ligbox-ops-platform/projects/ops-desk/api/app/agents/store.py

"""Persistence Agentic Ops."""
from __future__ import annotations
import json, sqlite3, time
from datetime import datetime, timezone
from typing import Any

from app.agents.messages import init_messages_schema

def _now():
    return datetime.now(timezone.utc).isoformat()

def _exec_retry(conn, sql, params=(), *, attempts=8):
    for attempt in range(attempts):
        try:
            return conn.execute(sql, params)
        except sqlite3.OperationalError as exc:
            if "locked" not in str(exc).lower() or attempt >= attempts - 1:
                raise
            time.sleep(0.25 * (attempt + 1))

def init_agent_schema(conn):
    init_messages_schema(conn)
    conn.executescript("""
        CREATE TABLE IF NOT EXISTS agent_scenarios (
            id TEXT PRIMARY KEY, title TEXT NOT NULL, schedule TEXT,
            severity_default TEXT NOT NULL DEFAULT 'warn', config_json TEXT NOT NULL,
            enabled INTEGER NOT NULL DEFAULT 1, updated_at TEXT NOT NULL);
        CREATE TABLE IF NOT EXISTS agent_runs (
            id INTEGER PRIMARY KEY, scenario_id TEXT NOT NULL, trigger TEXT NOT NULL DEFAULT 'cron',
            status TEXT NOT NULL, summary_text TEXT, llm_model TEXT, metadata_json TEXT,
            started_at TEXT NOT NULL, finished_at TEXT);
        CREATE TABLE IF NOT EXISTS agent_findings (
            id INTEGER PRIMARY KEY, run_id INTEGER NOT NULL, severity TEXT NOT NULL,
            category TEXT NOT NULL DEFAULT 'api', title TEXT NOT NULL, detail_md TEXT,
            evidence_json TEXT, suggested_human_action TEXT, kb_refs_json TEXT,
            acknowledged_at TEXT, acknowledged_by TEXT, created_at TEXT NOT NULL);
        CREATE TABLE IF NOT EXISTS agent_action_log (
            id INTEGER PRIMARY KEY, ts TEXT NOT NULL, agent_id TEXT NOT NULL DEFAULT 'sentinel',
            run_id INTEGER, event_type TEXT NOT NULL, message TEXT, payload_json TEXT);
        CREATE TABLE IF NOT EXISTS agent_kb_chunks (
            id INTEGER PRIMARY KEY, source_path TEXT NOT NULL, chunk_text TEXT NOT NULL, indexed_at TEXT NOT NULL);
        CREATE INDEX IF NOT EXISTS idx_agent_runs_scenario ON agent_runs(scenario_id);
        CREATE TABLE IF NOT EXISTS agent_incidents (
            id INTEGER PRIMARY KEY,
            scenario_id TEXT NOT NULL UNIQUE,
            primary_agent TEXT NOT NULL,
            severity TEXT NOT NULL,
            status TEXT NOT NULL DEFAULT 'open',
            title TEXT NOT NULL,
            latest_finding_id INTEGER,
            occurrence_count INTEGER NOT NULL DEFAULT 1,
            first_seen_at TEXT NOT NULL,
            last_seen_at TEXT NOT NULL,
            suggested_human_action TEXT,
            thread_id INTEGER,
            acknowledged_at TEXT,
            acknowledged_by TEXT);
        CREATE INDEX IF NOT EXISTS idx_agent_incidents_status ON agent_incidents(status);
    """)
    try:
        if conn.execute("SELECT COUNT(*) FROM agent_incidents").fetchone()[0] == 0:
            backfill_incidents(conn)
    except sqlite3.OperationalError:
        pass

def log_event(conn, *, event_type, message, agent_id="sentinel", run_id=None, payload=None):
    conn.execute("INSERT INTO agent_action_log (ts,agent_id,run_id,event_type,message,payload_json) VALUES (?,?,?,?,?,?)",
        (_now(), agent_id, run_id, event_type, message, json.dumps(payload or {})))
    try:
        conn.execute("INSERT INTO desk_security_audit (username,event_type,client_ip,payload,created_at) VALUES (?,?,?,?,?)",
            ("agentic", f"agent.{event_type}", "vm122", json.dumps({"message": message, **(payload or {})}), _now()))
    except sqlite3.OperationalError:
        pass

def upsert_scenario(conn, scenario):
    conn.execute("""INSERT INTO agent_scenarios (id,title,schedule,severity_default,config_json,enabled,updated_at)
        VALUES (?,?,?,?,?,1,?) ON CONFLICT(id) DO UPDATE SET title=excluded.title, schedule=excluded.schedule,
        severity_default=excluded.severity_default, config_json=excluded.config_json, updated_at=excluded.updated_at""",
        (scenario["id"], scenario["title"], scenario.get("schedule","*/5 * * * *"),
         scenario.get("severity_default","warn"), json.dumps(scenario), _now()))

def list_scenarios(conn):
    out = []
    for row in conn.execute("SELECT * FROM agent_scenarios WHERE enabled=1 ORDER BY id"):
        item = dict(row)
        item["config"] = json.loads(item.pop("config_json") or "{}")
        last = conn.execute("SELECT status,started_at FROM agent_runs WHERE scenario_id=? ORDER BY id DESC LIMIT 1", (row["id"],)).fetchone()
        item["last_run_status"] = last["status"] if last else None
        item["last_run_at"] = last["started_at"] if last else None
        out.append(item)
    return out

def get_scenario(conn, scenario_id):
    row = conn.execute("SELECT * FROM agent_scenarios WHERE id=? AND enabled=1", (scenario_id,)).fetchone()
    if not row: return None
    item = dict(row); item["config"] = json.loads(item.pop("config_json") or "{}"); return item

def create_run(conn, scenario_id, trigger):
    return int(conn.execute("INSERT INTO agent_runs (scenario_id,trigger,status,started_at) VALUES (?,?,?,?)",
        (scenario_id, trigger, "running", _now())).lastrowid)

def finish_run(conn, run_id, *, status, summary, llm_model=None, metadata=None):
    conn.execute("UPDATE agent_runs SET status=?,summary_text=?,llm_model=?,metadata_json=?,finished_at=? WHERE id=?",
        (status, summary, llm_model, json.dumps(metadata or {}), _now(), run_id))

def add_finding(conn, run_id, *, severity, category, title, detail_md="", evidence=None, human_action="", kb_refs=None):
    return int(conn.execute("""INSERT INTO agent_findings (run_id,severity,category,title,detail_md,evidence_json,
        suggested_human_action,kb_refs_json,created_at) VALUES (?,?,?,?,?,?,?,?,?)""",
        (run_id, severity, category, title, detail_md, json.dumps(evidence or {}), human_action,
         json.dumps(kb_refs or []), _now())).lastrowid)

def list_findings(conn, *, severity=None, limit=50, open_only=True):
    q, params = "SELECT * FROM agent_findings WHERE 1=1", []
    if severity: q += " AND severity=?"; params.append(severity)
    if open_only: q += " AND acknowledged_at IS NULL"
    q += " ORDER BY id DESC LIMIT ?"; params.append(limit)
    return [dict(r) for r in conn.execute(q, params)]

def list_action_log(conn, limit=100):
    return [dict(r) for r in conn.execute("SELECT * FROM agent_action_log ORDER BY id DESC LIMIT ?", (limit,))]

def index_kb_file(conn, source_path, text):
    _exec_retry(conn, "DELETE FROM agent_kb_chunks WHERE source_path=?", (source_path,))
    now = _now()
    for i in range(0, len(text), 1200):
        _exec_retry(conn, "INSERT INTO agent_kb_chunks (source_path,chunk_text,indexed_at) VALUES (?,?,?)",
            (source_path, text[i:i+1200], now))
    conn.commit()

def search_kb(conn, query, limit=8):
    terms = [t.strip().lower() for t in query.split() if len(t.strip()) > 2]
    if not terms: return []
    scored = []
    for row in conn.execute("SELECT source_path,chunk_text FROM agent_kb_chunks"):
        score = sum(1 for t in terms if t in row["chunk_text"].lower())
        if score: scored.append((score, {"source": row["source_path"], "snippet": row["chunk_text"][:400]}))
    scored.sort(key=lambda x: -x[0])
    return [s[1] for s in scored[:limit]]


SEVERITY_RANK = {"info": 0, "warn": 1, "high": 2, "critical": 3}


def _severity_max(a: str, b: str) -> str:
    return a if SEVERITY_RANK.get(a, 0) >= SEVERITY_RANK.get(b, 0) else b


def _incident_row(conn, incident_id: int) -> dict | None:
    row = conn.execute("SELECT * FROM agent_incidents WHERE id=?", (incident_id,)).fetchone()
    return dict(row) if row else None


def _enrich_incident(conn, row: dict) -> dict:
    from app.agents.catalog import AGENT_CATALOG

    item = dict(row)
    p = AGENT_CATALOG.get(item.get("primary_agent", ""))
    item["agent_name"] = p.name if p else item.get("primary_agent", "Vigia")
    return item


def upsert_incident(
    conn,
    *,
    scenario_id: str,
    finding_id: int,
    title: str,
    severity: str,
    primary_agent: str,
    suggested_human_action: str = "",
) -> tuple[dict, bool]:
    """Retorna (incidente, notify_operators). notify=True na 1ª ocorrência ou escalação."""
    from app.agents import messages as agent_messages

    now = _now()
    row = conn.execute("SELECT * FROM agent_incidents WHERE scenario_id=?", (scenario_id,)).fetchone()
    if row:
        item = dict(row)
        prev_sev = item["severity"]
        new_sev = _severity_max(prev_sev, severity)
        was_closed = item["status"] != "open"
        notify = was_closed or SEVERITY_RANK.get(new_sev, 0) > SEVERITY_RANK.get(prev_sev, 0)
        if was_closed:
            conn.execute(
                """UPDATE agent_incidents SET occurrence_count=occurrence_count+1, last_seen_at=?,
                   latest_finding_id=?, severity=?, title=?, suggested_human_action=?,
                   primary_agent=?, status='open', acknowledged_at=NULL, acknowledged_by=NULL WHERE id=?""",
                (now, finding_id, new_sev, title, suggested_human_action or item.get("suggested_human_action"), primary_agent, item["id"]),
            )
        else:
            conn.execute(
                """UPDATE agent_incidents SET occurrence_count=occurrence_count+1, last_seen_at=?,
                   latest_finding_id=?, severity=?, title=?, suggested_human_action=?,
                   primary_agent=? WHERE id=?""",
                (now, finding_id, new_sev, title, suggested_human_action or item.get("suggested_human_action"), primary_agent, item["id"]),
            )
        out = _incident_row(conn, item["id"])
        if out and not out.get("thread_id"):
            thread_id = agent_messages.create_thread(
                conn, subject=title, primary_agent=primary_agent, severity=new_sev, related_finding_id=finding_id
            )
            conn.execute("UPDATE agent_incidents SET thread_id=? WHERE id=?", (thread_id, item["id"]))
            out = _incident_row(conn, item["id"])
        return _enrich_incident(conn, out or {}), notify

    thread_id = agent_messages.create_thread(
        conn, subject=title, primary_agent=primary_agent, severity=severity, related_finding_id=finding_id
    )
    iid = int(
        conn.execute(
            """INSERT INTO agent_incidents
               (scenario_id, primary_agent, severity, status, title, latest_finding_id,
                occurrence_count, first_seen_at, last_seen_at, suggested_human_action, thread_id)
               VALUES (?,?,?,?,?,?,1,?,?,?,?)""",
            (scenario_id, primary_agent, severity, "open", title, finding_id, now, now, suggested_human_action, thread_id),
        ).lastrowid
    )
    return _enrich_incident(conn, _incident_row(conn, iid) or {}), True


def list_incidents(
    conn,
    *,
    status: str = "open",
    severity: str | None = None,
    agent_id: str | None = None,
    limit: int = 50,
) -> list[dict]:
    q = "SELECT * FROM agent_incidents WHERE 1=1"
    params: list[Any] = []
    if status != "all":
        q += " AND status=?"
        params.append(status)
    if severity:
        q += " AND severity=?"
        params.append(severity)
    if agent_id:
        q += " AND primary_agent=?"
        params.append(agent_id)
    q += " ORDER BY CASE severity WHEN 'critical' THEN 0 WHEN 'high' THEN 1 WHEN 'warn' THEN 2 ELSE 3 END, last_seen_at DESC LIMIT ?"
    params.append(limit)
    return [_enrich_incident(conn, dict(r)) for r in conn.execute(q, params)]


def get_incident(conn, incident_id: int) -> dict | None:
    row = conn.execute("SELECT * FROM agent_incidents WHERE id=?", (incident_id,)).fetchone()
    return _enrich_incident(conn, dict(row)) if row else None


def ack_incident(conn, incident_id: int, username: str) -> dict | None:
    inc = get_incident(conn, incident_id)
    if not inc:
        return None
    now = _now()
    conn.execute(
        "UPDATE agent_incidents SET status='ack', acknowledged_at=?, acknowledged_by=? WHERE id=?",
        (now, username, incident_id),
    )
    if inc.get("latest_finding_id"):
        conn.execute(
            "UPDATE agent_findings SET acknowledged_at=?, acknowledged_by=? WHERE id=? AND acknowledged_at IS NULL",
            (now, username, inc["latest_finding_id"]),
        )
    if inc.get("thread_id"):
        conn.execute(
            """UPDATE agent_messages SET acknowledged_at=?, acknowledged_by=?
               WHERE thread_id=? AND requires_human=1 AND acknowledged_at IS NULL""",
            (now, username, inc["thread_id"]),
        )
    return get_incident(conn, incident_id)


def resolve_incidents_for_scenario(conn, scenario_id: str) -> None:
    """Marca incidente resolvido quando cenário volta a OK."""
    conn.execute(
        "UPDATE agent_incidents SET status='resolved', last_seen_at=? WHERE scenario_id=? AND status='open'",
        (_now(), scenario_id),
    )


def get_overview(conn) -> dict:
    import os

    from app.agents import llm_client

    last_tick = conn.execute(
        "SELECT ts, message, payload_json FROM agent_action_log WHERE event_type='tick.complete' ORDER BY id DESC LIMIT 1"
    ).fetchone()
    open_counts = {r["severity"]: r["c"] for r in conn.execute(
        "SELECT severity, COUNT(*) c FROM agent_incidents WHERE status='open' GROUP BY severity"
    )}
    scenarios = list_scenarios(conn)
    ok_count = sum(1 for s in scenarios if s.get("last_run_status") == "ok")
    payload = {}
    if last_tick and last_tick["payload_json"]:
        try:
            payload = json.loads(last_tick["payload_json"])
        except json.JSONDecodeError:
            pass
    return {
        "tier": "t1" if llm_client.AGENTIC_LLM_ENABLED else "t0",
        "ollama": llm_client.ollama_available(),
        "model": llm_client.AGENTIC_LLM_MODEL,
        "last_tick_at": last_tick["ts"] if last_tick else None,
        "last_tick_status": "degraded" if payload.get("runs") and any(
            r.get("findings_count", 0) > 0 for r in payload.get("runs", []) if isinstance(r, dict)
        ) else "ok",
        "scenarios_total": len(scenarios),
        "scenarios_ok": ok_count,
        "incidents_open": {
            "critical": open_counts.get("critical", 0),
            "high": open_counts.get("high", 0),
            "warn": open_counts.get("warn", 0),
            "info": open_counts.get("info", 0),
        },
        "worker_interval_sec": int(os.getenv("AGENTIC_INTERVAL_SEC", "600")),
    }


def recent_runs_for_scenario(conn, scenario_id: str, limit: int = 12) -> list[dict]:
    return [
        dict(r)
        for r in conn.execute(
            """SELECT id AS run_id, status, started_at, summary_text,
                      (SELECT COUNT(*) FROM agent_findings f WHERE f.run_id=agent_runs.id) AS findings_count
               FROM agent_runs WHERE scenario_id=? ORDER BY id DESC LIMIT ?""",
            (scenario_id, limit),
        )
    ]


def backfill_incidents(conn) -> int:
    """Consolida findings abertos legacy em agent_incidents (one-time safe)."""
    rows = conn.execute(
        """SELECT f.id AS finding_id, f.severity, f.title, f.suggested_human_action, f.created_at,
                  r.scenario_id
           FROM agent_findings f
           JOIN agent_runs r ON r.id = f.run_id
           WHERE f.acknowledged_at IS NULL
           ORDER BY f.id ASC"""
    ).fetchall()
    seen: set[str] = set()
    n = 0
    for row in rows:
        sid = row["scenario_id"]
        if sid in seen:
            continue
        seen.add(sid)
        existing = conn.execute("SELECT id FROM agent_incidents WHERE scenario_id=?", (sid,)).fetchone()
        if existing:
            continue
        from app.agents.catalog import SCENARIO_AGENT_MAP

        agent_id = SCENARIO_AGENT_MAP.get(sid, "sentinel")
        upsert_incident(
            conn,
            scenario_id=sid,
            finding_id=row["finding_id"],
            title=row["title"],
            severity=row["severity"],
            primary_agent=agent_id,
            suggested_human_action=row["suggested_human_action"] or "",
        )
        n += 1
    conn.commit()
    return n