v0.8.0

What's New

• Async-safe pipeline state: Replaced threading.local() with contextvars.ContextVar for safe concurrent asyncio usage
• New context managers: GlacisTagScope and GlacisOperationContext with proper token-based cleanup
• Fixed get_evidence(): Now reads storage backend from glacis.yaml config instead of hardcoding SQLite
• Removed online attestation path
• Updated all provider integrations (OpenAI, Anthropic, Gemini, LiteLLM)

Breaking Changes

• GlacisCallContext renamed to GlacisTagScope (deprecated alias available for one release)
• online_attestation_demo and examples/online_openai.py removed

v0.6.0

What's New

Custom Metadata

• All integration wrappers (attested_openai, attested_anthropic, attested_gemini) now accept a metadata parameter for custom key-value pairs included in every attestation

Integration Refactor

• Consolidated ~270 lines of duplicate boilerplate across OpenAI, Anthropic, and Gemini integrations into shared helpers in base.py
• New IntegrationContext dataclass + setup_integration(), check_input_block(), check_output_block(), attest_and_store() helpers
• Single source of truth for the attestation pipeline

Controls & Models

• Staged input/output control pipeline (PII scanning, jailbreak detection, word filter)
• Control Plane Results (CPR) with per-control execution records
• Deterministic L1/L2 sampling with HMAC-based PRF

pip install glacis==0.6.0

v0.1.4

Initial clean release with all features:

• Offline and online attestation modes
• OpenAI and Anthropic integrations
• WASM runtime support
• Cross-runtime hash compatibility
• Async support

v0.1.3

Update pricing table to match website.

v0.1.2

Add missing get_runtime() function.

Changes

• Add get_runtime() convenience function to wasm_runtime module

v0.1.1

Fix lint errors and improve code quality.

Changes

• Remove unused imports
• Fix line length issues
• Sort imports correctly
• Clean up unused variables

v0.1.0

Initial release of GLACIS Python SDK.

Features

• Online mode: Server-witnessed attestations with Merkle proofs
• Offline mode: Local Ed25519 signing (UNVERIFIED status)
• OpenAI and Anthropic integrations with auto-attestation
• SQLite storage for offline receipts
• WASM runtime with PyNaCl fallback
• Streaming session support
• CLI for receipt verification

Installation

pip install glacis