Automatically diagnose Kubernetes incidents from logs, metrics, and traces.
Chat with an AI agent that can inspect and repair cluster and node issues.
Run safe, automatic remediation for routine problems while requiring approval for risky ones.
Keep a full audit trail of every diagnostic and repair action taken on a cluster.
| pro-deploy/aegil | 2417467487-hub/trend2video-pro | stanleynull/autohunter | |
|---|---|---|---|
| Stars | 111 | 111 | 111 |
| Language | Python | Python | Python |
| Setup difficulty | hard | moderate | moderate |
| Complexity | 5/5 | — | 4/5 |
| Audience | ops devops | vibe coder | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires a Kubernetes cluster, building and publishing container images, a language model endpoint, and manual review of manifests for your environment.
Aegil is a tool that watches over a Kubernetes cluster and helps fix problems on it, acting like an automated site reliability engineer. Instead of an engineer manually digging through logs and dashboards when something breaks, Aegil connects to the cluster, the logs, and the monitoring systems, figures out what is actually wrong, and can take action to repair it, all while the person operating it just chats with it in plain language. Under the hood, a deterministic pipeline reads log lines, gathers facts, runs a set of built in checks, and builds a conclusion where every claim is backed by an exact quote from a real log line rather than a guess. A small trained classifier helps sort routine requests from harder ones that get escalated to a larger language model. The language model itself does not have free rein: a separate rule based classifier decides how risky any action is, so the model cannot skip past a required confirmation step for anything destructive, like deleting data or removing storage volumes. The project ships as four separate services plus an optional set of open source monitoring tools. One service runs with elevated access on cluster nodes to inspect disk space, processes, and memory and to clean things up when needed. Another is the chat panel where an operator talks to the agent and reviews an audit trail of everything it has done. A safety mode called dry run is on by default, meaning the agent will observe and suggest but not act until autonomous mode is explicitly turned on, and even then, safe fixes happen automatically while destructive ones always wait for a human to confirm. Setting it up means building container images, applying Kubernetes manifests, and creating secrets for authentication, and it currently expects the operator to review cluster specific details like storage classes before deploying. The project describes itself as an early extraction from a larger platform, version 0.1.0, with the core pieces tested but some parts, like a Helm chart and full configuration options, still incomplete.
An autonomous Kubernetes SRE agent that observes cluster logs and metrics, diagnoses problems, and can safely fix them through chat, with built in guardrails against risky actions.
Mainly Python. The stack also includes Python, Kubernetes, FastAPI.
No license information is stated in the visible README, so reuse and redistribution terms are unclear.
Setup difficulty is rated hard, with roughly 1day+ to a first successful run.
Mainly ops devops.
This repo across BitVibe Labs
double-check against the repo, no cap.