0.5.0

MemoryVault 0.5.0

• defines 1.0 as a local-first memory-learning workbench
• adds onboarding, transfer, refresh, and release benchmark workflows
• adds the first fixed public release benchmark bundle and artifact compatibility markers
• updates the README and repository presentation

v0.4.0

Added

• Python logging-based lifecycle logging with CLI-controlled log level and optional log file output.
• Per-run observability.json artifacts with timings, counts, and summary metrics.
• Per-run wind_tunnel_observability.json artifacts for wind tunnel runs.
• A logging and observability guide in docs/observability.md.
• A release-version sync check that keeps pyproject.toml aligned with the latest released section in Chaneglog.md.
• A platform-neutral integration strategy covering HTTP, MCP, multi-agent use, and caching in docs/integration_strategy.md.

Changed

• The CLI now supports --log-level and --log-file.
• The pipeline and wind tunnel now emit run lifecycle logs and save basic observability data.
• The local quality gate now also checks release-version consistency before commit.
• The project strategy, research summary, and product brief now reflect the planned hybrid integration model.

v0.3.0

Added

• A first local discovery prototype for interrupted-task memory evaluation.
• A small memoryvault/ package with local run storage, candidate memory extraction, resume packet generation, and evaluation logging.
• Built-in sample scenarios for coding and document work.
• A JSON intake path for imported interrupted-task traces.
• Example imported traces modeled after coding and long-memory work.
• A synthetic tool-use trace for domain-agnostic interrupted-task testing.
• A Hugging Face benchmark lead registry for later external evaluation.
• A durable-field suggestion step that highlights repeated misses.
• Automated tests covering the local discovery loop.
• A strategy note describing the tool-first, zero-domain-knowledge development phases.
• A Memory Wind Tunnel that removes memory fields and measures the damage.
• A repo-local quality gate with a pre-commit hook, Python linting, Markdown linting, and a 90% coverage threshold.

Changed

• Replaced the placeholder root entry point with a working CLI.
• Promoted assumptions into the resume packet after repeated misses across scenarios.
• Reframed the project from a domain-specific product direction to a tool-first memory-learning direction.
• Updated project docs and .codex notes to reflect the new prototype state.