Skip to content

roadmap.md

Latest commit

 

History

History
218 lines (174 loc) · 10.7 KB

roadmap.md

File metadata and controls

218 lines (174 loc) · 10.7 KB

Development Roadmap

This roadmap documents the evolution of the Research Project Template infrastructure. Architecture details: architecture.md and workflow.md.

Last verified: 2026-06-13 (package and latest published release at v3.4.0). Measured metrics defer to TO-DO.md and docs/_generated/COUNTS.md unless this file is re-measured.

Completed Releases

v3.4.0 — Public Template Scope and Release Baseline (2026-06-12)

  • published the v3.4.0 release and resolved the prior pending-tag gap
  • rebaselined public exemplar scope to nine tracked templates under projects/templates/
  • closed the v3.3 follow-up backlog sweep and moved live measured facts into generated docs and TO-DO.md
  • see CHANGELOG.md for the full entry

v3.3.1 — DOCX Completion and Reproducible Exemplar Outputs (2026-06-07)

  • completed Pandoc DOCX rendering so output embeds figures and resolves cross-references (infrastructure/rendering/pipeline.py)
  • ran a deep per-exemplar quality pass across the eight tracked public templates and completed + cross-linked sidecar publication metadata for all nine
  • reconciled the generated project-scope collection count in docs/_generated/COUNTS.md to 216
  • committed refreshed rendered output/ artifacts for the public exemplars alongside source so the repo ships reproducible, inspectable deliverables
  • see CHANGELOG.md for the full entry

v3.3.0 — Deterministic Evidence and Anti-Hallucination Infrastructure (2026-06-07)

  • added reference-existence verification (infrastructure/reference/verification) — a deterministic anti-hallucination gate resolving each cited reference against Crossref → OpenAlex / arXiv (offline-first, opt-in live, SQLite cache)
  • added an AI-writing fingerprint detector (infrastructure/validation/content/ai_writing.py, validation.cli prose-quality)
  • shipped the evidence graph (EVIDENCE-GRAPH-1), reproduction bundle (REPRO-BUNDLE-1), release-readiness dashboard (DASHBOARD-1), pipeline plugin stages (PLUGIN-STAGES-1), and incremental pipeline skipping (INCREMENTAL-PIPELINE-1) — all opt-in, default plan unchanged
  • parallelized CI infrastructure tests (pytest-xdist -n auto) and derived the test-project matrix dynamically from public_scope (CI-MATRIX-DYNAMIC-1)
  • quieted terminal logging (LOG-CLEAN-1), consolidated the safe markdown reader (READFILE-SAFE-1), and ran documentation-accuracy passes across docs/ and infrastructure {SKILL,README,AGENTS}.md
  • see CHANGELOG.md for the full entry

v3.2.0 — Format, Logging, and Exemplar Hardening (2026-06-04)

  • added optional Pandoc-backed DOCX/EPUB rendering with per-format toggles (default off)
  • extended the Active Inference validation spine with producer completeness, stale-artifact checks, and dependency graph v2
  • hardened the public SIA harness boundary and re-baselined coverage gaps
  • closed GitHub supply-chain hygiene (SHA-pinned actions, actionlint gate, guarded Dependabot automerge) and the XML-parser policy
  • see CHANGELOG.md for the full entry

v3.1.0 — Public Exemplar / Validation-Spine Release (2026-05-30)

  • added the sixth public exemplar, projects/templates/template_sia/, plus the reusable infrastructure.sia harness and validation docs
  • promoted the first Active Inference validation-spine tracks: provenance, reproducibility, and counterexample
  • refreshed public project signposting to projects/templates/... and checked folder-level AGENTS.md / README.md coverage across public exemplars
  • hardened public project coverage orchestration by pinning subprocess coverage behavior across project virtual environments
  • shipped curated release notes and a 3.1.0 metadata bump

Roster note (post-v3.1.0): the git-tracked public exemplar set under projects/templates/ has since grown beyond the six referenced above — template_autoscientists, template_newspaper, and template_textbook were added (each double-published as a standalone GitHub repo + Zenodo DOI), bringing the current public exemplar roster to nine. The authoritative, always-current count and names live in docs/_generated/active_projects.md and docs/_generated/COUNTS.md — consult those rather than this historical release log.

v3.0.0 — Production / Stable (2026-02-22)

  • mypy strict adopted as the baseline gate for infrastructure/ (live counts in TO-DO.md)
  • Ruff format enforcement
  • Security hardening: Bandit MEDIUM+ gate in CI; pip-audit blocking since v0.7.2 (ignore list + retries — see CHANGELOG.md)
  • Dockerfile modernised to python:3.12 + uv

v2.x — Foundation Series (2025–2026)

Release Theme
v2.0.0 Two-layer architecture, thin orchestrator pattern, declared DAG pipeline, multi-project support
v2.1.0 Unified intelligent logging — ProjectLogger, structured format, log_operation(), format_duration()
v2.1.1 CI Zero-Mock gate (verify_no_mocks.py); mock/fake patterns eliminated from suite
v2.2.0 Orchestration hermeticity — script discovery, get_subprocess_env(), hermetic subprocess env
v2.3.0 Type safety — TypedDicts for config, ResolvedTestingConfig, ProjectInfo dataclass
v2.4.0 Test isolation — real tmp_path + env-isolation fixtures (pytest's monkeypatch fixture remains permitted by the no-mocks policy and is still used for boundary substitution and env isolation; it was not eliminated)
v2.5.0 Structured log assertions — caplog-based, log_parser.py
v2.6.0 Ruff lint remediation: 710 → 0 errors across infrastructure/, scripts/, tests/
v2.7.0 Type narrowing & mypy baseline: 100 → 0 errors across core/
v2.8.0 Error reporting & resilience — typed InfraError constants, standardized error format
v2.9.0 Documentation parity — python3 → uv run python, auto-generated API reference

v0.6.0 — Desloppify Code-Health Campaign (2026-03-10)

161-commit systematic blind-review campaign across all infrastructure packages:

  • Import hygiene (unused imports, sys.path mutations, TYPE_CHECKING guards)
  • Exception narrowing (specific types, context restoration, no silent swallowing)
  • Dead-code removal (coverage_reporter.py, stub wrappers, passthrough methods)
  • Type annotations modernised (legacy typing → built-in generics)
  • API surface consolidation (OllamaClientConfig, PerformanceMetrics, ProjectLogger)
  • Bug fixes (inverted bool, stall detection, path bugs, broken imports)
  • Structural: eliminated core.py hub, extracted _build_stage_list, broke circular dep
  • Logging noise reduction; docstring bloat; test name collisions resolved

Planned

The active planning surface is TO-DO.md. It is intentionally more specific than this roadmap and now groups work as Minor, Medium, and Major.

Ongoing — Hygiene and Backlog Accuracy

  • keep TO-DO.md, generated facts, release metadata, and public-scope docs synchronized after each release or major verifier pass
  • keep forkability checks and regression pins green, finish the active-inference fixed-point cluster, and only then add new scientific surface

The GitHub supply-chain hygiene items (SHA-pinned actions, actionlint, safe Dependabot automerge) shipped in v3.2.0; terminal logging cleanup (LOG-CLEAN-1), the safe markdown-read helper (READFILE-SAFE-1), and the dynamic CI project matrix (CI-MATRIX-DYNAMIC-1) shipped in v3.3.0. See Completed Releases above.

Next (open backlog)

The authoritative open backlog lives in TO-DO.md. Current themes are verifier-first maintenance items:

  • Active-inference semantic fixed point (AI-SEMANTIC-FIXPOINT-1): finish the variable-generation and composed sheaf/roadmap cluster before expanding scientific scope.
  • Active-inference gate runtime (AI-GATE-PERF-2): reduce redundant slow refreshes only after the semantic fixed-point and negative controls are green.
  • AutoResearch report and benchmark ergonomics (AR-REPORT-ERGONOMICS-1, AR-BENCHMARK-ERGONOMICS-1): clarify reviewer-facing evidence and benchmark boundaries before adding new evidence types.
  • AutoResearch source-ledger contract (AR-SOURCE-LEDGER-2): promote claim/source drift checks into project-local gates.

Next Generation (vision)

The prior vision list — evidence graph (EVIDENCE-GRAPH-1), incremental pipeline (INCREMENTAL-PIPELINE-1), plugin architecture (PLUGIN-STAGES-1), hermetic release bundles (REPRO-BUNDLE-1), and local dashboard (DASHBOARD-1) — all shipped in v3.3.0 (opt-in/default-off). See Completed Releases above. The live follow-up work for these capabilities is tracked in the open backlog above and in TO-DO.md; new long-horizon vision items should replace this note as they are defined.

Next Up

Use TO-DO.md as the authoritative backlog and live snapshot. The current open top items are:

  • AI-SEMANTIC-FIXPOINT-1
  • AI-GATE-PERF-2
  • AR-REPORT-ERGONOMICS-1
  • AR-BENCHMARK-ERGONOMICS-1
  • AR-SOURCE-LEDGER-2

Shipped and not re-tracked here: the GitHub supply-chain hygiene set (GH-PIN-1, GH-ACTIONLINT-1, GH-AUTOMERGE-1), LOG-CLEAN-1, READFILE-SAFE-1, CI-MATRIX-DYNAMIC-1, FMT-BUNDLE-1, AI-SPINE-V2, COVERAGE-REBASE-1, and the five v3.3.0 Major capabilities (EVIDENCE-GRAPH-1, INCREMENTAL-PIPELINE-1, PLUGIN-STAGES-1, REPRO-BUNDLE-1, DASHBOARD-1), plus pip-audit blocking CI, Bandit LOW triage, docs-lint CI, the per-project test runner, and SIA public-exemplar promotion. See CHANGELOG.md for the release history.

Quality Metrics

Authoritative counts and gate outputs live in the Live state snapshot table in TO-DO.md (re-baseline there after substantive changes). This roadmap avoids duplicating numbers that drift between audits.

Topic Where to verify
mypy / ruff / Bandit / pip-audit / health TO-DO.md
CI wiring .github/workflows/ci.yml, .github/AGENTS.md
Coverage gaps coverage-gaps.md

See Also