Tags

v5.7.1

v5.7.1 — PyPI trove classifiers (Python 3.11–3.14)

v5.7.0

v5.6.0

Release 5.6.0 — first PyPI publish

v9.15.0

v9.15.0 — Production mission release

Six verified turbo research proposals, pipeline hardening,
GitLab Pages discoveries gallery, macOS CI shell-runner fixes.

v9.14.1

v9.14.1 (2026-06-22) — tech-debt cleanup, 12 atomic commits + remote polish merge

Branch: feat/production-upgrade

This tag now points to the merge commit (6d33f4f) that integrates:
- My 12 atomic commits (H-8 Tier 1 follow-up, observability, tech debt)
- Remote's 10+ polish commits (splash, budget, desktop, key centralization)

Summary of changes since v9.14.0:
- H-8 Tier 1 follow-up: cost tracking for guarded_call + BaseLLMClient
  + AsyncLLMClient now actually works (4 silent-failure bugs fixed).
- 5 missing Prometheus counters wired (API_REQUESTS, RATE_LIMIT_HITS,
  CACHE_HITS, VERIFICATION_RUNS, DISCOVERIES_GENERATED).
- OpenAPI contract: 9 missing operationIds fixed.
- mypy regression gate: scripts/check_mypy_regression.py with
  61-error baseline.
- 24 Phase 1 reorg branches DELETED; archive/phase1-reorg-2026-06-08
  tag preserves the work.
- pre-commit hook installed locally; .secrets.baseline refreshed
  (103 known false positives).
- docs/DEFERRED.md captures 5 explicit deferrals (P2-D, P4, P2-A A2,
  P3-3, Phase 1 reorg) with verbatim REWORK_PLAN rationale.
- Remote polish: splash animation, budget key centralization, desktop
  port (mac/win), TUI v9 minor improvements.

Audit ref: 2026-06-22

archive/phase1-reorg-2026-06-08

Archive of Phase 1 reorg + stabilization work, 2026-06-08.

Series of 22 commits across reorg/01..12 + stab/01..12 (the MERGE_PLAN
documented only 26 commits on stab/01..08; the user continued the
stack through stab/12 in the same session). The full work:
- reorg/01..12: 12 commits (subtraction ~84.5k LOC, core de-cycling
  refactor, architecture doc, etc.)
- stab/01..12: 12 commits (bug sweep, test-runnable, requirements,
  hang-proof, pandas-deps, terminal-rm, tmax-conn, import-sweep,
  import-guard, test-hygiene, deshadow-pipeline, fake-llm-test).

Status: ABANDONED per docs/DEFERRED.md. Since this work was frozen
(2026-06-08), feat/production-upgrade accumulated 91+ commits of
Round 5 audit work that touches the same modules. The reorg was
likely significantly drifted. The substance of the reorg (dead-code
deletion, dependency de-cycling) was achieved post-hoc by the
Round 1 P1 subtraction (155 files deleted) and Round 5 audit.

This tag preserves the work for reference. The reorg/* + stab/*
working branches have been deleted (after this tag) so the
abandoned branches don't clutter the local branch list.

To recover any of the deleted branches (if a future maintainer
wants to rebase them onto current feat/production-upgrade):
  git tag -d archive/phase1-reorg-2026-06-08
  git branch reorg/01-subtraction <sha-of-reorg/01-tip>
  ... (same for each deleted branch)

The SHAs of the deleted tips are reachable from this tag's tree.

Audit ref: 2026-06-22 O-3

reorg-core-decycled

Core fully de-cycled (2026-06-07)

The pathological 24-package core-blob SCC is fully decomposed across 10 stacked,
test-validated branches (reorg/01-subtraction .. reorg/10-devmode), all pushed.

Dissolved: the 24-blob, then the engine cycle {agents,pipeline}, then {api,auth}.
Freed to clean layers: c4 (a true leaf), c4_analysis, triz, metamodels, archetypes,
core, llm, knowledge, verification, exploration, discovery, validation, api, auth,
publishing, social, graph, export, analogy, plugins.
New foundational leaf homes: contracts, security, infrastructure.
Removed ~84.5k lines of dead code (v6_legacy + the v8/ parallel app).
Decomposed the 1189-line discovery god-module (api router -> discovery domain).

Remaining non-trivial SCCs are 2 small, GENUINE cohesive couplings, intentionally
LEFT AS-IS (not misplaced code): {agent,cli,codegen,mcp_server} (CLI/MCP tooling)
and {patterns,simulations} (runner subclass + GPU bridges). Breaking them would be
cosmetic graph-gaming or a deliberate redesign for the author to own.

Recover original state from tag pre-reorg-baseline.

reorg-decycle-24to8

End of structural de-cycling phase (2026-06-07)

Core-blob SCC reduced 24 -> 8 packages across 5 stacked branches
(reorg/01-subtraction .. reorg/05-eventbus), all pushed.

Freed from the cycle: c4 (now a clean leaf — the key goal), c4_analysis,
triz, metamodels, archetypes, core, analogy, graph, export, plugins, llm,
publishing, social, knowledge, verification, exploration.
New foundational leaf homes: contracts (pipeline_config), security (guardian),
infrastructure (events).

Also removed ~84.5k lines of dead code (v6_legacy + the v8/ parallel app).

Residual 8-blob = agents, api, auth, discovery, litintel, pipeline, solver,
validation — the genuine orchestration core; further cuts need interface/contract
design, not file moves.

Recover original state from tag pre-reorg-baseline.

pre-reorg-baseline

Baseline before Phase 1 reordering (2026-06-07)

Snapshot of c4reqber/turbo-cdi v5.6.0 as-is, immediately before the
cleanup/reordering work begins. Return here to compare against the
original state.

Phase 1 plan: in-place restructuring (NOT a fresh parallel tree).
  1. Subtraction: delete v6_legacy, the dead v8/ parallel app, tui shim, merge dups
  2. Component analysis on the survivor
  3. Reorg into ~16 target modules with boundary contracts
  4. Carve the worker boundary (heavy Python -> isolated workers)
  5. New features born into the clean structure
Baseline metrics: src/ ~200k LOC Python, 86 packages.