OpenUBA v0.0.2

complete ground-up modernization of the openuba platform.

video walkthrough:
https://youtu.be/tMppVt2v1nI

backend:

• migrated from flask to fastapi with uvicorn asgi server
• full rest api with interactive swagger/redoc docs
• pydantic schema validation on all endpoints
• sqlalchemy 2.0 orm with postgresql as system of record
• postgraphile auto-generated graphql api from postgres schema
• jwt authentication with bcrypt password hashing
• role-based access control (admin, manager, triage, analyst)
• per-page granular permissions configurable by admins
• apscheduler-based model execution scheduling

database:

• postgresql with 15+ tables (models, model_versions, model_artifacts, model_runs, anomalies, cases, rules, users, roles, permissions, integration_settings, notifications, system_logs, etc.)
• sql triggers for audit logging and data consistency
• schema migration system with backward compatibility from v1 json files
• default rbac roles seeded on initialization

model lifecycle:

• model installation system with multi-registry adapters (github, openuba hub, huggingface, local filesystem)
• model manifest system with sha hash verification at install and pre-execution
• multi-version model support with artifact tracking
• model orchestrator dispatching training and inference as ephemeral jobs
• cron-based scheduled model execution
• framework-aware artifact serialization (sklearn/joblib, pytorch/torch.save, tensorflow/savedmodel, networkx/pickle)

kubernetes-native execution:

• custom openuba operator (kopf) watching ubatraining and ubainference crds
• ephemeral k8s jobs for all model execution (no long-lived per-model services)
• dual execution drivers: kubernetesjobexecutiondriver and localdockerexecutiondriver
• framework-specific runner images (model-runner:sklearn, model-runner:pytorch, model-runner:tensorflow, model-runner:networkx)
• shared persistent volumes for model library, saved artifacts, and datasets
• kind cluster support for local development with host mount

data integration:

• elasticsearch connector for event storage, search, and anomaly indexing
• apache spark integration for distributed data processing
• data loader abstraction (local pandas csv, elasticsearch, spark)
• data source management with connection testing
• source groups for organizing data sources
• data ingestion pipeline with status monitoring

frontend:

• migrated from react/electron to next.js 14 with app router
• typescript throughout with tailwindcss and shadcn/ui component library
• dark mode default with light mode toggle
• comprehensive page structure: home dashboard, models, anomalies, cases, data, entities, rules, alerts, schedules, settings, user management
• real-time updates via graphql subscriptions
• apollo client for graphql data layer, axios for rest
• recharts for data visualization
• zustand for ui state management
• responsive layout with mobile support
• global command palette, system log dock, toast notifications
• login page with jwt session management

rules engine:

• flow-based visual rule builder with drag-and-drop canvas
• rule composition combining model outputs with logical operators
• deterministic json serialization of rule logic circuits
• rule-triggered alert generation
• rule enable/disable toggle
• crud api for rules management

llm assistant:

• always-available draggable/resizable chat overlay
• multi-provider support (ollama, openai, claude, gemini)
• streaming response handling with sse
• context-aware of current route, selected entities, and filters
• conversation persistence across sessions
• provider/model selection dropdown
• integration settings ui with connectivity testing

alerting and notifications:

• alert generation from rule execution
• persistent notification system with channels
• notification management ui

cases and anomalies:

• case management with anomaly linking and timeline
• anomaly detection result browsing with filtering and pagination
• feedback system for true/false positive marking
• case status tracking and assignment

infrastructure:

• kubernetes manifests for all components (backend, frontend, operator, postgres, elasticsearch, postgraphile, spark)
• crd definitions for ubatraining and ubainference
• persistent volume setup for model library, saved models, datasets
• docker images for backend, frontend, operator, and all runner variants
• makefile-driven workflow for all build, deploy, and test operations

testing:

• comprehensive pytest test suite
• end-to-end tests for model lifecycle, anomalies, cases, rules, data display
• unit tests for api routers, repositories, services
• integration tests for registry adapters and database operations
• k8s integration tests with crd and job monitoring

cleanup:

• removed accidentally committed pycache, .mypy_cache, and .ds_store files
• updated .gitignore for build artifacts and runtime data