fix(ocr): file page headers and footers under the furniture content layer - #252

Open

JonnyTran wants to merge 16 commits into

mainfrom

feat/retrieval-hybrid-search

Commits on Aug 26, 2026

build(server): add liteparse + chonkie, bake the retrieval payloads into the image
JonnyTran
committed
30a08df
build(server): fetch the lance extension instead of INSTALL-ing it
JonnyTran
committed
99ec6a2

Commits on Aug 27, 2026

build(server): split the builder stage and apply the uv Docker guide
JonnyTran
committed
b8b3948
feat(ocr): read `items` rows back as typed elements, with a liftable table header
JonnyTran
committed
a25b574
build(server): pull the lance fetch out of the per-build path, bump uv, drop the conda vars
JonnyTran
committed
9db0f0f
fix(server): verify the lance extension over TLS against a pinned digest
JonnyTran
committed
6c23323
fix(ocr): keep a table's caption in the table element
JonnyTran
committed
59ba87d
build: advance the hf-space submodule, and keep --seed after all
JonnyTran
committed
b282657

Commits on Aug 31, 2026

refactor(ocr): project elements columnar, as one DuckDB statement
JonnyTran
committed
c6568d3

Commits on Sep 3, 2026

refactor(ocr): delete the docling reimplementations, render elements with docling
JonnyTran
committed
5af997b

Commits on Sep 4, 2026

Commits on Sep 7, 2026

refactor(ocr): unwind the element layer; docling's chunker owns it
JonnyTran
committed
2406ed8

Commits on Sep 9, 2026