AI Configuration Management Software is a full-stack, self-hostable AI-powered configuration management software built for engineering teams that need to manage environment variables, secrets, feature flags, and service settings across multiple environments — with an AI layer that actively monitors, explains, and improves your configuration state. You buy the complete source code once and run it on your own infrastructure.
The software is built on a Python/FastAPI backend, a React/TypeScript frontend (Vite + Tailwind CSS), PostgreSQL, Redis, Celery for background task processing, and ChromaDB as a vector store for semantic search. Everything ships as a Docker Compose stack, including a production configuration with Nginx as a reverse proxy. RBAC (Owner, Editor, Viewer roles) controls access at the workspace level. All secrets are encrypted at rest with AES-256-GCM and are never returned in plaintext through list endpoints — a reveal action is required and is individually audited. Every API mutation is recorded in a full audit log with user, action, resource, IP, and timestamp.
The AI layer ships nine capabilities powered by the Anthropic Claude API and a sentence-transformer embedding pipeline:
- Natural language config chat — ask questions about your actual config data in plain English
- Anomaly detection — an IsolationForest model extracts feature vectors from config change records (timestamp, value entropy, change frequency, sensitivity flag) and scores each change 0–1 for unusual patterns
- AI config recommendations — on-demand analysis for missing keys, insecure values, duplicates, and performance issues
- Semantic config search — find keys by describing what they do
- AI-powered compliance checker — evaluates configs against built-in security rules (no DEBUG in prod, no plaintext credentials, no wildcard CORS, and more) and generates Claude-written remediation advice
- Environment summary generation — one-click markdown summary of any environment's full configuration
- Predictive drift alert — forecasts when dev/prod divergence will become critical based on recent drift history
- AI change explainer — explains what any config change means operationally
- Smart config validator — validates values for type, format, and contextual correctness on import or entry
Config drift detection runs on a configurable schedule (default every 15 minutes), compares keys across all environments, and produces a drift score from 0 to 100. Configs can be imported and exported in JSON, YAML, TOML, and .env formats. Feature flags support boolean, string, and number types with per-environment overrides.
To improve AI model coverage and protect real user information, the software uses synthetic data generation — a seed script populates the database with representative config entries, secrets, feature flags, and workspace structures across dev, staging, and production environments. This synthetic data is used to bootstrap the anomaly detection model and the vector store without requiring exposure of any real credentials or configuration values.
