turboprivate-ai 0.1.1

Unified platform for self-hosted LLM inference + enterprise safety governance

TurboPrivate AI

Unified platform for self-hosted LLM inference + enterprise safety governance.

Features

• Inference Engine — vLLM/llama.cpp backends with auto hardware detection, TurboQuant-v3 INT4 quantization, QoS routing, dynamic batching, KV/prompt caching
• Safety Gate — Pre-flight and post-flight verification with 7 verifiers: anti-hacking, PII detection, prompt injection, overengineering, patch analysis, vulnerability scanning, hallucination scoring
• Memory & RAG — Vector store with cosine similarity search, multi-format document parser (PDF, DOCX, HTML, CSV, JSON, XML, YAML, Markdown)
• Infrastructure — K3s provisioning via SSH/Terraform, age-encrypted backups, Helm chart deployment
• Frontend — React SPA dashboard with model management, safety governance, memory search, and settings
• Auth — JWT-based authentication with RBAC
• Observability — Prometheus metrics, OpenTelemetry tracing, structured logging
• CLI — 20+ commands for model management, safety policies, infrastructure, and backups

Quick Start

pip install turboprivate-ai
turbo serve --reload

Changelog

0.1.1 (2026-05-11)

• Migrated to hatchling build system
• Fixed missing InferenceEngine import in turbo.inference
• Fixed TracerProvider bug in OpenTelemetry instrumentation
• Added structured logging to all exception handlers
• Consolidated Celery workers into shared worker.celery_app
• Added CI workflow with ruff linting + pytest
• Improved graceful shutdown (audit trail flush)
• Updated dependencies (replaced unstructured with actual used libs)

Documentation

• turbo/cli.py — CLI entry point and command reference
• turbo/api/main.py — FastAPI application
• turbo/safety/gate.py — Safety gate configuration
• frontend/ — React SPA
• helm/turboprivate/ — Kubernetes deployment