feat: remove H1 match pipeline, use TOC leaf-based shard planning - #203

Merged

EricNGOntos merged 30 commits into

mainfrom

feat/wuchengke/remove-h1-match

Jul 7, 2026

Commits on Jun 29, 2026

feat: implement node assembler for page memory hierarchy to support shared page content and VLM-based node summarization
EricNGOntos
committed
6773fe3
feat: implement page asset integration into node assembly and chunk connections with support for custom probes
EricNGOntos
committed
0035b4a
feat: enhance document profiling and page memory integration with skip shard plan option
EricNGOntos
committed
6d7042c
feat: enhance retrieval and document ingestion with new chunk types and effective parse track handling
EricNGOntos
committed
a0ab8ca
refactor: decouple page memory from image URIs and simplify node metadata structure
EricNGOntos
committed
c8f88ce
feat: integrate summary engine for enhanced document parsing and asset handling
EricNGOntos
committed
3bf5d69
feat: implement concurrent page tagging and enrich image and table asset metadata with LLM-extracted entities and titles
EricNGOntos
committed
eea74df
feat: standardize page_memory track for PDF/PPTX, implement page-count limits, and refactor section summary publication to initialization.
EricNGOntos
committed
dfc10f1
chore: align parser env defaults
EricNGOntos
committed
5dbd084
refactor: replace legacy data_type integer with flexible chunk_types set across retrieval services
EricNGOntos
committed
3531581
feat: implement intermediate node chain collapse and add utilities for cross-page table continuity analysis
EricNGOntos
committed
10e6ba8
Add thread-safe trace logging, remove unused heading dataframes, and implement a text-only tagging mode for page memory processing.
EricNGOntos
committed
e11a41f
refactor: remove budget tracking dependency from page memory services
EricNGOntos
committed
ab73cc7
chore: fix lint and type errors caught by ruff and pyright
EricNGOntos
committed
48ce6ea
fix: resolve alembic revision ID conflict causing cycle detection failure
EricNGOntos
committed
dd0a25a
test: fix failing tests from page-memory native hierarchy branch
EricNGOntos
committed
81fe498

Commits on Jul 1, 2026

Commits on Jul 2, 2026

Commits on Jul 3, 2026

fix: resolve PR #200 lint and typecheck failures

EricNGOntos
and
cursoragent
committed
a42b189
feat: add toc_page_offset to PageAnatomyMap and AgentBlackboard

EricNGOntos
and
cursoragent
committed
411c115
feat: prune out-of-scope TOC nodes before offset-guided anchoring

EricNGOntos
and
claude
committed
625bd9a
feat: add chapter boundary detection and toc_chapter_boundary anchor type

EricNGOntos
and
claude
committed
f870572
Merge remote-tracking branch 'origin/main' into feat/wuchengke/remove-h1-match
EricNGOntos
committed
8a72bb7
fix: use private _calibrate_offset_via_vlm import after merge rename

EricNGOntos
and
claude
committed
f583eaf

Commits on Jul 7, 2026

feat: restore use_agentic switch with classic top-K retrieval route

EricNGOntos
and
claude
committed
00d944f
fix: update contract test to expect classic_topk when use_agentic=False

EricNGOntos
and
claude
committed
5af9c63

Navigation Menu

Uh oh!

feat: remove H1 match pipeline, use TOC leaf-based shard planning - #203

feat: remove H1 match pipeline, use TOC leaf-based shard planning#203
EricNGOntos merged 30 commits into
mainfrom
feat/wuchengke/remove-h1-match

Commits on Jun 29, 2026

feat: implement node assembler for page memory hierarchy to support shared page content and VLM-based node summarization

feat: implement page asset integration into node assembly and chunk connections with support for custom probes

feat: enhance document profiling and page memory integration with skip shard plan option

feat: enhance retrieval and document ingestion with new chunk types and effective parse track handling

refactor: decouple page memory from image URIs and simplify node metadata structure

feat: integrate summary engine for enhanced document parsing and asset handling

feat: implement concurrent page tagging and enrich image and table asset metadata with LLM-extracted entities and titles

feat: standardize page_memory track for PDF/PPTX, implement page-count limits, and refactor section summary publication to initialization.

chore: align parser env defaults

refactor: replace legacy data_type integer with flexible chunk_types set across retrieval services

feat: implement intermediate node chain collapse and add utilities for cross-page table continuity analysis

Add thread-safe trace logging, remove unused heading dataframes, and implement a text-only tagging mode for page memory processing.

refactor: remove budget tracking dependency from page memory services

chore: fix lint and type errors caught by ruff and pyright

fix: resolve alembic revision ID conflict causing cycle detection failure

test: fix failing tests from page-memory native hierarchy branch

Commits on Jul 1, 2026

refactor: remove page_memory_work_summary_20260624.md file and streamline page memory processing documentation

Merge remote-tracking branch 'origin/feat/wuchengke/page-memory-native-hierarchy' into feat/wuchengke/page-memory-native-hierarchy

Commits on Jul 2, 2026

feat: enhance skeleton extractor with native hierarchy support and agentic navigation improvements

Merge remote-tracking branch 'origin/main' into feat/wuchengke/page-memory-native-hierarchy

fix: remove dead commented code and propagate PageLocateConfig to pending TOC fallback

refactor: remove PageLocateResidualAgent fallback from skeleton extractor

Commits on Jul 3, 2026

fix: resolve PR #200 lint and typecheck failures

feat: add toc_page_offset to PageAnatomyMap and AgentBlackboard

feat: prune out-of-scope TOC nodes before offset-guided anchoring

feat: add chapter boundary detection and toc_chapter_boundary anchor type

Merge remote-tracking branch 'origin/main' into feat/wuchengke/remove-h1-match

fix: use private _calibrate_offset_via_vlm import after merge rename

Commits on Jul 7, 2026

feat: restore use_agentic switch with classic top-K retrieval route

fix: update contract test to expect classic_topk when use_agentic=False