Skip to content

Changelog

Full changelog: CHANGELOG.md on GitHub

FactorForge v3.0+ is the current public release line.

v3.3.2 — 2026-07-07

Adds structured cai_authority metadata to CAI-bearing outputs (labels the CAI reference relationship without changing CAI values or formulas), a benchmark runner overwrite guard, and fixes the profile-engine CLI's --output/--reference-id output contract so it's self-describing and consumable by downstream tooling.

v3.3.1 — 2026-07-03

Added

  • Exposed the DP feasibility_best engine's CAI target as a configurable CLI/API parameter (--cai-target / constraints.cai_target), alongside the existing GC-range controls.
  • Added mfe_status_reason and a checksum-validated --reference-id expert CLI flag.

Fixed

  • Fixed single-sequence profile-engine --output FASTA files to include the standard metadata header instead of writing only the raw sequence.

v3.3.0 — 2026-07-01

Changed

  • Activated NbeV1.1 high-confidence CDS-derived codon usage as the current N. benthamiana software default with a 40-47% GC policy band. Public reference overrides remain disabled; the legacy Kazusa/SGN composite is retained as a historical comparator. This is an in-silico software-default change only, not experimental validation or comparative biological-performance evidence.

v3.2.8 — 2026-06-29

Reverted

  • Default N. benthamiana GC reference band / codon-reference table reverted to legacy (55-65% GC), undoing a default-codon-reference promotion that shipped in v3.2.7. Newer candidate/reference-comparator assets remain packaged for provenance and research comparison, but are not public production defaults or selectable UI options pending an MFE re-sensitivity + 2x2 factorial recheck

v3.2.7 — 2026-06-29

Changed

  • Recalibrated the MFE-score normalization clamp from [-0.5, 0.0] to [-0.40, -0.15] based on a measured MFE distribution (n=135) showing the wider clamp was range-compressing scoring discriminability

Fixed

  • calculate_composite_score() now raises on an unrecognized profile name instead of silently falling back to balanced; ScoringConfig rejects negative weights and gc_min > gc_max (library-direct callers only — no product-path behavior change)
  • Guarded calculate_mfe() against unbounded ViennaRNA RNA.fold() runtime on long sequences (algorithmic-complexity DoS); sequences over 1000nt now skip MFE folding instead of risking an indefinite worker hang (correction: this also shipped in v3.2.7 but was undocumented here at release time)

Added

  • Benchmark runs now record whether ViennaRNA bindings were active, going forward

v3.2.6 — 2026-06-27

Changed

  • Aligned the public-safe wet-lab feedback intake surface with the actual GitHub issue template, removing fields that were never collected
  • Disabled blank GitHub issues to reduce the chance of sensitive wet-lab/construct details landing in an untemplated public issue

v3.2.5 — 2026-06-25

Fixed

  • The published Docker image's POST /api/optimize crashed with an AttributeError due to a routing bug in the local dev/Docker server (hosted web app and PyPI CLI were unaffected)
  • CITATION.cff's doi field corrected to the v3.2.4 exact-release DOI

v3.2.4 — 2026-06-24

Added

  • Runtime validation registry exposing all 17 sequence checks (9 advisory scanners + restriction-site + MoClo-overhang) via the web "Sequence Checks" panel and /api/optimize response
  • Explicit consent modal before sending sequences to AlphaFold DB / ESM Atlas structure-prediction links

Fixed

  • Web UI "AA Preserved" badge showed assembly/restriction-site review status instead of actual amino acid identity
  • /api/optimize/compare and /api/optimize/batch now reject host/host_profile fields with HTTP 400 instead of silently dropping them
  • high_cai profile and feasibility_best objective are now consistently disclosed/rejected for non-default hosts across CLI, REST, and web UI

v3.2.3 — 2026-06-19

Fixed

  • Release provenance hashing now computed from the committed git blob instead of local working-tree bytes, fixing CRLF/LF drift on Windows that could silently produce incorrect SHA-256 values in reproducibility manifests
  • Public-surface DOI references switched from version-pinned Zenodo DOIs to the concept DOI, which always resolves to the latest release

Added

  • A public-surface audit and a CHANGELOG consistency check now run on every push/PR in CI

v3.2.2 — 2026-06-18

Fixed

  • multi_constraint_pass definition corrected (scoring_contract v1.1) — now requires GC-in-target-range in addition to biological/assembly pass; corrected ablation values (L3=3.5%, L4=5.6%) supersede the previous inflated figures. All benchmark artifacts regenerated (seed=320, N=49,257); Zenodo benchmark_results.csv v2 (DOI: 10.5281/zenodo.20676276) supersedes v1
  • Benchmark: source-profile codon-table injection now flows into both the design and scoring paths, fixing a prior gap where design always used the default table regardless of an injected profile

Added

  • Data: three genome-annotated N. benthamiana codon-usage profiles (SGN QLD183 v103 CDS-derived; SGN NbeV1.1 all-CDS-derived; SGN NbeV1.1 high-confidence-CDS-derived), alongside the existing packaged reference profile

Changed

  • Web: Host System cards now render dynamically from the API's supported_hosts/host_metadata instead of being hardcoded

Docs

  • Aligned public README/docs/web/citation/packaging/roadmap/benchmark wording with the in-silico CDS design claim boundary
  • Clarified that public GitHub Issues must not contain raw sequences, confidential construct details, or other sensitive data

v3.2.1 — 2026-06-16

Added

  • Protein risk annotation — transmembrane helix (Kyte-Doolittle, window=19) and signal peptide heuristics flag HIGH / MEDIUM / LOW / UNKNOWN risk sequences before wet-lab submission

Fixed

  • CAI provenance annotation in benchmark output
  • Type IIS restriction site warning status
  • Manifest SHA-256 reproducibility drift on Windows
  • wet-lab result GitHub template: protein_class options + validation consent checkbox

v3.2.0 — 2026-06-11

Benchmark

  • Formal benchmark — N. benthamiana SGN CDS dataset (N=49,257 sequences), seed=320 for deterministic reproduction. All metrics are in-silico; no wet-lab validation is claimed.
  • Seed injection — benchmark runner now accepts --seed for fully deterministic reruns; most_frequent_codon tie-breaking deduplication included

Data / Provenance

  • Codon table provenance disclosure — nbenthamiana_codons.json origin documented as a legacy Kazusa + SGN v1.0.1-era reference; codon_table_manifest.json added with sha256 pin, build_path_status: incomplete, and known limitations recorded

API / Schema

  • Design Package schema v1.0.0 — formal IUPAC/FASTA I/O contracts and MFE null invariant established
  • Registry constants export — DEFAULT_CAI_TARGET, DEFAULT_GC_LOW, DEFAULT_GC_HIGH importable as public production constants

Fixed

  • Input validator — IUPAC ambiguous DNA/AA sequence misclassification corrected

Documentation

  • Benchmark Trust Pack — registry/spec integrity tests, metric correctness tests, and deterministic baseline runner added
  • Claim wording alignment — public-facing API and CLI output wording unified; no expression-level or yield improvement claims

v3.1.9 — 2026-06-04

Documentation

  • Internal housekeeping: project tracking references updated. No engine changes.

v3.1.8 — 2026-06-01

Breaking

  • gc_target default — now targets host GC midpoint (~60%) instead of legacy 42.5%

Changed

  • GC scoring replaced with band function; assembly_friendly weights updated
  • feasibility.py defaults realigned to 55–65% output distribution

Fixed

  • CAI reference unified to golden-set weights throughout
  • Duplicate root data/ files removed

v3.1.7 — 2026-05-31

Added

  • Host selector UI — N. benthamiana / BY-2 Experimental toggle; BY-2 disables Feasibility Best
  • E2E smoke tests — 5 Playwright tests auto-run after each deployment
  • Eijex MCP — MCP-compatible access via mcp.eijex.com

v3.1.6 — 2026-05-30

Added

  • BY-2 host (experimental) — --host by2, N. tabacum Kazusa codon table
  • Profile comparison — --compare-profiles CLI + POST /api/optimize/compare
  • Batch API — POST /api/optimize/batch (최대 20개 서열)
  • Structure links — AlphaFold DB + ESM Atlas 링크 결과 화면에 추가
  • Experimental scoring hooks — non-public optional scoring hooks; public profiles unchanged
  • GFP Tutorial — docs/tutorials/gfp-nbenthamiana.md

Fixed

  • API GC 기본값 CLI와 일치 (40-55% → 55-65%)

v3.1.5 — 2026-05-28

Fixed

  • Data file packaging — PyPI wheel에 JSON 데이터 파일 미포함 버그 수정 (pip install 후 경로 오류)
  • Path resolution — parents[N] 하드코딩 → get_data_path() 통일 (pip/Docker/Vercel 모두 정상 작동)
  • CI matrix — Python 3.13 + Windows 테스트 추가

v3.1.4 — 2026-05-27

Added

  • CITATION.cff — GitHub "Cite this repository" button; version-bumped automatically
  • SECURITY.md — vulnerability reporting policy
  • ROADMAP.md — public development direction, validation scope, and planned host/profile work
  • scripts/release.py — automates version string updates

Changed

  • Public engine naming — factorforge.engines.profile / CLI --engine profile formalized
  • Benchmark corrected — CAI 0.76, GC% 59.77%, target 55–65% (N=3,876 SGN CDS, balanced profile)

Fixed

  • Stale test version — test_design_package.py product_version 3.1.1 → 3.1.3

v3.1.3 — 2026-05-26

Fixed

  • Disabled profile cards — tooltips now accessible; "Pending wet-lab validation" notice added to 5' Ramp and Viral Delivery
  • Viral Delivery tooltip — corrected citation reference; updated to Peccoud et al. 2024 (PMC11718241)
  • Analytics notice — clarified: "Submitted sequences are not logged or stored"

v3.1.2 — 2026-05-26

Fixed

  • viral_delivery scoring — corrected citation reference; w_mfe 0.40→0.30 per PMC11718241 (Peccoud et al. 2024)
  • 5' Ramp deoptimization — N-terminal deoptimization bottom 50%→25% (mild, per PMC11718241)

v3.1.1 — 2026-05-24

Added

  • Wet-lab feedback link — Submit result button opens a public-safe feedback path
  • JSON Copy button — one-click copy of full optimization JSON output

Changed

  • Design Objective order — reordered to match recommended wet-lab testing sequence
  • Validation fields — Issue template updated for structured public-safe feedback

Fixed

  • Vercel deployment — resolved /api/optimize 404 caused by incorrect Root Directory setting

v3.1.0 — 2026-05-24

Added

  • Custom restriction site removal — synonymous substitution of user-specified restriction sites
  • Rare codon run detection — detects consecutive rare codons for ribosome stalling risk
  • Dinucleotide reduction — CAI-budgeted CpG/TpA reduction (aggressive / balanced / cai_preserving modes)

Fixed

  • Pipeline metric accuracy — CAI/GC/score re-measured from final sequence after dinucleotide fix
  • CAI guard weight consistency — guard now uses Sharp & Li 1987 golden set reference weights

v3.0.0 — 2026-05-23

First official release. DP Feasibility Engine, CLI, Validator, open-source under AGPL-3.0.