Ragdoc 1.8.0 · Ragdrop 0.6 — Your scientific library. Connected.

Ragdoc 1.8.0 · Ragdrop 0.6

A native PDF review workspace for your scientific library, with an English
interface, light and dark appearances, and a clearer path from import to evidence.

In this release

• Ragdrop for macOS. Import PDFs from Finder or Zotero, review the extraction
  beside the original, inspect Markdown and visual artifacts, then approve items
  for indexing. Queue progress and failed steps remain visible.
• A complete English interface. Navigation, settings, queue messages, review,
  errors and synthetic demos now share consistent English labels. Saved queue
  identifiers, preferences and document metadata retain their existing format.
  Original document text and external diagnostic messages are not translated.
• Light, dark and system appearance. Restrained typography and spacing keep
  reading and review at the center of the workspace.
• Evidence retrieval through MCP. Ragdoc connects compatible assistants to a
  personal article library with hybrid retrieval, canonical snapshots, hashes and
  source context. Provenance does not certify OCR or scientific accuracy.
• A clearer project home. A new banner, real screenshots and English guides
  explain the workflow and service dependencies. Setup guides live under docs/;
  earlier guides and Windows helpers are separated from current onboarding.

Download and requirements

Ragdrop-0.6-macos-arm64.zip contains Ragdrop 0.6, build 26 for Apple silicon,
requiring macOS 14 or later. It is a development build with an ad-hoc signature;
it is not Developer ID signed or notarized, and macOS may block a downloaded
copy. A notarized installer and an Intel binary are not provided in this release.

The archive includes both OCR helper scripts. Real imports also require a
configured SSH-accessible Linux Ragdoc backend, local Python tooling and OCR
credentials. The default SSH host and server directory are placeholders that must
be configured in Settings. See the macOS guide and
backend installation. Source builds require Swift 6.2.

Mistral or MinerU receives selected PDFs for OCR before review. Voyage receives
text and queries for embeddings; optional Cohere reranking receives queries and
candidate passages. The isolated synthetic demo needs none of these services.

SHA256SUMS.txt records the downloadable app archive checksum. GitHub also provides
source archives for the release tag.

Existing installations

Back up your library and queue before upgrading. Ragdrop retains persisted state
keys and displays known earlier French status messages in English without rewriting
stored document text. Review your converter path, SSH host and server directory in
Settings; an existing custom deployment may use different defaults.

For backend changes, follow the migration and recovery guidance
and keep a backup before replacing an index. Do not treat a successful conversion
or approval as proof that indexing and verification have finished.

Validation

• 43 native Swift tests passed, including queue recovery, review, appearance,
  Zotero monitoring and legacy-state compatibility.
• 127 Python unit and retrieval-metric tests passed with service credentials
  removed and external connections blocked.
• Native synthetic demo views checked in light and dark appearances: home,
  review, rendered extraction, source, tables, settings, errors, Zotero and a mixed
  batch. The four public screenshots are native captures of synthetic data.
• The packaged app is built in release mode; converter scripts are bundled and
  the ad-hoc signature is verified. Private compilation paths are excluded from
  the distributed executable.

These checks do not constitute a live OCR/import test or a retrieval-quality
benchmark. No real article was imported for this release validation.