Skip to content

DWIN Research Adoption v0.1.0

Latest

Choose a tag to compare

@vidvinianinov vidvinianinov released this 07 Sep 22:26

First public release

DWIN Research Adoption turns version-pinned AI research into reviewable local setup experiments without silently promoting paper text into durable memory.

Distribution

  • npm: dwin-research-adoption@0.1.0
  • MCP Registry: io.github.vidvinianinov/dwin-research-adoption@0.1.0
  • skills.sh: vidvinianinov/dwin-research-adoption
  • runtime: Node.js 22 or newer, local stdio transport

Release evidence

  • 6 governed capsules behind one public MCP entrypoint
  • 27 tools in the default profile
  • 68/68 automated tests passed
  • clean npm-registry install returned the expected 27-tool surface
  • 5 positive and 3 negative marketplace scenarios passed
  • release audit found 0 personal-path, credential, employer-artifact, symlink, or metadata-drift findings
  • source-tree receipt: 7c08a4f6710954eb187e1e251a5d08f875a8d245cdb5697960091a0000deef18

Reproduce

git clone --branch v0.1.0 https://github.com/vidvinianinov/dwin-research-adoption.git
cd dwin-research-adoption
npm ci
npm run verify
npm run demo

The offline demo discovers the unified surface, routes an AI-memory question, compiles an inspectable arXiv query, searches a sanitized fixture, and reports durable_change: false.

Boundaries and limitations

  • This is a local single-user stdio release, not a hosted multi-tenant service.
  • Default retrieval is lexical and source-cited. Local embeddings are optional and experimental.
  • Paper text, PDF extraction, and graph edges are candidates rather than trusted instructions or scientific truth.
  • Durable memory requires an accepted evaluation record plus explicit human approval.
  • The demo validates deterministic wiring and safety boundaries; it is not yet evidence of user-level productivity improvement.

Rollback

Remove the MCP entry from the client configuration and uninstall dwin-research-adoption. Local application data is not silently deleted; inspect, export, or delete it explicitly according to your client and operating-system policy.