Evidently ML Observability Platform

This is a repackaged open source software product wherein additional charges apply for cloudimg support services.

## Evidently ML Observability Platform - Preinstalled and Production-Ready

Evidently is an open source platform for evaluating, testing, and monitoring machine learning, large language model, and data pipelines. This image delivers Evidently fully installed and configured as a system service on EC2, so a private observability dashboard is running within minutes of launch - no manual setup, no Docker orchestration, no dependency troubleshooting.

## Why This Image Over a Manual Install

Installing Evidently from source requires configuring Python environments, setting up process supervision, configuring a reverse proxy, provisioning storage, and implementing authentication - the upstream project ships no login mechanism. This image eliminates that entire workflow:

• Evidently installed into a dedicated Python virtual environment on the current LTS Python
• Dedicated unprivileged service account, supervised by systemd with automatic restart
• Monitoring dashboard bound to loopback only, published on port 80 by an nginx reverse proxy with WebSocket support
• nginx HTTP Basic Auth gate with a per-instance password generated on first boot - the open source UI ships no login of its own
• Dedicated resizable data disk for the workspace, projects, and Report snapshots - survives instance replacement

## Real Data on First Login

The open source Evidently UI ships empty. On first boot, a one-shot service seeds a demo project containing a genuine data drift and data quality Report over a synthetic sample. The dashboard shows a real, explorable Report the moment you log in rather than a blank workspace. To monitor your own data, compute Evidently Reports over your datasets with the bundled Evidently Python SDK and add their snapshots to a project in the workspace.

## Secure First Boot

On first boot a one-shot service generates a fresh dashboard password unique to that instance, stores it as a bcrypt hash in an nginx HTTP Basic Auth file, and writes the plaintext password to a root-only file. Unauthenticated requests to the dashboard are refused; an unauthenticated health endpoint is exposed for load balancer probes only. No shared or default credentials ship in the image, and the round trip is verified on every boot.

## Security Practices

• No default or shared credentials - every instance generates its own password
• Dashboard never exposed directly; nginx reverse proxy is the sole entry point
• Service runs as an unprivileged account with no shell access
• Workspace data isolated on a dedicated EBS volume, independently encryptable via AWS KMS
• Built on a fully patched base image with the OS security baseline applied at build time

## Use Cases

• Production ML model monitoring: track data drift and performance decay for fraud detection, recommendation, or pricing models
• Data quality validation: run automated data quality Reports after nightly batch jobs write to S3, catching schema changes and distribution shifts before they reach downstream models
• LLM and retrieval pipeline observability: monitor embedding drift and response quality metrics for RAG pipelines
• Team-wide observability dashboard: share an explorable data and model health dashboard across data science, ML engineering, and platform teams
• Versioned audit trail: keep a timestamped history of data and model checks on your own infrastructure for compliance and reproducibility

## Getting Started

1. Launch the AMI from AWS Marketplace

2. SSH into the instance and retrieve your unique dashboard password from the root-only file

3. Open port 80 in your browser, sign in, and explore the seeded demo Report

4. Install the Evidently Python SDK in your pipeline environment and push Report snapshots to the workspace

For a guided pilot deployment or instance sizing consultation, contact the cloudimg team through the support channels listed on this page.

## Evaluation Resources

The seeded demo project serves as a hands-on evaluation artifact - explore a real data drift and data quality Report immediately after launch without writing any code. When ready to evaluate with your own data, use the bundled Evidently Python SDK to compute Reports over CSV files and Pandas DataFrames; the seeded demo project's own generator script ships on the image as a working example.

## About cloudimg

cloudimg provides 24/7 technical support by email and live chat. Our engineers help with deployment, workspace configuration, building data drift and model quality Reports, wiring Evidently into batch or streaming pipelines, TLS configuration, data disk resizing, and runtime performance tuning.