Skip to content

RELEASE-NOTES-v1.1.x.md

Latest commit

 

History

History
238 lines (168 loc) · 6.09 KB

RELEASE-NOTES-v1.1.x.md

File metadata and controls

238 lines (168 loc) · 6.09 KB
Raw

TealTiger Python SDK v1.1.x Release Notes

Overview

TealTiger SDK v1.1.x introduces enterprise-grade features for organizational adoption while maintaining 100% backwards compatibility with v1.0.x.

Release Date

February 2026

What's New

P0.1: Policy Rollout Modes

Safe policy deployment with three evaluation modes:

  • ENFORCE: Block violations (production mode)
  • MONITOR: Log violations, allow requests (observability mode)
  • REPORT_ONLY: Log decisions without evaluation (testing mode)
from tealtiger.core.engine.types import PolicyMode, ModeConfig

mode_config = ModeConfig(
    default_mode=PolicyMode.MONITOR,  # Start with MONITOR
    environment_modes={'production': PolicyMode.ENFORCE},
    policy_modes={'tools.file_delete': PolicyMode.ENFORCE}
)

engine = TealEngine(policy, mode=mode_config)

Benefits:

  • Test policies without breaking production
  • Gradual rollout (MONITOR → ENFORCE)
  • Environment-specific configuration

P0.2: Deterministic Decision Contract

Stable typed Decision object for reliable flows:

from tealtiger.core.engine.types import Decision, DecisionAction, ReasonCode

decision = engine.evaluate(request_context)

# Deterministic fields
assert decision.action in [DecisionAction.ALLOW, DecisionAction.DENY]
assert len(decision.reason_codes) > 0
assert 0 <= decision.risk_score <= 100
assert decision.correlation_id  # Always present

Benefits:

  • Predictable decision structure
  • Type-safe integration
  • Standardized reason codes

P0.3: Correlation IDs + Traceability

ExecutionContext with auto-generated correlation IDs:

from tealtiger.core.context.context_manager import ContextManager

# Auto-generate correlation ID
context = ContextManager.create_context(
    tenant_id='acme-corp',
    app='customer-support'
)

# Use across all operations
response = await client.chat.completions.create(
    model='gpt-4',
    messages=[...],
    context=context
)

# Query audit logs by correlation ID
events = audit.query(correlation_id=context.correlation_id)

Benefits:

  • End-to-end request tracing
  • Distributed tracing integration (OpenTelemetry)
  • HTTP header propagation

P0.4: Audit Schema + Redaction Guarantees

Versioned audit events with security-by-default redaction:

from tealtiger.core.audit.teal_audit import TealAudit, AuditConfig
from tealtiger.core.audit.types import RedactionLevel

audit = TealAudit(
    outputs=[FileOutput('./logs/audit.log')],
    config=AuditConfig(
        input_redaction=RedactionLevel.HASH,  # SHA-256 hash (default)
        output_redaction=RedactionLevel.HASH,
        detect_pii=True  # Enabled by default
    )
)

Benefits:

  • No raw prompts/responses in logs
  • PII detection and redaction
  • Compliance-ready audit trails
  • Versioned schema (1.0.0)

P0.5: Policy Test Harness

CLI/library test runner for CI/CD integration:

from tealtiger.core.engine.testing.policy_tester import PolicyTester
from tealtiger.core.engine.testing.test_corpora import TestCorpora

# Use starter test corpora
test_suite = TestCorpora.prompt_injection()

# Run tests
tester = PolicyTester(engine)
report = tester.run_suite(test_suite)

# Export to JUnit XML for CI/CD
xml_report = tester.export_report(report, format='junit')

CLI Integration:

python -m tealtiger.cli.test ./policies/*.json --format=junit --output=results.xml

Benefits:

  • Validate policies before deployment
  • 25+ starter test cases
  • CI/CD integration
  • Coverage tracking

Performance Characteristics

All features meet performance targets:

Feature Target Actual (p99)
Mode resolution < 1ms ✅ 0.5ms
Decision evaluation < 10ms ✅ 8ms
Context propagation < 0.5ms ✅ 0.3ms
Content redaction (10KB) < 5ms ✅ 4ms
Audit logging (async) < 2ms ✅ 1.5ms
Policy test execution < 100ms ✅ 50ms

Security Guarantees

✅ No raw prompts/responses in audit logs (HASH redaction by default) ✅ PII detection enabled by default ✅ Security-by-default configuration ✅ Explicit opt-in required for debug mode ✅ SHA-256 hashing for content redaction ✅ Cryptographically random UUID v4 generation

Compliance Alignment

TealTiger v1.1.x aligns with:

  • OWASP Top 10 for Agentic Applications (ASI01-ASI10)
  • NIST AI RMF 1.0 (Govern, Map, Measure, Manage)
  • Google SAIF (Secure AI Framework)

Backwards Compatibility

✅ 100% backwards compatible with v1.0.x ✅ No breaking changes ✅ All new features are opt-in ✅ Existing code continues to work without modification

Migration

See Migration Guide for detailed upgrade instructions.

Quick Start:

# v1.0.x code - still works in v1.1.x
client = TealOpenAI(TealOpenAIConfig(api_key="..."))
response = await client.chat.completions.create(...)

# v1.1.x - add enterprise features incrementally
context = ContextManager.create_context()
response = await client.chat.completions.create(..., context=context)

Examples

See examples/ directory for complete examples:

  • correlation_ids_tracing.py - 9 examples for ExecutionContext
  • audit_redaction.py - 8 examples for redaction levels
  • policy_testing.py - 8 examples for PolicyTester
  • enterprise_integration.py - Complete end-to-end setup

Documentation

Known Issues

None.

Deprecations

None. All v1.0.x APIs remain supported.

Contributors

TealTiger team and community contributors.

License

MIT License

Support

Next Steps

  1. Read the Migration Guide
  2. Review Best Practices
  3. Try the Examples
  4. Integrate with your CI/CD pipeline

Happy securing! 🎉