v0.4.0

[0.4.0] — 2026-11-03

Fourth versioned state: MINOR feature batch (v0.4.0 batch) — the
reading pipeline §1b in full (RFP-D1-3, issue #4111), multi-course
intake, exam countdowns, and .ics export (RFP-D1-4, issue #4112),
and the memory-methods expansion (RFP-D1-5, issue #4113).

Provider credit: command-dispatch (Swarm Optimizations &
Implementation Assurance seat), session
6628d41d-9519-4881-9063-3afe501df815.

Added

• Reading pipeline (study_pipeline.py): the six §1b stages —
  edition finder over legal open sources (OpenLibrary, archive.org)
  with a student-confirm step before anything plans from an edition;
  upload intake in any form (text native, photos/scans via OCR,
  PDF/EPUB via an injected structured extractor); structure
  extraction (chapters -> sections) mapped onto the reader core's
  Section type; per-chapter Index+Glossary term bank implementing
  the reader's term-lookup hook and feeding SR cards; term emphasis
  segmentation + tap payloads; hypothes.is annotations as W3C Web
  Annotation JSON-LD with an API client (the API-only option from
  the #3697 evaluation; selections convert to SR cards and flagged
  terms to annotations). All network/OCR touches are injected
  callables — the module is offline by default.
• Tests: tests/test_study_pipeline.py covers every public
  function in study_pipeline.py.
• Multi-course intake (study_courses.py): N syllabus documents ->
  N plans (via the existing plan builder) in one stamped bundle
  with merged totals and deadline conflict flags (same-week exams;
  exams within 7 days).
• Exam countdown per course: from the syllabus exam_date —
  days/weeks remaining, urgency tier, the implied study pace
  (sessions per day), and countdown milestones (T-14, T-7, T-3,
  exam day).
• .ics export: every plan session plus countdown milestones and
  exam days as iCalendar VEVENTs (CRLF, stable UIDs, RFC 5545
  escaping), with a parse-back validator (validate_ics).
• Tests: tests/test_study_courses.py covers every public
  function in study_courses.py.
• Memory methods (study_memory.py), all six RFP-D1-5 items, each
  carrying its evidence grade and claims ceiling in code, receipts,
  and FEATURES: (1) method of loci — chapters -> palace route,
  term-bank items at loci, walk-mode review of ordered recall
  (ceiling: ordered/concrete recall, not understanding);
  (2) Pomodoro session engine — 25/5 blocks x4 + long break,
  check-in = block completion, adaptive block length (ceiling:
  pacing/starting, not memory); (3) retrievability notifications —
  fire at the FSRS threshold with the cue + one-tap grade in the
  reminder, implementation-intention phrasing; (4) Feynman
  explain-back graded against verified evidence only (abstain with
  no evidence, escalate on thin coverage); (5) interleaved
  quiz/card sessions (deterministic rotation, mixed topics);
  (6) Leitner boxes — the plain-language onboarding scheduler,
  receipt-stamped.
• Tests: tests/test_study_memory.py covers every public
  function in study_memory.py.