v0.2.0

First curated release on top of the public v0.1.0 baseline. Adopts a
forward-only cleanup model (Scenario 1B): no published commits are ever
rewritten; improvements land as new commits on main.

No functional changes since v0.1.0 — this is a hygiene and policy
baseline release. Application behaviour, MCP tool surface, and deployment
shape are unchanged.

Added

• Repository hygiene baseline: .gitignore, .editorconfig, CHANGELOG.md,
  SECURITY.md, CONTRIBUTING.md, CODE_OF_CONDUCT.md (Contributor
  Covenant 2.1).
• Security policy with private vulnerability reporting via GitHub Security
  Advisories, secondary contact at lf@logosflux.io.
• Committed bun.lock for reproducible installs across contributors.
• Social preview card for GitHub repo metadata.

Security

• Established a coordinated-disclosure baseline and supported-versions
  policy (see SECURITY.md).

v0.1.0 — initial public release

First public release of Mnemo — an MCP server that uses Gemini's context caching as extended memory for AI assistants.

What's in this release

• MCP server exposing tools to load and query Gemini-cached contexts
• Loaders for GitHub repositories, URLs, and PDFs
• Cache lifecycle management (create / refresh / list / delete)
• Designed to pair with Claude (or any MCP-aware assistant) as the orchestrator while Gemini holds the long context

Why use it

Skip RAG-style chunking and embeddings for codebase / docset Q&A — load the whole thing into Gemini's 1M-token cache once, then query it conversationally with perfect recall.

See the README for setup, supported loaders, and example MCP client config.