v0.2.1 - Security Patches & Integrations

What's Changed

• fix: patch CVSS vulnerabilities and migrate to postcard by @CodNoob100 in #27
• ci: fix gh cli authentication token by @CodNoob100 in #28
• security: update dependencies to latest secure versions by @CodNoob100 in #77
• ci: automate security vulnerability tracking by @CodNoob100 in #78

Full Changelog: v0.2.0...v0.2.1

v0.2.0 - Multi-Tenant Architecture & Native Integrations

[0.2.0] - 2026-06-11

Added

• Multi-Tenant Architecture: Converted the singleton DaemonState to a moka::future::Cache storing isolated ThreadState environments, scaling to thousands of concurrent agents on a single daemon.
• LangGraph & CrewAI Native Integrations: Added nulldrift_agents.py containing drop-in SDK adapters (NullDriftCrewStorage, NullDriftLangGraphStore) to provide seamless memory sync for popular AI frameworks.
• Time-To-Idle (TTI) Disk Paging: Inactive ThreadState instances are automatically swapped to disk (state_{thread_id}.nd) after 24 hours of inactivity to save RAM, with instant on-demand hydration.
• Deterministic Projector: The global 15MB 10,000D projection matrix is now seeded deterministically via rand_pcg::Pcg64, guaranteeing survival across daemon reboots.
• Docker Compose Test Architecture: Created docker-compose.test.yml for fully isolated native ML framework integration testing.

Changed

• Strict $O(1)$ Memory Fix: Completely purged step_history: Vec<String> from the core ThreadState. The daemon now strictly guarantees $< 50$ KB memory limits per thread regardless of injection count.
• Updated all API routes (/inject, /recall, /snapshot, /restore) to use the ?thread_id= parameter.

Fixed

• Fixed incompatible bincode::deserialize library mismatch when loading the phase space from disk during /restore.

v0.1.0 - Initial Beta Release

Added

• SECURITY.md, CONTRIBUTING.md, and CHANGELOG.md to establish community guidelines and open-source standards.
• High-concurrency demo.gif to the README demonstrating the performance of the pure Rust daemon.
• $O(1)$ memory scaling benchmark graph (benchmark_memory.png) comparing null-drift against standard VectorDBs like Chroma and pgvector.
• Integration tests (docker_load_test.py and load_test.py) for stress testing the Axum memory daemon and FastAPI gateway.
• Initial release of the Holographic Reversible State Accumulator (HRSA).
• tokio::sync::RwLock implementation for lock-free, highly concurrent memory state reads.
• Bipolar phase space projection using a Gaussian Random Matrix ($W_{proj}$).
• Geometric decay implementation for temporal permutations (fading noise).
• Attractor Memory Network (AMN) associative cleanup for high-salience anchors.
• Microsecond zero-loss binary state serialization using bincode (state.nd).
• Dockerized FastAPI Python ML Gateway using sentence-transformers (all-MiniLM-L6-v2) to act as the embedding ingestion layer.