v0.7.0 — Tier 6 DX & Distribution + Tier 7 KB Quality Engineering

What's New

Tier 6 — DX & Distribution

• backendpro lint — 18 backend anti-pattern rules (SARIF, pre-commit, GitHub Action)
• backendpro export — Obsidian (BM25 similarity graph), Notion CSV, Org-mode
• backendpro learn — SM-2 spaced-repetition flashcards
• VS Code extension — Search, Explain Selection, CodeLens stack guidelines
• JetBrains plugin stub — Tools menu actions
• Web playground at backendpro.cc

Tier 7 — KB Quality Engineering

• Golden query test suite — 239 retrieval assertions across all 34 domains, 12 stacks, cross-domain, and compare mode
• backendpro dedup — BM25 near-duplicate detection (within + cross-domain, threshold, allowlist)
• backendpro coverage — per-domain gap/thin analysis with shields.io badge

Numbers

• 652 tests passing (up from 332)
• 34 domains, 12 stacks
• Zero dependencies (pure stdlib)

pip install --upgrade backendpro

Full changelog: https://github.com/shashankswe2020-ux/backend-pro-max-skill/blob/main/CHANGELOG.md

v0.2.0 — compare, REPL, freshness, validator, tests, CI

[0.2.0] — 2026-04-22

Added

• Compare mode — backendpro compare "<A>" "<B>" [...] [--domain <d>] outputs a side-by-side markdown table for tradeoff decisions / ADRs. Programmatic API: core.compare().
• Interactive REPL — backendpro --interactive (-i) with slash commands: /d, /s, /all, /cmp, /stale, /list, /help, /quit.
• Synonym / hybrid search — query expansion bridges keyword gaps (e.g. "partial failure" → compensation, saga, fault). Disable with --no-expand.
• Confidence scores — every result now carries a _score field (BM25). Markdown output adds a (score: 7.69, confidence: high) label per result. Hide with --no-scores.
• Freshness tracking — optional Last Updated column convention (YYYY-MM-DD). New flags: --max-age-months N (filter), --stale --domain <d> --max-age-months N (audit). New API: core.find_stale().
• In-memory index cache — lazy, mtime-invalidated. No more BM25 recompute per query. New API: core.clear_cache().
• CSV validator — backendpro-validate console script (also wired into CI). Validates every CSV against its declared schema (columns, ragged rows, date formats).
• Test suite — 37 pytest cases covering BM25 ranking quality, edge cases (empty CSV, missing columns), domain detection, cache behavior, compare, stale filtering, and CSV validation.
• GitHub Actions CI — .github/workflows/ci.yml runs ruff lint + CSV validator + pytest + CLI smoke tests on Python 3.9 / 3.11 / 3.12.
• CONTRIBUTING.md — content & schema rules, severity definitions, ranking-quality test contract, synonym guidance.
• Landing page (index.html) with new "Power Tools" section, refreshed terminal demo, and CI/test stats.
• [project.optional-dependencies] dev — pytest, pytest-cov, ruff. Plus [tool.pytest.ini_options] and [tool.ruff] config in pyproject.toml.

Changed

• core.search(), core.search_stack(), core.search_all() accept new keyword-only args (min_score, max_age_months, expand). Existing positional signatures are unchanged → fully backwards compatible.
• Pattern domain rows now include a Last Updated column (set to 2026-01-15 for the initial 22 rows).

Fixed

• 9 pre-existing CSV data-corruption bugs (caught by the new validator): unquoted commas mis-aligned fields in api.csv, architecture.csv, cache.csv, iac.csv, messaging.csv (×3), observability.csv, plus csharp-aspnet, elixir-phoenix, nodejs-express, php-laravel, rust-axum stack files. Search was silently degraded for these rows.
• architecture domain's search_cols referenced a non-existent Use Case column → corrected to When to Use.
• Replaced deprecated datetime.utcnow() with datetime.now() (Python 3.13+ deprecation).