Skip to content

CHANGELOG.md

Latest commit

 

History

History
100 lines (78 loc) · 5.35 KB

CHANGELOG.md

File metadata and controls

100 lines (78 loc) · 5.35 KB
Raw

Changelog

All notable changes to meta-pipe are documented in this file.

[Unreleased]

Fixed

  • Replace deprecated datetime.utcnow() with timezone-aware datetime.now(timezone.utc) across 13 files (29 call sites). Output string shape preserved (...Z suffix, no +00:00). Eliminates all pytest DeprecationWarnings
  • tests/test_project_status.py::test_all_stages_complete — update fixture to create 04_fulltext/manifest.csv matching the current validation lambda (was creating round-01/unpaywall_results.csv from an earlier schema)

Added

  • Sprint 3 of pipeline design fixes (#38)
    • session_log.py append subcommand — writes append-only provenance stamps to 09_qa/sessions/artifact_stamps.jsonl (who/stage/deviations) and attaches to the active session when one exists
    • ma-end-to-end/references/artifact-stamping.md — documents the stage-exit provenance convention
    • Stage-exit stamping snippets wired into ma-screening-quality, ma-data-extraction, ma-meta-analysis, and ma-manuscript-quarto SKILL files
  • Sprint 2 of pipeline design fixes (#38)
    • init_project.py: --mode {strict,draft} flag. Draft mode writes .ma_meta.json with quality_mode: draft and a DRAFT_MODE.md notice so fast-prototype runs are explicit and non-publishable
    • tooling/python/project_meta.py: shared helper that reads .ma_meta.json (single source of truth for project quality mode)
    • validate_pipeline.py: reads quality mode and reports unchecked items as NOTES (exit 0) in draft mode, FAILURES (exit 2) in strict mode; added --json output for programmatic consumers
    • ma-data-extraction/templates/prognostic_factor.yaml + .md: 19-field 2×2 extraction template for binary exposure → binary outcome studies. Plugs directly into create_extraction_template.py
  • Sprint 1 of pipeline design fixes (#38)
    • ma-search-bibliography/scripts/enrich_abstracts.py: new abstract enrichment stage (Entrez → CrossRef → OpenAlex fallbacks) so dedupe.bib records reach screening with abstracts populated
    • tooling/python/CLAUDE_CLI_FLAGS.md: documents required claude CLI flags, minimum version (2.1.100), and the --bare + ANTHROPIC_API_KEY auth interaction
    • scopus_fetch.py: reports opensearch:totalResults, warns on silent truncation, adds --strict-cap flag
    • ai_screen.py: startup _assert_claude_cli() check verifies --bare and --output-format flags exist

Changed

  • ai_screen.py: _invoke_claude() now uses claude -p --bare --output-format json when ANTHROPIC_API_KEY is set, cutting per-call input tokens from ~10k → ~1.5k; falls back to non-bare with a one-time warning when only OAuth is available
  • ai_screen.py: META_PIPE_ROOT resolved from $MA_PIPE_ROOT or module-relative path instead of hardcoded /Users/htlin/meta-pipe
  • tooling/python/pyproject.toml: pinned rapidfuzz, lxml (biopython, bibtexparser, requests were already present); ran uv lock

Fixed

  • Fix hooks schema in .claude/settings.local.json (matcher + hooks array structure)

[2026-03-24]

Added

  • Agent teams orchestration framework for parallel meta-analysis (#28)
    • Hook scripts for TaskCompleted and TeammateIdle events
    • Team spawn helper (tooling/python/team_spawn_helper.py)
    • Role-specific prompts in ma-agent-teams/prompts/
    • G-CSF neutropenia NMA test project for agent teams validation
  • Academic and Non-Commercial Use License (English + Chinese)

Fixed

  • Verification script bugs (#26)
  • setup.sh fails on macOS ARM: R package 'fs' requires cmake (#24)
  • Missing file noted when creating new projects (#22)

[2026-03-22]

Added

  • Peters' test for binary outcomes and low-power caveat in bias analysis
  • Semi-automated GRADE with computed suggestions and rationale (#16)
  • Stage 04b full-text eligibility screening (PRISMA item 16)
  • CITATION.cff with Lin HT as first author (#14)
  • Manuscript .gitignore and Vancouver citation style

[2026-03-21]

Added

  • Manuscript scaffold, dashboard plan, and Python test suite
  • Publish workflow to high-IF journal article (#9)

Fixed

  • League table heatmap not included (#3)

[2026-03-19]

Added

  • .gitignore for ici-breast-cancer project
  • Script to create GitHub private repositories

[2026-02-17]

Added

  • Phase 2 enhancements (AI automation 95-98%)
    • publication_readiness_score.py — objective 0-100% score
    • validate_nma_outputs.py — NMA-specific validation (7 checks)
    • Enhanced claim_audit.py — overclaim detection (12 patterns)
    • nma-completion-checklist.md — 25-item pre-submission checklist

[2026-02-08]

Changed

  • Migrated all projects to projects/<project-name>/ structure
  • Legacy data moved to projects/legacy/

[2026-02-06]

Added

  • Initial release of meta-pipe
  • Complete meta-analysis pipeline: 10 stages from topic intake to submission
  • Skills: ma-topic-intake, ma-search-bibliography, ma-screening-quality, ma-fulltext-management, ma-data-extraction, ma-meta-analysis, ma-network-meta-analysis, ma-manuscript-quarto, ma-peer-review, ma-publication-quality
  • End-to-end orchestration (ma-end-to-end)
  • Module registry validation and management
  • LLM-assisted data extraction tools (Stage 05)
  • PROSPERO protocol generator, RoB 2 and ROBINS-I assessment tools
  • Generalized PDF retrieval tools for systematic reviews
  • Example project: ICI in triple-negative breast cancer (5 RCTs, N=2,402)