Intent Drift Detector

AI-assisted detection of configuration drift between intended state and runtime state in ONES-FM–managed fabrics

---

Overview

Intent Drift Detector identifies mismatches between the intended network configuration (defined via user-provided YAML intents and stored in ONES-FM) and the actual runtime configuration deployed on network devices.

In real production environments, configuration drift occurs due to:

• Manual CLI changes outside the controller
• Partial or failed orchestration runs
• Automation bugs
• Emergency hotfixes applied directly on devices

This project automates drift detection and reduces MTTR from hours to minutes.

---

Problem Statement

ONES-FM assumes that:

The configuration stored in the database always matches the configuration running on the devices.

In practice, this assumption frequently breaks.

Current Challenges

• No automated validation between intent and runtime
• Manual DB exports and CLI inspections
• Drift detection is slow, reactive, and error-prone
• High operational risk in large fabrics

---

Solution Overview

The Intent Drift Detector:

1. Fetches intended state from ONES-FM
2. Collects runtime state from network devices
3. Normalizes both into a common data model
4. Detects configuration drift
5. Generates structured drift reports with severity and remediation hints

---

Architecture

        +----------------------+
        |  ONES-FM MCP Server  |
        |  (Intent & Runtime)  |
        +----------+-----------+
                   |
                   v
        +----------------------+
        | FabricIntentAgent    |
        | - Intent fetch       |
        | - Runtime fetch      |
        | - Normalization      |
        +----------+-----------+
                   |
                   v
        +----------------------+
        | Drift Engine         |
        | - JSON diff          |
        | - Severity scoring   |
        +----------+-----------+
                   |
                   v
        +----------------------+
        | ReportAgent          |
        | - Markdown / CSV     |
        | - Patch suggestions  |
        +----------------------+

Screenshots

Figure 1. Intent Drift Check

Figure 2. Sample ONES-FM Intent