nexo-brain 7.30.22 → 7.30.23

package/src/opportunity_orchestrator.py

@@ -0,0 +1,933 @@
+from __future__ import annotations
+
+"""Opportunity Orchestrator MVP.
+
+This module builds a sparse, evidence-backed proactive queue. It is deliberately
+not a psychological profile layer: it never ranks proposals from mood labels,
+sentiment labels, or personality inferences, and Phase 1 never performs external
+actions.
+"""
+
+import datetime as _dt
+import hashlib
+import json
+import re
+from pathlib import Path
+from typing import Any
+
+
+NORMAL_PROPOSAL_LIMIT = 3
+DEFAULT_PROPOSAL_THRESHOLD = 2.45
+SAFE_ACTION_CLASSES = {"read_only", "prepare_artifact", "local_reversible"}
+VALID_FEEDBACK = {
+    "accepted",
+    "ignored",
+    "snoozed",
+    "dismissed",
+    "false_positive",
+    "useful_but_later",
+}
+FORBIDDEN_VISIBLE_TERMS = {
+    "anxious",
+    "depressed",
+    "vulnerable",
+    "unstable",
+    "burnout",
+    "manipulable",
+    "compliance_likely",
+    "frustrated",
+    "mood",
+    "tension",
+}
+
+
+def _now_iso() -> str:
+    return _dt.datetime.now(_dt.timezone.utc).replace(microsecond=0).isoformat().replace("+00:00", "Z")
+
+
+def _today() -> str:
+    return _dt.datetime.now(_dt.timezone.utc).strftime("%Y-%m-%d")
+
+
+def _expires(days: int = 14) -> str:
+    return (
+        _dt.datetime.now(_dt.timezone.utc)
+        + _dt.timedelta(days=max(1, int(days or 14)))
+    ).replace(microsecond=0).isoformat().replace("+00:00", "Z")
+
+
+def _hash_id(prefix: str, value: str, length: int = 20) -> str:
+    digest = hashlib.sha256(str(value).encode("utf-8", errors="ignore")).hexdigest()[:length]
+    return f"{prefix}-{digest}"
+
+
+def _safe_json(payload: Any) -> str:
+    try:
+        return json.dumps(payload, ensure_ascii=False, sort_keys=True)
+    except Exception:
+        return "{}"
+
+
+def _parse_json(value: str | None, default: Any) -> Any:
+    try:
+        parsed = json.loads(value or "")
+        return parsed if parsed is not None else default
+    except Exception:
+        return default
+
+
+def _row(row: Any) -> dict[str, Any]:
+    if row is None:
+        return {}
+    if hasattr(row, "keys"):
+        return {key: row[key] for key in row.keys()}
+    return dict(row)
+
+
+def _parse_time(value: Any) -> _dt.datetime | None:
+    if value in ("", None):
+        return None
+    if isinstance(value, (int, float)):
+        try:
+            return _dt.datetime.fromtimestamp(float(value), tz=_dt.timezone.utc)
+        except Exception:
+            return None
+    text = str(value).strip()
+    if not text:
+        return None
+    for candidate in (text, text.replace("Z", "+00:00")):
+        try:
+            parsed = _dt.datetime.fromisoformat(candidate)
+            if parsed.tzinfo is None:
+                parsed = parsed.replace(tzinfo=_dt.timezone.utc)
+            return parsed
+        except Exception:
+            pass
+    for fmt in ("%Y-%m-%d %H:%M:%S", "%Y-%m-%d"):
+        try:
+            parsed = _dt.datetime.strptime(text[:19] if fmt != "%Y-%m-%d" else text[:10], fmt)
+            return parsed.replace(tzinfo=_dt.timezone.utc)
+        except Exception:
+            continue
+    return None
+
+
+def _expired(value: Any) -> bool:
+    parsed = _parse_time(value)
+    return bool(parsed and parsed < _dt.datetime.now(_dt.timezone.utc))
+
+
+def _clamp(value: Any, default: float = 0.0) -> float:
+    try:
+        raw = float(value)
+    except Exception:
+        raw = default
+    return max(0.0, min(1.0, raw))
+
+
+def _sanitize_text(value: Any, limit: int = 360) -> str:
+    text = re.sub(r"\s+", " ", str(value or "")).strip()
+    for term in sorted(FORBIDDEN_VISIBLE_TERMS, key=len, reverse=True):
+        text = re.sub(rf"\b{re.escape(term)}\b", "operational signal", text, flags=re.IGNORECASE)
+    if len(text) > limit:
+        text = text[: max(0, limit - 1)].rstrip() + "..."
+    return text
+
+
+def _source_hash(payload: Any) -> str:
+    return hashlib.sha256(_safe_json(payload).encode("utf-8", errors="ignore")).hexdigest()
+
+
+def _score(payload: dict[str, Any]) -> float:
+    score = (
+        float(payload.get("impact") or 0)
+        + float(payload.get("urgency") or 0)
+        + float(payload.get("confidence") or 0)
+        + float(payload.get("readiness") or 0)
+        + float(payload.get("user_burden_reduction") or 0)
+        + float(payload.get("strategic_alignment") or 0)
+        - float(payload.get("risk") or 0)
+        - float(payload.get("interruption_cost") or 0)
+        - float(payload.get("repetition_penalty") or 0)
+    )
+    return round(max(0.0, score), 4)
+
+
+def _table_exists(conn, table: str) -> bool:
+    row = conn.execute(
+        "SELECT 1 FROM sqlite_master WHERE type='table' AND name=? LIMIT 1",
+        (table,),
+    ).fetchone()
+    return row is not None
+
+
+def _opportunity_type_from_closure(item: dict[str, Any]) -> str:
+    source = str(item.get("source_primary") or "").lower()
+    kind = str(item.get("kind") or "").lower()
+    if "debt" in source or "blocked" in kind:
+        return "remediation"
+    if "outcome" in source:
+        return "closure"
+    if "followup_due" in kind:
+        return "deadline"
+    if "mcp_write_queue" in source:
+        return "remediation"
+    if "release" in (str(item.get("title") or "") + " " + str(item.get("summary") or "")).lower():
+        return "product"
+    return "closure"
+
+
+def _domain_from_item(item: dict[str, Any]) -> str:
+    text = (str(item.get("title") or "") + " " + str(item.get("summary") or "")).lower()
+    if "desktop" in text:
+        return "desktop"
+    if "release" in text or "publish" in text:
+        return "release"
+    if "email" in text or "correo" in text:
+        return "email"
+    if "ads" in text or "google" in text:
+        return "ads"
+    if "nexo" in text or "brain" in text:
+        return "brain"
+    return "general"
+
+
+def _authorization_status(action_class: str) -> str:
+    clean = str(action_class or "read_only").strip() or "read_only"
+    return "not_required" if clean in SAFE_ACTION_CLASSES else "needs_permission"
+
+
+def _candidate_from_closure(item: dict[str, Any]) -> dict[str, Any]:
+    source_id = str(item.get("id") or item.get("dedupe_key") or "")
+    source_type = "closure_items"
+    title_source = _sanitize_text(item.get("title") or "Operational opportunity", 160)
+    opportunity_type = _opportunity_type_from_closure(item)
+    domain = _domain_from_item(item)
+    impact = _clamp(item.get("impact_score"), 0.55)
+    urgency = _clamp(item.get("urgency_score"), 0.45)
+    confidence = max(0.45, _clamp(item.get("confidence_score"), 0.75))
+    risk = _clamp(item.get("risk_score"), 0.15)
+    readiness = 0.72 if str(item.get("evidence_required") or "").strip() else 0.58
+    burden = 0.68 if str(item.get("next_action") or "").strip() else 0.52
+    strategic = 0.6 if domain in {"release", "desktop", "brain"} else 0.42
+    interruption_cost = 0.32
+    action_class = "read_only"
+    dedupe_key = f"{source_type}:{source_id}"
+    opportunity_id = _hash_id("OPP", dedupe_key)
+    now = _now_iso()
+    source_payload = {
+        "source_primary": item.get("source_primary"),
+        "kind": item.get("kind"),
+        "state": item.get("state"),
+        "deadline_at": item.get("deadline_at"),
+    }
+    score_payload = {
+        "impact": impact,
+        "urgency": urgency,
+        "confidence": confidence,
+        "readiness": readiness,
+        "user_burden_reduction": burden,
+        "strategic_alignment": strategic,
+        "risk": risk,
+        "interruption_cost": interruption_cost,
+        "repetition_penalty": 0.0,
+    }
+    opportunity = {
+        "id": opportunity_id,
+        "title": f"Prepared review: {title_source}",
+        "hypothesis": "This open operational item has enough evidence for a read-only preparation.",
+        "domain": domain,
+        "opportunity_type": opportunity_type,
+        "dedupe_key": dedupe_key,
+        "impact": impact,
+        "urgency": urgency,
+        "confidence": confidence,
+        "risk": risk,
+        "effort": 0.25,
+        "readiness": readiness,
+        "user_burden_reduction": burden,
+        "interruption_cost": interruption_cost,
+        "strategic_alignment": strategic,
+        "repetition_penalty": 0.0,
+        "score": _score(score_payload),
+        "state": "candidate",
+        "owner": "nero",
+        "why_now": _sanitize_text(
+            item.get("blocker_reason")
+            or item.get("evidence_required")
+            or "The source remains open and can be reduced to a short review.",
+            360,
+        ),
+        "next_action": _sanitize_text(
+            item.get("next_action")
+            or "Inspect evidence and choose accept, snooze, or suppress.",
+            360,
+        ),
+        "action_class": action_class,
+        "authorization_status": _authorization_status(action_class),
+        "created_at": now,
+        "updated_at": now,
+        "expires_at": item.get("deadline_at") or _expires(14),
+        "last_proposed_at": "",
+        "source_payload_json": _safe_json(source_payload),
+    }
+    signal = {
+        "id": _hash_id("SIG", f"{dedupe_key}:signal"),
+        "source_type": source_type,
+        "source_id": source_id,
+        "entity_ref": _sanitize_text(domain, 80),
+        "summary": _sanitize_text(item.get("title") or item.get("summary") or source_id, 280),
+        "signal_kind": _sanitize_text(item.get("kind") or opportunity_type, 80),
+        "urgency": urgency,
+        "confidence": confidence,
+        "privacy_level": "normal",
+        "source_hash": _source_hash(source_payload),
+        "created_at": now,
+        "expires_at": opportunity["expires_at"],
+    }
+    evidence = {
+        "id": _hash_id("OPE", f"{opportunity_id}:{source_type}:{source_id}"),
+        "opportunity_id": opportunity_id,
+        "source_type": source_type,
+        "source_id": source_id,
+        "evidence_summary": _sanitize_text(item.get("summary") or item.get("evidence_required") or item.get("title"), 360),
+        "confidence": confidence,
+        "created_at": now,
+    }
+    preparation = {
+        "id": _hash_id("PREP", f"{opportunity_id}:decision_card"),
+        "opportunity_id": opportunity_id,
+        "artifact_type": "decision_card",
+        "artifact_ref": f"nexo://opportunity/{opportunity_id}",
+        "safe_mode": 1,
+        "approval_required": 0,
+        "status": "ready",
+        "created_at": now,
+        "expires_at": opportunity["expires_at"],
+    }
+    return {
+        "signal": signal,
+        "opportunity": opportunity,
+        "evidence": [evidence],
+        "preparations": [preparation],
+    }
+
+
+def _collect_from_closure(conn, limit_per_source: int = 250) -> list[dict[str, Any]]:
+    from closure_plane import closure_next, refresh_closure_items
+
+    refresh_closure_items(conn, limit_per_adapter=limit_per_source)
+    items = closure_next(conn, limit=limit_per_source, include_waiting=True)
+    return [_candidate_from_closure(item) for item in items]
+
+
+def _filtered_sources(sources: str) -> set[str]:
+    return {part.strip().lower() for part in str(sources or "").split(",") if part.strip()}
+
+
+def collect_candidates(conn=None, *, sources: str = "", limit_per_source: int = 250) -> list[dict[str, Any]]:
+    if conn is None:
+        from db import get_db
+        from db._schema import run_migrations
+
+        conn = get_db()
+        run_migrations(conn)
+    selected = _filtered_sources(sources)
+    candidates: list[dict[str, Any]] = []
+    if not selected or "closure" in selected or "closure_items" in selected:
+        candidates.extend(_collect_from_closure(conn, limit_per_source=limit_per_source))
+    return candidates
+
+
+def _upsert_signal(conn, signal: dict[str, Any]) -> None:
+    conn.execute(
+        """
+        INSERT INTO nexo_signals (
+            id, source_type, source_id, entity_ref, summary, signal_kind,
+            urgency, confidence, privacy_level, source_hash, created_at, expires_at
+        ) VALUES (
+            :id, :source_type, :source_id, :entity_ref, :summary, :signal_kind,
+            :urgency, :confidence, :privacy_level, :source_hash, :created_at, :expires_at
+        )
+        ON CONFLICT(source_type, source_id, signal_kind) DO UPDATE SET
+            entity_ref = excluded.entity_ref,
+            summary = excluded.summary,
+            urgency = excluded.urgency,
+            confidence = excluded.confidence,
+            privacy_level = excluded.privacy_level,
+            source_hash = excluded.source_hash,
+            expires_at = excluded.expires_at
+        """,
+        signal,
+    )
+
+
+def _upsert_opportunity(conn, opportunity: dict[str, Any]) -> None:
+    conn.execute(
+        """
+        INSERT INTO nexo_opportunities (
+            id, title, hypothesis, domain, opportunity_type, dedupe_key,
+            impact, urgency, confidence, risk, effort, readiness,
+            user_burden_reduction, interruption_cost, strategic_alignment,
+            repetition_penalty, score, state, owner, why_now, next_action,
+            action_class, authorization_status, created_at, updated_at,
+            expires_at, last_proposed_at, source_payload_json
+        ) VALUES (
+            :id, :title, :hypothesis, :domain, :opportunity_type, :dedupe_key,
+            :impact, :urgency, :confidence, :risk, :effort, :readiness,
+            :user_burden_reduction, :interruption_cost, :strategic_alignment,
+            :repetition_penalty, :score, :state, :owner, :why_now, :next_action,
+            :action_class, :authorization_status, :created_at, :updated_at,
+            :expires_at, :last_proposed_at, :source_payload_json
+        )
+        ON CONFLICT(dedupe_key) DO UPDATE SET
+            title = excluded.title,
+            hypothesis = excluded.hypothesis,
+            domain = excluded.domain,
+            opportunity_type = excluded.opportunity_type,
+            impact = excluded.impact,
+            urgency = excluded.urgency,
+            confidence = excluded.confidence,
+            risk = excluded.risk,
+            effort = excluded.effort,
+            readiness = excluded.readiness,
+            user_burden_reduction = excluded.user_burden_reduction,
+            interruption_cost = excluded.interruption_cost,
+            strategic_alignment = excluded.strategic_alignment,
+            score = excluded.score,
+            why_now = excluded.why_now,
+            next_action = excluded.next_action,
+            action_class = excluded.action_class,
+            authorization_status = excluded.authorization_status,
+            updated_at = excluded.updated_at,
+            expires_at = excluded.expires_at,
+            source_payload_json = excluded.source_payload_json
+        WHERE nexo_opportunities.state NOT IN ('closed', 'discarded', 'suppressed', 'stale')
+        """,
+        opportunity,
+    )
+
+
+def _upsert_evidence(conn, evidence: dict[str, Any]) -> None:
+    conn.execute(
+        """
+        INSERT INTO nexo_opportunity_evidence (
+            id, opportunity_id, source_type, source_id, evidence_summary, confidence, created_at
+        ) VALUES (
+            :id, :opportunity_id, :source_type, :source_id, :evidence_summary, :confidence, :created_at
+        )
+        ON CONFLICT(opportunity_id, source_type, source_id) DO UPDATE SET
+            evidence_summary = excluded.evidence_summary,
+            confidence = excluded.confidence
+        """,
+        evidence,
+    )
+
+
+def _upsert_preparation(conn, preparation: dict[str, Any]) -> None:
+    conn.execute(
+        """
+        INSERT INTO nexo_preparations (
+            id, opportunity_id, artifact_type, artifact_ref, safe_mode,
+            approval_required, status, created_at, expires_at
+        ) VALUES (
+            :id, :opportunity_id, :artifact_type, :artifact_ref, :safe_mode,
+            :approval_required, :status, :created_at, :expires_at
+        )
+        ON CONFLICT(opportunity_id, artifact_type, artifact_ref) DO UPDATE SET
+            safe_mode = excluded.safe_mode,
+            approval_required = excluded.approval_required,
+            status = excluded.status,
+            expires_at = excluded.expires_at
+        """,
+        preparation,
+    )
+
+
+def _persist_candidates(conn, candidates: list[dict[str, Any]]) -> dict[str, int]:
+    counts = {"signals": 0, "opportunities": 0, "evidence": 0, "preparations": 0}
+    for candidate in candidates:
+        _upsert_signal(conn, candidate["signal"])
+        counts["signals"] += 1
+        _upsert_opportunity(conn, candidate["opportunity"])
+        counts["opportunities"] += 1
+        for evidence in candidate.get("evidence") or []:
+            _upsert_evidence(conn, evidence)
+            counts["evidence"] += 1
+        for preparation in candidate.get("preparations") or []:
+            _upsert_preparation(conn, preparation)
+            counts["preparations"] += 1
+    conn.commit()
+    return counts
+
+
+def _active_suppression(conn, scope_type: str, scope_key: str) -> dict[str, Any] | None:
+    if not scope_key:
+        return None
+    rows = conn.execute(
+        """
+        SELECT *
+        FROM nexo_suppression_rules
+        WHERE scope_type = ? AND scope_key = ?
+        ORDER BY created_at DESC
+        """,
+        (scope_type, scope_key),
+    ).fetchall()
+    for raw in rows:
+        row = _row(raw)
+        if not _expired(row.get("expires_at")):
+            return row
+    return None
+
+
+def _opportunity_evidence(conn, opportunity_id: str) -> list[dict[str, Any]]:
+    rows = conn.execute(
+        """
+        SELECT *
+        FROM nexo_opportunity_evidence
+        WHERE opportunity_id = ?
+        ORDER BY confidence DESC, created_at DESC
+        """,
+        (opportunity_id,),
+    ).fetchall()
+    return [_row(row) for row in rows]
+
+
+def _opportunity_preparations(conn, opportunity_id: str) -> list[dict[str, Any]]:
+    rows = conn.execute(
+        """
+        SELECT *
+        FROM nexo_preparations
+        WHERE opportunity_id = ? AND status NOT IN ('stale', 'deleted')
+        ORDER BY created_at DESC
+        """,
+        (opportunity_id,),
+    ).fetchall()
+    return [_row(row) for row in rows]
+
+
+def _proposal_copy(opportunity: dict[str, Any], evidence: list[dict[str, Any]]) -> str:
+    evidence_count = len(evidence)
+    return _sanitize_text(
+        f"{opportunity.get('title')}. Evidence refs: {evidence_count}. "
+        f"Why now: {opportunity.get('why_now')}",
+        520,
+    )
+
+
+def _create_or_update_proposal(conn, opportunity: dict[str, Any], surface: str, evidence: list[dict[str, Any]]) -> dict[str, Any]:
+    now = _now_iso()
+    proposal_id = _hash_id("PROP", f"{opportunity['id']}:{surface}")
+    payload = {
+        "id": proposal_id,
+        "opportunity_id": opportunity["id"],
+        "surface": surface,
+        "copy": _proposal_copy(opportunity, evidence),
+        "cta_primary": "Inspect evidence",
+        "cta_secondary": "Snooze",
+        "shown_at": "",
+        "feedback": "",
+        "created_at": now,
+    }
+    conn.execute(
+        """
+        INSERT INTO nexo_proposals (
+            id, opportunity_id, surface, copy, cta_primary, cta_secondary,
+            shown_at, feedback, created_at
+        ) VALUES (
+            :id, :opportunity_id, :surface, :copy, :cta_primary, :cta_secondary,
+            :shown_at, :feedback, :created_at
+        )
+        ON CONFLICT(opportunity_id, surface) DO UPDATE SET
+            copy = excluded.copy,
+            cta_primary = excluded.cta_primary,
+            cta_secondary = excluded.cta_secondary
+        """,
+        payload,
+    )
+    conn.execute(
+        "UPDATE nexo_opportunities SET state = 'proposed', last_proposed_at = ?, updated_at = ? WHERE id = ?",
+        (now, now, opportunity["id"]),
+    )
+    conn.commit()
+    row = conn.execute("SELECT * FROM nexo_proposals WHERE id = ?", (proposal_id,)).fetchone()
+    return _row(row)
+
+
+def _proposal_payload(conn, proposal: dict[str, Any]) -> dict[str, Any]:
+    opportunity = get_opportunity(proposal.get("opportunity_id") or "", include_evidence=True, conn=conn)
+    item = {
+        "id": proposal.get("id"),
+        "surface": proposal.get("surface") or "",
+        "copy": _sanitize_text(proposal.get("copy"), 600),
+        "cta_primary": proposal.get("cta_primary") or "",
+        "cta_secondary": proposal.get("cta_secondary") or "",
+        "feedback": proposal.get("feedback") or "",
+        "created_at": proposal.get("created_at") or "",
+        "opportunity": opportunity.get("opportunity"),
+    }
+    if item["opportunity"]:
+        item["confidence"] = item["opportunity"].get("confidence")
+        item["evidence_refs"] = [
+            f"{ev.get('source_type')}:{ev.get('source_id')}"
+            for ev in item["opportunity"].get("evidence", [])
+        ]
+    else:
+        item["confidence"] = 0
+        item["evidence_refs"] = []
+    return item
+
+
+def _eligible_opportunities(conn, *, include_snoozed: bool = False) -> list[dict[str, Any]]:
+    rows = conn.execute(
+        """
+        SELECT *
+        FROM nexo_opportunities
+        WHERE state IN ('candidate', 'prepared', 'proposed')
+          AND score >= ?
+          AND authorization_status IN ('not_required', 'needs_permission')
+        ORDER BY score DESC, urgency DESC, updated_at DESC
+        LIMIT 100
+        """,
+        (DEFAULT_PROPOSAL_THRESHOLD,),
+    ).fetchall()
+    results: list[dict[str, Any]] = []
+    for raw in rows:
+        opportunity = _row(raw)
+        if _expired(opportunity.get("expires_at")):
+            continue
+        if _active_suppression(conn, "opportunity", opportunity.get("id") or ""):
+            continue
+        if _active_suppression(conn, "domain", opportunity.get("domain") or ""):
+            continue
+        if _active_suppression(conn, "type", opportunity.get("opportunity_type") or ""):
+            continue
+        if not include_snoozed and _active_suppression(conn, "snooze", opportunity.get("id") or ""):
+            continue
+        evidence = _opportunity_evidence(conn, opportunity["id"])
+        if not evidence:
+            continue
+        opportunity["evidence"] = evidence
+        results.append(opportunity)
+    return results
+
+
+def opportunity_queue(
+    conn=None,
+    *,
+    surface: str = "home",
+    limit: int = NORMAL_PROPOSAL_LIMIT,
+    refresh: bool = False,
+    include_snoozed: bool = False,
+) -> dict[str, Any]:
+    if conn is None:
+        from db import get_db
+        from db._schema import run_migrations
+
+        conn = get_db()
+        run_migrations(conn)
+    if refresh:
+        refresh_opportunities(conn, dry_run=False)
+    clean_surface = _sanitize_text(surface or "home", 80) or "home"
+    clean_limit = min(NORMAL_PROPOSAL_LIMIT, max(0, int(limit or NORMAL_PROPOSAL_LIMIT)))
+    if clean_limit <= 0:
+        return {"ok": True, "schema": "nexo.opportunity.queue.v1", "surface": clean_surface, "proposals": []}
+    selected = _eligible_opportunities(conn, include_snoozed=include_snoozed)[:clean_limit]
+    proposals = [
+        _proposal_payload(conn, _create_or_update_proposal(conn, opportunity, clean_surface, opportunity["evidence"]))
+        for opportunity in selected
+    ]
+    return {
+        "ok": True,
+        "schema": "nexo.opportunity.queue.v1",
+        "surface": clean_surface,
+        "proposal_limit": NORMAL_PROPOSAL_LIMIT,
+        "proposals": proposals,
+        "zero_proposals_ok": len(proposals) == 0,
+    }
+
+
+def refresh_opportunities(
+    conn=None,
+    *,
+    dry_run: bool = True,
+    sources: str = "",
+    limit_per_source: int = 250,
+    write_report: bool = False,
+) -> dict[str, Any]:
+    if conn is None:
+        from db import get_db
+        from db._schema import run_migrations
+
+        conn = get_db()
+        run_migrations(conn)
+    clean_limit = max(1, min(int(limit_per_source or 250), 500))
+    candidates = collect_candidates(conn, sources=sources, limit_per_source=clean_limit)
+    persisted = {"signals": 0, "opportunities": 0, "evidence": 0, "preparations": 0}
+    if not dry_run:
+        persisted = _persist_candidates(conn, candidates)
+    candidate_summary = [
+        {
+            "opportunity_id": item["opportunity"]["id"],
+            "title": item["opportunity"]["title"],
+            "score": item["opportunity"]["score"],
+            "evidence_refs": [
+                f"{ev.get('source_type')}:{ev.get('source_id')}"
+                for ev in item.get("evidence", [])
+            ],
+            "selected": item["opportunity"]["score"] >= DEFAULT_PROPOSAL_THRESHOLD,
+        }
+        for item in candidates
+    ]
+    result = {
+        "ok": True,
+        "schema": "nexo.opportunity.refresh.v1",
+        "dry_run": bool(dry_run),
+        "sources": sorted(_filtered_sources(sources)) if _filtered_sources(sources) else ["closure_items"],
+        "observed_candidates": len(candidates),
+        "persisted": persisted,
+        "candidates": candidate_summary,
+        "zero_proposals_ok": True,
+    }
+    if write_report:
+        result["report"] = write_daily_report(conn, refresh_result=result)
+    return result
+
+
+def get_opportunity(opportunity_id: str, *, include_evidence: bool = True, conn=None) -> dict[str, Any]:
+    if conn is None:
+        from db import get_db
+        from db._schema import run_migrations
+
+        conn = get_db()
+        run_migrations(conn)
+    clean_id = str(opportunity_id or "").strip()
+    if not clean_id:
+        return {"ok": False, "error": "opportunity_id is required"}
+    row = conn.execute(
+        "SELECT * FROM nexo_opportunities WHERE id = ? OR dedupe_key = ?",
+        (clean_id, clean_id),
+    ).fetchone()
+    if not row:
+        return {"ok": False, "error": "opportunity not found"}
+    opportunity = _row(row)
+    opportunity["source_payload"] = _parse_json(opportunity.pop("source_payload_json", "{}"), {})
+    if include_evidence:
+        opportunity["evidence"] = _opportunity_evidence(conn, opportunity["id"])
+        opportunity["preparations"] = _opportunity_preparations(conn, opportunity["id"])
+    return {"ok": True, "opportunity": opportunity}
+
+
+def opportunity_feedback(
+    proposal_id: str,
+    feedback: str,
+    *,
+    note: str = "",
+    snooze_until: str = "",
+    conn=None,
+) -> dict[str, Any]:
+    if conn is None:
+        from db import get_db
+        from db._schema import run_migrations
+
+        conn = get_db()
+        run_migrations(conn)
+    clean_feedback = str(feedback or "").strip()
+    if clean_feedback not in VALID_FEEDBACK:
+        return {"ok": False, "error": f"invalid feedback: {clean_feedback}"}
+    proposal = _row(conn.execute("SELECT * FROM nexo_proposals WHERE id = ?", (str(proposal_id or ""),)).fetchone())
+    if not proposal:
+        return {"ok": False, "error": "proposal not found"}
+    now = _now_iso()
+    conn.execute("UPDATE nexo_proposals SET feedback = ? WHERE id = ?", (clean_feedback, proposal["id"]))
+    event_id = _hash_id("OPEV", f"{proposal['id']}:{clean_feedback}:{now}:{note}", 24)
+    conn.execute(
+        """
+        INSERT INTO nexo_proposal_events (
+            id, proposal_id, event_type, feedback, note, metadata_json, created_at
+        ) VALUES (?, ?, 'feedback', ?, ?, ?, ?)
+        """,
+        (
+            event_id,
+            proposal["id"],
+            clean_feedback,
+            _sanitize_text(note, 300),
+            _safe_json({"snooze_until": snooze_until}),
+            now,
+        ),
+    )
+    suppression = None
+    if clean_feedback in {"dismissed", "false_positive", "ignored", "snoozed"}:
+        expires_at = snooze_until if clean_feedback == "snoozed" and snooze_until else _expires(7 if clean_feedback != "false_positive" else 30)
+        suppression = suppress(
+            "snooze" if clean_feedback == "snoozed" else "opportunity",
+            proposal["opportunity_id"],
+            reason=clean_feedback,
+            expires_at=expires_at,
+            conn=conn,
+        )
+    conn.commit()
+    return {
+        "ok": True,
+        "proposal_id": proposal["id"],
+        "feedback": clean_feedback,
+        "event_id": event_id,
+        "suppression": suppression,
+    }
+
+
+def suppress(
+    scope_type: str,
+    scope_key: str,
+    *,
+    reason: str = "",
+    expires_at: str = "",
+    conn=None,
+) -> dict[str, Any]:
+    if conn is None:
+        from db import get_db
+        from db._schema import run_migrations
+
+        conn = get_db()
+        run_migrations(conn)
+    clean_type = _sanitize_text(scope_type, 80)
+    clean_key = _sanitize_text(scope_key, 160)
+    if not clean_type or not clean_key:
+        return {"ok": False, "error": "scope_type and scope_key are required"}
+    clean_reason = _sanitize_text(reason or "manual", 240)
+    clean_expires = str(expires_at or _expires(14)).strip()
+    row_id = _hash_id("OSR", f"{clean_type}:{clean_key}:{clean_reason}", 24)
+    conn.execute(
+        """
+        INSERT INTO nexo_suppression_rules (
+            id, scope_type, scope_key, reason, expires_at, created_at
+        ) VALUES (?, ?, ?, ?, ?, ?)
+        ON CONFLICT(scope_type, scope_key, reason) DO UPDATE SET
+            expires_at = excluded.expires_at
+        """,
+        (row_id, clean_type, clean_key, clean_reason, clean_expires, _now_iso()),
+    )
+    conn.commit()
+    return {
+        "ok": True,
+        "id": row_id,
+        "scope_type": clean_type,
+        "scope_key": clean_key,
+        "reason": clean_reason,
+        "expires_at": clean_expires,
+    }
+
+
+def write_daily_report(conn=None, *, refresh_result: dict[str, Any] | None = None) -> dict[str, Any]:
+    if conn is None:
+        from db import get_db
+        from db._schema import run_migrations
+
+        conn = get_db()
+        run_migrations(conn)
+    from paths import operations_dir
+
+    queue = opportunity_queue(conn, surface="morning_briefing", limit=NORMAL_PROPOSAL_LIMIT, refresh=False)
+    root = operations_dir() / "opportunity-orchestrator"
+    root.mkdir(parents=True, exist_ok=True)
+    json_path = root / f"{_today()}-opportunities.json"
+    md_path = root / f"{_today()}-opportunities.md"
+    payload = {
+        "schema": "nexo.opportunity.report.v1",
+        "generated_at": _now_iso(),
+        "refresh": refresh_result or {},
+        "queue": queue,
+    }
+    json_path.write_text(json.dumps(payload, indent=2, ensure_ascii=False) + "\n", encoding="utf-8")
+    lines = [
+        f"# Opportunity Orchestrator -- {_today()}",
+        "",
+        f"- Generated at: {payload['generated_at']}",
+        f"- Proposals: {len(queue.get('proposals') or [])}",
+        "- Zero proposals is valid when evidence is not strong enough.",
+        "",
+    ]
+    for proposal in queue.get("proposals") or []:
+        opportunity = proposal.get("opportunity") or {}
+        lines.extend([
+            f"## {opportunity.get('title', proposal.get('id'))}",
+            "",
+            f"- Score: {opportunity.get('score')}",
+            f"- Confidence: {opportunity.get('confidence')}",
+            f"- Why now: {opportunity.get('why_now')}",
+            f"- Next action: {opportunity.get('next_action')}",
+            f"- Evidence: {', '.join(proposal.get('evidence_refs') or [])}",
+            "",
+        ])
+    md_path.write_text("\n".join(lines), encoding="utf-8")
+    return {"json": str(json_path), "markdown": str(md_path)}
+
+
+def handle_opportunity_refresh(
+    dry_run: bool = True,
+    sources: str = "",
+    limit_per_source: int = 250,
+    write_report: bool = False,
+) -> str:
+    return json.dumps(
+        refresh_opportunities(
+            dry_run=bool(dry_run),
+            sources=sources,
+            limit_per_source=limit_per_source,
+            write_report=bool(write_report),
+        ),
+        indent=2,
+        ensure_ascii=False,
+    )
+
+
+def handle_opportunity_queue(
+    surface: str = "home",
+    limit: int = NORMAL_PROPOSAL_LIMIT,
+    refresh: bool = False,
+    include_snoozed: bool = False,
+) -> str:
+    return json.dumps(
+        opportunity_queue(
+            surface=surface,
+            limit=limit,
+            refresh=bool(refresh),
+            include_snoozed=bool(include_snoozed),
+        ),
+        indent=2,
+        ensure_ascii=False,
+    )
+
+
+def handle_opportunity_get(opportunity_id: str, include_evidence: bool = True) -> str:
+    return json.dumps(
+        get_opportunity(opportunity_id, include_evidence=bool(include_evidence)),
+        indent=2,
+        ensure_ascii=False,
+    )
+
+
+def handle_opportunity_feedback(
+    proposal_id: str,
+    feedback: str,
+    note: str = "",
+    snooze_until: str = "",
+) -> str:
+    return json.dumps(
+        opportunity_feedback(proposal_id, feedback, note=note, snooze_until=snooze_until),
+        indent=2,
+        ensure_ascii=False,
+    )
+
+
+def handle_opportunity_suppress(
+    scope_type: str,
+    scope_key: str,
+    reason: str = "",
+    expires_at: str = "",
+) -> str:
+    return json.dumps(
+        suppress(scope_type, scope_key, reason=reason, expires_at=expires_at),
+        indent=2,
+        ensure_ascii=False,
+    )