pn-new-crm/src/lib/services/error-events.service.ts

/**
 * Error event capture + retrieval.
 *
 * `captureErrorEvent(...)` is called from `errorResponse(...)` whenever
 * an unhandled (5xx) error fires inside a route handler. It pulls the
 * request context from AsyncLocalStorage, sanitizes the payload, and
 * inserts one row into `error_events`. Failure to write must NEVER
 * throw — the caller is already in the error path.
 *
 * `listErrorEvents` / `getErrorEventById` back the super-admin inspector.
 */

import { and, desc, eq, gte, lte } from 'drizzle-orm';

import { db } from '@/lib/db';
import { errorEvents, type ErrorEvent } from '@/lib/db/schema/system';
import { logger } from '@/lib/logger';
import { getRequestContext } from '@/lib/request-context';

const STACK_MAX_BYTES = 4 * 1024;
const BODY_MAX_BYTES = 1 * 1024;

/** Keys whose values are never persisted to the body excerpt. */
const SENSITIVE_KEYS = new Set([
  'password',
  'newPassword',
  'oldPassword',
  'token',
  'secret',
  'apiKey',
  'accessKey',
  'secretKey',
  'creditCard',
  'cardNumber',
  'cvv',
  'ssn',
  'authorization',
]);

/** Drop sensitive keys + cap the JSON length. */
function sanitizeBody(body: unknown): string | null {
  if (body === null || body === undefined) return null;
  let cloned: unknown;
  try {
    cloned = JSON.parse(JSON.stringify(body));
  } catch {
    return null;
  }
  function walk(value: unknown): unknown {
    if (Array.isArray(value)) return value.map(walk);
    if (value && typeof value === 'object') {
      const out: Record<string, unknown> = {};
      for (const [k, v] of Object.entries(value)) {
        if (SENSITIVE_KEYS.has(k)) {
          out[k] = '[REDACTED]';
        } else {
          out[k] = walk(v);
        }
      }
      return out;
    }
    return value;
  }
  const sanitized = walk(cloned);
  let serialized: string;
  try {
    serialized = JSON.stringify(sanitized);
  } catch {
    return null;
  }
  if (Buffer.byteLength(serialized, 'utf8') > BODY_MAX_BYTES) {
    serialized = serialized.slice(0, BODY_MAX_BYTES) + '…[truncated]';
  }
  return serialized;
}

interface CaptureArgs {
  statusCode: number;
  error: unknown;
  /** Optional structured metadata (e.g. zod issues parsed from a ZodError). */
  metadata?: Record<string, unknown>;
  /** Sanitized request body (already JSON-serializable). Optional. */
  body?: unknown;
}

/**
 * Persist an error_events row tied to the active request context.
 * Best-effort — silently swallows any DB failure (the caller is
 * already returning the user an error response; we do NOT want to
 * mask the original error with a logging-pipeline failure).
 */
export async function captureErrorEvent(args: CaptureArgs): Promise<void> {
  const ctx = getRequestContext();
  if (!ctx) {
    // Outside a request context (e.g. queue worker). Log but skip — the
    // queue has its own failure-capture in BullMQ.
    return;
  }
  try {
    const err = args.error;
    const errorName = err instanceof Error ? err.name : typeof err;
    const errorMessage = err instanceof Error ? err.message : err === undefined ? '' : String(err);
    const stack = err instanceof Error && err.stack ? err.stack.slice(0, STACK_MAX_BYTES) : null;
    const durationMs = Date.now() - ctx.startedAt;

    // Pull through any well-known fields the upstream library decorated
    // onto the error — Postgres driver uses `code` (SQLSTATE) and
    // `severity`, fetch errors carry `cause.code`, etc. The classifier
    // reads from `metadata.code` to drive the "likely culprit" badge.
    const enriched: Record<string, unknown> = { ...(args.metadata ?? {}) };
    if (err && typeof err === 'object') {
      const e = err as { code?: unknown; severity?: unknown; cause?: { code?: unknown } };
      if (typeof e.code === 'string') enriched.code = e.code;
      if (typeof e.severity === 'string') enriched.severity = e.severity;
      if (e.cause && typeof e.cause === 'object' && typeof e.cause.code === 'string') {
        enriched.causeCode = e.cause.code;
      }
    }

    await db
      .insert(errorEvents)
      .values({
        requestId: ctx.requestId,
        portId: ctx.portId || null,
        userId: ctx.userId || null,
        statusCode: args.statusCode,
        method: ctx.method,
        path: ctx.path,
        errorName,
        errorMessage,
        errorStack: stack,
        requestBodyExcerpt: sanitizeBody(args.body),
        metadata: enriched,
        durationMs,
      })
      .onConflictDoNothing();
  } catch (writeErr) {
    // Logged but never thrown — the caller is in the error path already.
    logger.error({ err: writeErr }, 'Failed to persist error_events row');
  }
}

export interface ListErrorEventsFilter {
  portId?: string;
  statusCode?: number;
  /** ISO date strings; defaults to last 7 days. */
  from?: string;
  to?: string;
  limit?: number;
}

export async function listErrorEvents(filter: ListErrorEventsFilter): Promise<ErrorEvent[]> {
  const conditions = [];
  if (filter.portId) conditions.push(eq(errorEvents.portId, filter.portId));
  if (filter.statusCode) conditions.push(eq(errorEvents.statusCode, filter.statusCode));
  if (filter.from) conditions.push(gte(errorEvents.createdAt, new Date(filter.from)));
  if (filter.to) conditions.push(lte(errorEvents.createdAt, new Date(filter.to)));

  return db
    .select()
    .from(errorEvents)
    .where(conditions.length ? and(...conditions) : undefined)
    .orderBy(desc(errorEvents.createdAt))
    .limit(filter.limit ?? 100);
}

export async function getErrorEventById(requestId: string): Promise<ErrorEvent | null> {
  const row = await db.query.errorEvents.findFirst({
    where: eq(errorEvents.requestId, requestId),
  });
  return row ?? null;
}