ANKR-GURU and DevBrain: Level-3 Autonomous AI Operations Daemon with Self-Learning Auto-Healing

Read full text Summarise with VIRAT
Authors: Anil Sharma
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This document is submitted as a defensive publication. No patent rights are asserted. The submission date establishes prior art for each mechanism described. We describe two interrelated innovations developed in the course of building AnkrStack — a production AI platform comprising 192 registered services across multiple domains including maritime, logistics, legal, and enterprise operations. The two contributions are: (1) ANKR-GURU — Universal AI Daemon operating at Level 3 autonomy, a 24/7 autonomous monitoring and remediation system implementing the full Observe → Analyze → Prescribe → Execute operational loop without human intervention, with self-learning feedback from resolved incidents; and (2) DevBrain — AI-Powered Auto-Healing System, a three-phase cascading diagnosis and repair engine integrated with PM2 process management, implementing pattern-based instant repair (Phase 1, zero cost), LLM-augmented diagnosis for unrecognized error patterns (Phase 2), and a self-building pattern library that grows from each resolved incident (Phase 3). The central novelty of this combined system is the placement on the autonomy taxonomy. Competitive monitoring solutions — Datadog, New Relic, Grafana, Splunk, PagerDuty — operate at Level 2: they detect conditions and notify humans. ANKR-GURU operates at Level 3: it detects conditions, analyzes root cause, prescribes a repair action, and executes that action autonomously, notifying humans only when the repair succeeds or when the confidence threshold for autonomous action is not met. DevBrain adds the learning dimension: each resolved incident, whether through pattern match or LLM diagnosis, updates the pattern library so that future occurrences of the same error are resolved faster and with higher confidence. Each contribution is described with sufficient technical precision — including algorithms, data models, confidence thresholds, and implementation evidence from production code — to constitute a complete prior art di

Fields of Study & Subjects

Software System Performance and Reliability Distributed systems and fault tolerance Advanced Software Engineering Methodologies

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
25 March 2026
Language
EN
Document type
Preprint