Skip to main content

​
Overview

TealTiger v1.1.0 is a foundational release focused on deterministic AI governance, policy enforcement, and auditability. This version establishes the core contract between TealTiger and its users:
  • What the system does
  • What it guarantees
  • What it explicitly does not do
All future versions build on the guarantees defined here.

​
Release Intent

The intent of v1.1.0 is to:
  • Establish a stable policy evaluation core
  • Enable security, cost, and governance enforcement
  • Provide enterprise‑grade audit evidence
  • Remain SDK‑first and developer‑friendly
v1.1.0 prioritizes correctness, determinism, and trust over advanced automation.

​
In‑Scope Capabilities

v1.1.0 includes the following capabilities.

​
Policy Evaluation

  • Deterministic condition → action evaluation
  • Logical composition (AND, OR, NOT)
  • Policy modes:
    • monitor
    • enforce

​
Conditions and Actions

  • Input, model, cost, risk, and execution context conditions
  • Enforcement actions:
    • allow
    • restrict
    • block
  • Transformation actions:
    • redact
    • mask
  • Cost control actions:
    • token limits
    • model switching
  • Observability actions:
    • logging
    • metric emission

​
Risk Scoring

  • Normalized risk scores (0–100)
  • Domain scores:
    • security
    • cost
    • governance
    • reliability
  • Composite risk evaluation
  • Policy‑driven risk thresholds

​
Reason Codes

  • Stable, machine‑readable identifiers
  • Multiple reason codes per decision
  • Domain‑based taxonomy
  • Audit‑safe and automation‑friendly

​
Cost Metadata

  • Token usage (input / output / total)
  • Estimated vs actual cost
  • Model and provider attribution
  • Budget and quota context
  • Cost attribution (service, agent, environment)

​
Audit and Telemetry

  • Structured audit events
  • Audit event schema v1.1.0
  • Policy decision events
  • Action execution events
  • Correlation identifiers
  • Redaction metadata

​
Logging Behavior

  • Structured logging
  • Policy‑aware logging
  • Safe‑by‑default (metadata‑only)
  • Deterministic redaction behavior

​
Execution Identity (Foundational)

  • Actor identity (user / service / agent)
  • Environment context
  • Deployment metadata
  • Attribution for audit and cost events

​
Explicitly Out of Scope

The following are intentionally excluded from v1.1.0.
  • Autonomous agent orchestration
  • Agent planning or reasoning loops
  • Runtime policy mutation or learning
  • Advanced model governance (fairness, drift, provenance)
  • Deep content logging by default
  • Hardened multi‑tenant isolation
  • Automated GRC or compliance reporting
  • Performance or latency SLAs
These exclusions are by design, not limitations.

​
Stability Guarantees

Within the v1.1.x line, TealTiger guarantees:
  • Deterministic policy evaluation
  • Stable condition and action semantics
  • Stable risk score ranges and meaning
  • Stable reason code identifiers
  • Backward‑compatible audit schemas
  • Non‑breaking SDK behavior
Bug fixes may improve correctness but will not change behavior.

​
Compatibility Contract

TealTiger follows semantic versioning.
  • Patch (v1.1.x)
    Bug fixes only, no breaking changes
  • Minor (v1.x)
    Additive, backward‑compatible changes
  • Major (v2.0+)
    Breaking changes allowed with migration guidance
All guarantees in this document apply only within v1.1.x.

​
Audit Defensibility

v1.1.0 guarantees that:
  • Policy decisions are reproducible
  • Inputs to decisions are traceable
  • Enforcement outcomes are explainable
  • Risk and cost signals are attributable
  • Historical decisions remain defensible
This enables regulatory, legal, and internal audits.

​
Non‑Goals

v1.1.0 does not attempt to:
  • Replace application logic
  • Act as an AI runtime or agent framework
  • Automatically remediate unsafe behavior
  • Optimize business outcomes autonomously
Its role is control, visibility, and governance.

​
Forward Compatibility

v1.1.0 intentionally lays the groundwork for:
  • v1.2.x — governance expansion
  • v1.3.x — advanced cost, reliability, and agent governance
  • Runtime and platform layers beyond SDK‑only usage
No v1.1.0 guarantees will be silently weakened in future versions.

​
Release Status

v1.1.0 is frozen. All changes beyond this point must:
  • Be backward compatible, or
  • Target a future version explicitly
This document is the authoritative reference for v1.1.0 behavior.