---
title: "April 2026 - AI / Insights"
slug: "april-2026-ai-insights"
updated: 2026-06-29T20:41:35Z
published: 2026-06-29T20:41:35Z
canonical: "kb.kentik.com/april-2026-ai-insights"
---

> ## Documentation Index
> Fetch the complete documentation index at: https://kb.kentik.com/llms.txt
> Use this file to discover all available pages before exploring further.

# April 2026 - AI / Insights

The following updates were released to the Kentik platform for AI / Insights during April 2026.

## Features

- **Durable tool approval flow**: AI Advisor tool approvals now persist across page refreshes, deployments, and long approval windows, enabling multi-hour human-in-the-loop Auto Investigation scenarios.
- **Auto-approved tools support**: Added support for auto-approved tools so certain remediation actions can execute without per-turn human approval, plus a new endpoint listing all available tools and their status.
- **Approval-pending visibility**: AI Advisor now surfaces an awaiting-approval state at the session level when any message needs user approval.
- **Faster, asynchronous chat**: AI Advisor chat now accepts user messages with an immediate response and lets clients poll for results, improving responsiveness.
- **AI Advisor traceroute analysis for Synthetics**: You can now ask AI Advisor to analyze traceroute results for a given IP and agent, with hop-by-hop latency returned for troubleshooting.
- **Improved NMS query tool reliability**: The AI Advisor NMS tool was ported to TypeScript and hardened against schema validation errors that could cause natural-language query failures.
- **RBAC enforcement for AI tools**: AI Advisor now respects role-based permissions, so users only see tools they are entitled to and cannot access restricted data through natural-language queries.

## Bug Fixes

- **Syslog & Trap tool now returns results**: Fixed the AI Advisor Syslog & Trap tool failing to return data by validating queries against dynamically loaded dictionary values.
- **Overlay mode reconnect crash fixed**: Resolved a crash when entering a prompt after a server restart or websocket reconnection.
- **Reduced response cutoffs**: Increased AI Advisor output token limits and made turn compression less aggressive so investigation responses are no longer truncated.
- **Chat memory leak fixed**: Fixed a memory leak in chat sessions where reconnect listeners were not cleaned up, improving stability.
- **Reliable market intelligence queries**: Market Intelligence (KMI) insights now use a fixed lookback date instead of a dynamic timestamp for consistent, predictable results.
- **Protected insights from rogue restrictions**: Hardened the insights restriction endpoint to prevent security scanners from inadvertently creating DENY rules that disabled insights.