feat(doc-agent): Modular Document Anatomy Agent — bootstrap, planner, executor, persist - #108

Merged

EricNGOntos merged 11 commits into

feat/wuchengke/dev

May 28, 2026

Commits on May 28, 2026

feat: add DocumentAnatomyAgent and PDF shard splitter
EricNGOntos
committed
c9de1d0
refactor: remove Phase 1 shard planning legacy
EricNGOntos
committed
9f8a880
refactor: streamline DocumentAnatomyAgent by removing legacy components
EricNGOntos
committed
a487bc7
refactor: update DocumentAgent to use boundary candidates and streamline transitions
EricNGOntos
committed
a5950c3
feat: implement H1-to-page matching tools and integrate TOC hierarchies into document agent state.
EricNGOntos
committed
bf519da
refactor: restructure document agent into modular bootstrap, executor, planner, and persist components
EricNGOntos
committed
e5d8d7e
fix(doc-agent): resolve pyright type error in inspect_pages.py setdefault call
EricNGOntos
committed
4d77026
feat: integrate budget-aware VLM token usage tracking and improve shard planning resilience with validation fallback logic.
EricNGOntos
committed
838167c
refactor: improve heading LLM prompt accuracy with strict TOC rules and implement PDF shard splitting and merging logic.
EricNGOntos
committed
f39b9ba
refactor: remove page numbers from assets and optimize heading detection logic for markdown parsing
EricNGOntos
committed
ccbe4ba
fix: resolve heading detection NEG condition inconsistencies and clean up stale pipeline code
EricNGOntos
committed
f34958a