AEGIS Framework

Open-source autonomous reliability engineering framework for AI-powered incident investigation, reasoning, remediation, and reliability automation.

---

🚀 Status: v0.1.0 Public Release
AEGIS Framework is a production-grade, open-source autonomous reliability engineering framework providing modular provider interfaces, deterministic safety guardrails, AI reasoning engines, and automated remediation.

---

---

What is AEGIS?

AEGIS is an open-source, AI-native reliability and controlled autonomous remediation framework for production systems (aegis-resilience on PyPI). It correlates multimodal observability signals (metrics, logs, traces), performs causal failure reasoning, evaluates fail-closed safety policies, and executes approval-gated remediation.

Why Use AEGIS?

• Stop Reinventing SRE Bots: Standardize incident triage, hypothesis generation, and blast-radius assessment.
• Fail-Closed Safety: Built-in LLM firewalls (llmfirewall-core), strict policy gates, and bounded autonomy budgets (LEVEL_0 to LEVEL_3).
• Production Observability: Native support for Prometheus, Grafana Loki, OpenTelemetry, and Kubernetes.
• Zero Hallucinated Actions: Typed, deterministic remediation actions with pre-execution safety checks and post-remediation recovery validation.

Installation

pip install aegis-resilience

---

Consumption Paths

🧑‍💻 Path 1: Human Developers

Initialize and test autonomous reliability locally in 60 seconds with zero external infrastructure:

aegis init --example
aegis doctor
aegis demo

🤖 Path 2: AI Coding Agents (Claude Code, Cursor, Codex, Copilot)

AI coding agents can discover, configure, and safely integrate AEGIS without inventing parallel reliability subsystems:

• Canonical Agent Guide: docs/ai/integration.md
• Agent Rules & Invariants: AGENTS.md
• Machine-Readable Manifest: docs/ai/capabilities.yaml
• AI Agent Quickstart: docs/ai/quickstart.md

---

Relationship to AEGIS AI

Component	Role	Description
AEGIS Framework	Reusable Library	The modular, lightweight, open-source Python framework (pip install aegis-resilience) providing foundational abstractions, evidence models, reasoning interfaces, security guardrails, and autonomy boundaries.
AEGIS AI	Reference Platform	The enterprise flagship application providing full end-to-end production deployment, databases, web UI, vector search, and integrated microservices.

AEGIS Framework
    =
Reusable framework

AEGIS AI
    =
Full reference platform built around the same reliability architecture

---

Architecture Overview

Application / Client Workflows
              │
              ▼
       AEGIS Framework
              │
  ┌───────────┼───────────┬───────────┐
  ▼           ▼           ▼           ▼
Incident   Evidence   AI Reasoning   AI Security
Management Engineering (Hypotheses)  (Fail-Closed)
  │           │           │           │
  └───────────┼───────────┴───────────┘
              │
  ┌───────────┼───────────┬───────────┐
  ▼           ▼           ▼           ▼
Reliability Policy     Autonomy    Remediation
(SLO/Budgets) (Gates)  (Levels 0-3)(Verified)
              │
              ▼
          Evaluation
       (Golden/Release)
              │
              ▼
          Providers
 (LLM / Telemetry / Vector)

---

Core Capabilities (Roadmap)

• Incident Lifecycle Management: Structured representations of incidents, phases, status, and impact.
• Multimodal Evidence Engineering: Ingestion, temporal correlation, and normalization across metrics, logs, and distributed traces.
• Root Cause & Causal Reasoning: Structured hypothesis generation, causal DAGs, and validation loops.
• Fail-Closed AI Security: Prompt injection defense, LLM firewalls, deterministic boundary validations, and strict RBAC.
• Tiered Autonomy: Strictly bounded execution loops (LEVEL_0 observation to LEVEL_3 autonomous actions) with hard budgets.
• Deterministic Policy & Approval: Non-bypassable human approval gates, change freeze locks, and kill switches.
• Verified Remediation: Dry-run capabilities, safe action execution, rollback strategies, and deployment verification.
• Continuous Evaluation: Golden datasets, adversarial evaluation, and strict release gates.

---

Installation

# Clone the repository
git clone https://github.com/Livesh-L-28/AEGIS-Framework.git
cd AEGIS-Framework

# Install in editable mode
pip install -e .

---

Quickstart & Developer Experience

AEGIS provides an out-of-the-box CLI to scaffold, validate, and simulate autonomous incident engineering in seconds without requiring external infrastructure:

# 1. Initialize a new AEGIS project (or generate a standalone in-memory template)
aegis init --example

# 2. Run comprehensive framework, configuration, and security diagnostics
aegis doctor

# 3. Simulate the complete 7-step autonomous reliability lifecycle locally
aegis demo

Python API Integration

from aegis import Aegis

# Initialize directly from human-readable configuration with env var interpolation
aegis = Aegis.from_config("aegis.yaml")

# Run autonomous lifecycle
print(f"AEGIS Framework active: {aegis.config.project.name} (Autonomy Level {aegis.config.autonomy.level})")

---

Development

# Install development dependencies
pip install -e ".[dev]"

# Run linters and type checkers
ruff check .
mypy aegis

# Run tests
pytest

---

Contributing

We welcome community contributions! Please review CONTRIBUTING.md and our CODE_OF_CONDUCT.md before submitting pull requests.

---

Security

Security is foundational to AEGIS. For vulnerability reporting guidelines and our fail-closed security architecture, see SECURITY.md.

---

License

AEGIS Framework is licensed under the Apache License, Version 2.0.