Skip to content

docs(migration): upstream-reader summary + how the port tracks docling releases - #29

Merged
artiz merged 1 commit into
masterfrom
claude/docling-rust-review-profile-d26fdq
Jul 6, 2026
Merged

docs(migration): upstream-reader summary + how the port tracks docling releases#29
artiz merged 1 commit into
masterfrom
claude/docling-rust-review-profile-d26fdq

Conversation

@artiz

@artiz artiz commented Jul 6, 2026

Copy link
Copy Markdown
Contributor

Summary

Prep for the docling-org collaboration sync — MIGRATION.md now works as the technical one-pager for an upstream reader, without adding a new document:

  • "At a glance" table right under the status block: what the port is, conformance and performance in one line each (4.3× warm / 4.7× end-to-end vs Python docling at 2.3–2.6× less RAM, with links to the full methodology), where the models come from (docling's own checkpoints, format-converted to ONNX — no retraining), and what is intentionally not ported.
  • New §9 "Keeping up with upstream docling" — the answer to "continuously updated with the new features": conformance.sh measures against the latest published docling so upstream releases surface as concrete per-fixture diffs; model bumps flow through the existing export → quantize → publish scripts; the snapshot + 133-source regression suites re-gate every sync. Also states what can't be absorbed automatically (VLM/enrichment architectures, documented model differences).
  • Housekeeping: fleischwolf-rag added to the crate tree, "Four layers" wording fixed.

Docs-only change.

🤖 Generated with Claude Code

https://claude.ai/code/session_017fKFbatf55BK5j8vr3dZjj


Generated by Claude Code

…ocess

Ahead of the docling-org collaboration: an 'at a glance' table up top
(what the port is, conformance and performance in one line each, where
the models come from, what is intentionally not ported), fleischwolf-rag
in the crate tree, and a new §9 describing how the port tracks upstream
releases — conformance.sh against the latest PyPI docling surfaces
drift as per-fixture diffs, model bumps re-export/re-quantize/republish
via the existing scripts, and the snapshot + regression suites re-gate
every sync.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017fKFbatf55BK5j8vr3dZjj
@artiz
artiz merged commit 80ec0d7 into master Jul 6, 2026
3 checks passed
@artiz
artiz deleted the claude/docling-rust-review-profile-d26fdq branch July 6, 2026 08:57
artiz pushed a commit to artiz/docling.rs that referenced this pull request Jul 13, 2026
…order fixes

The pdf table/text fixtures pick up the quote normalization and the
docling-parse index-order cell join; several of them (2206, 2305,
normal_4pages, right_to_left_03, llncsdoc, mets_gbs, scanned/*) had been
stale since PR docling-project#29 - the odf panic (fixed two commits ago) aborted every
full snapshot regeneration before it reached them. The scanned/OCR
fixtures were re-pinned from a source-rebuilt model set (models-v1 assets
are unreachable from this environment); if the release-model environment
disagrees on OCR text, re-pin them there once.

Signed-off-by: artiz <artem.kustikov@gmail.com>

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EY5KAiquN4YpVf2PXEQkVT
Signed-off-by: artiz <artem.kustikov@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants