressed_low_confidence += 1
-                continue
-            if recipe_id:
-                seen_recipe_ids.add(recipe_id)
-            suggestion_id = _stable_id("sug-q", pattern["id"])
-            workflow = by_suggestion.get(suggestion_id) or (
-                by_recipe.get(str(recipe_id)) if recipe_id else None
-            )
-            items.append({
-                "id": suggestion_id,
-                "kind": "recurring_question",
-                "intent": pattern["intent"],
-                "recipe_id": recipe_id,
-                "title": pattern["representative"],
-                "reason": {
-                    "type": "repeated_question",
-                    "count": pattern["count"],
-                    "last_asked": pattern["last_asked"],
-                    "examples": pattern["examples"],
-                },
-                "cadence": "daily",
-                "confidence": confidence,
-                "confidence_factors": factors,
-                "low_confidence": confidence < _LOW_CONFIDENCE_THRESHOLD,
-                "installed": workflow is not None,
-                "workflow_id": (workflow or {}).get("id"),
-            })
-
-        for source in self._knowledge_sources():
-            file_status = source.get("file_status") or {}
-            indexed = sum(int(v or 0) for v in file_status.values())
-            if indexed <= 0:
-                continue
-            confidence, factors = _source_confidence(
-                indexed, bool(source.get("watch_enabled"))
-            )
-            if confidence < _MIN_SUGGESTION_CONFIDENCE:
-                # A barely-indexed folder is not enough evidence for a digest
-                # automation yet — do not suggest.
-                suppressed_low_confidence += 1
-                continue
-            suggestion_id = _stable_id("sug-src", str(source.get("id") or source.get("root_path") or ""))
-            items.append({
-                "id": suggestion_id,
-                "kind": "knowledge_source",
-                "intent": "source_digest",
-                "recipe_id": None,
-                "title": str(source.get("label") or source.get("root_path") or "knowledge folder"),
-                "reason": {
-                    "type": "connected_source",
-                    "root_path": source.get("root_path"),
-                    "indexed_files": indexed,
-                    "watch_enabled": bool(source.get("watch_enabled")),
-                },
-                "cadence": "when new knowledge arrives",
-                "confidence": confidence,
-                "confidence_factors": factors,
-                "low_confidence": confidence < _LOW_CONFIDENCE_THRESHOLD,
-                "installed": suggestion_id in by_suggestion,
-                "workflow_id": (by_suggestion.get(suggestion_id) or {}).get("id"),
-            })
-
-        return {
-            "suggestions": items,
-            "questions_scanned": pattern_report["questions_scanned"],
-            "quality": {
-                "min_confidence": _MIN_SUGGESTION_CONFIDENCE,
-                "low_confidence_threshold": _LOW_CONFIDENCE_THRESHOLD,
-                "suppressed_low_confidence": suppressed_low_confidence,
-                "suppressed_duplicates": suppressed_duplicates,
-            },
-            "consent": {
-                "default_state": "draft_disabled",
-                "local_only": True,
-                "external_actions": False,
-                "requires_user_enable": True,
-                "review_before_run": True,
-            },
-            "generated_at": _now(),
-        }
-
-    def find_suggestion(
-        self,
-        suggestion_id: str,
-        *,
-        user_email: Optional[str] = None,
-        workspace_id: Optional[str] = None,
-    ) -> Optional[Dict[str, Any]]:
-        for suggestion in self.suggestions(user_email=user_email, workspace_id=workspace_id)["suggestions"]:
-            if suggestion["id"] == suggestion_id:
-                return suggestion
-        return None
-
-    # ── workflow building ────────────────────────────────────────────────
-
-    def build_suggestion_workflow(
-        self, suggestion: Dict[str, Any], *, enabled: bool = False
-    ) -> Dict[str, Any]:
-        """Build a consent-first workflow definition from a suggestion.
-
-        Mirrors the starter-recipe shape: trigger → draft agent → review
-        output, installed disabled by default, review-queue gated, local
-        only, and stamped with provenance so installs stay idempotent.
-        """
-        kind = suggestion.get("kind")
-        title = str(suggestion.get("title") or "automation")[:80]
-        if kind == "knowledge_source":
-            trigger: Dict[str, Any] = {"trigger": "brain_event"}
-            trigger_name = "New knowledge in connected folder"
-            name = f"Folder digest: {title}"
-            reason = suggestion.get("reason") or {}
-            prompt = (
-                "New knowledge arrived from the connected folder "
-                f"'{reason.get('root_path') or title}'. Draft a short digest of what "
-                "was added, what changed, and any follow-ups it suggests. Use only "
-                "the local Brain; do not contact external services."
-            )
-            creates = ["folder digest", "change summary", "follow-up suggestions"]
-        else:
-            trigger = {"trigger": "interval", "interval_seconds": 86_400}
-            trigger_name = "User-enabled schedule"
-            name = f"Scheduled answer: {title}"
-            prompt = (
-                "The user repeatedly asks this question:\n"
-                f"“{suggestion.get('title')}”\n"
-                "Answer it from the current Brain (memories, knowledge graph, "
-                "recent activity) as a concise draft the user can review. "
-                "Use only local knowledge; do not contact external services."
-            )
-            creates = ["scheduled answer draft", "supporting evidence", "suggested follow-ups"]
-
-        trigger_config = {
-            **trigger,
-            "enabled": bool(enabled),
-            "review_queue": True,
-            "consent_required": True,
-            "local_only": True,
-            "external_actions": False,
-        }
-        return {
-            "name": name,
-            "nodes": [
-                {
-                    "id": "trigger",
-                    "type": "trigger",
-                    "name": trigger_name,
-                    "config": trigger_config,
-                    "next": "draft",
-                },
-                {
-                    "id": "draft",
-                    "type": "agent",
-                    "name": "Draft for review",
-                    "config": {
-                        "agent": "agent:planner",
-                        "goal": prompt,
-                        "prompt": prompt,
-                        "roles": ["researcher", "planner", "executor", "reviewer"],
-                        "mode": "draft",
-                        "local_only": True,
-                        "external_actions": False,
-                        "requires_review": True,
-                    },
-                    "next": "output",
-                },
-                {
-                    "id": "output",
-                    "type": "output",
-                    "name": "Review before saving",
-                    "config": {
-                        "value": "Draft ready for review. Save, edit, or discard it before it becomes durable memory.",
-                    },
-                    "next": None,
-                },
-            ],
-            "metadata": {
-                "created_from": "automation_suggestion",
-                "suggestion_id": suggestion.get("id"),
-                "suggestion_kind": kind,
-                "suggestion_title": title,
-                "suggestion_reason": suggestion.get("reason"),
-                "suggestion_confidence": suggestion.get("confidence"),
-                "automation_state": "enabled" if enabled else "draft_disabled",
-                "local_only": True,
-                "external_actions": False,
-                "requires_user_enable": not enabled,
-                "creates": creates,
-            },
-        }
-
-    # ── overview for the automation surface ──────────────────────────────
-
-    def overview(
-        self, *, user_email: Optional[str] = None, workspace_id: Optional[str] = None
-    ) -> Dict[str, Any]:
-        suggestion_report = self.suggestions(user_email=user_email, workspace_id=workspace_id)
-        installed: List[Dict[str, Any]] = []
-        if self._store is not None:
-            try:
-                workflows = self._store.list_workflows(workspace_id=workspace_id).get("workflows") or []
-            except Exception:
-                LOGGER.exception("automation intelligence workflow read failed")
-                workflows = []
-            from latticeai.services.automation_execution import last_execution_view
-
-            for workflow in workflows:
-                metadata = (workflow or {}).get("metadata") or {}
-                if metadata.get("created_from") not in {"automation_suggestion", "brain_automation_recipe"}:
-                    continue
-                installed.append({
-                    "id": workflow.get("id"),
-                    "name": workflow.get("name"),
-                    "created_from": metadata.get("created_from"),
-                    "suggestion_id": metadata.get("suggestion_id"),
-                    "recipe_id": metadata.get("recipe_id"),
-                    "enabled": metadata.get("automation_state") == "enabled",
-                    "requires_user_enable": bool(metadata.get("requires_user_enable", True)),
-                    "creates": metadata.get("creates") or [],
-                    # Execution log surfacing (backlog #6): the newest of the
-                    # run-now stamp vs. the latest persisted workflow run.
-                    "last_execution": last_execution_view(
-                        workflow, store=self._store, workspace_id=workspace_id
-                    ),
-                })
-        return {
-            "suggestions": suggestion_report["suggestions"],
-            "questions_scanned": suggestion_report["questions_scanned"],
-            "installed": installed,
-            "quality": suggestion_report["quality"],
-            "consent": suggestion_report["consent"],
-            "generated_at": _now(),
-        }
-
-
-__all__ = ["AutomationIntelligenceService"]