Overview
model-quality-gate is described as an AI Quality & Model-Risk Platform and a production-promotion eval / red-team gate and model-risk evidence system for APAC banking. It takes a target consisting of a model, a prompt version, and a golden dataset, and returns four artifacts with supporting evidence: EvalReport, RedTeamReport, GateDecision, and ModelCard plus PromptVersion. A target passes only if the EvalReport clears every metric threshold and the RedTeamReport blocks every probe.
The product is built ports-and-adapters on the Gemini Enterprise Agent Platform, with a configurable deployment region defaulting to asia-southeast1. Its domain core owns orchestration and speaks only to ports, while four interchangeable adapter families implement those ports. Switching the entire managed stack to the offline local one, or to the on-prem migration target, is a one-line profile change with no domain edits.
The local profile binds every port to a real, deterministic, SDK-free in-process implementation and runs the whole promotion gate end to end with no Google Cloud, no API key, and no emulator. The onprem profile binds every port to a placeholder adapter that raises NotImplementedError, and the gcp profile uses the real managed stack in asia-southeast1. Every artifact, metric and decision is a pure-stdlib dataclass with zero dependency on Google Cloud, ADK, or any framework.
Pricing
Free
Detailed plans have not been confirmed in our catalog. Check the official website for current limits and billing terms.
Visit WebsitePrices and limits may change. Confirm the currency, billing period, seat minimum and usage caps on the official website.
Use Cases
- Teams use model-quality-gate to evaluate a target model, prompt version, and golden dataset and produce four evidence-backed artifacts.
- Teams use model-quality-gate to gate production promotion based on metric thresholds and red-team probes.
- Teams use model-quality-gate to run red-team probes across five attack families.
- Teams use model-quality-gate to publish golden datasets at runtime or read bundled JSONL.
- Teams use model-quality-gate to read recorded quality drift and return a DriftEscalation.
- Teams use model-quality-gate to run the promotion gate offline with no Google Cloud, API key, or emulator.
- Teams use model-quality-gate to switch between managed, local, and on-prem adapters via a one-line profile change.
- Teams use model-quality-gate to produce versioned MRM evidence as ModelCard and PromptVersion.
Features
AI assistant
Natural language interface
Workflow automation
Reviews
No reviews yet. Be the first to share your experience!