[dbccpage]

What this app does

1. Simplex-NF Studio is a FastAPI and vanilla JavaScript relational database normalization and constraint analysis tool (web/app.py, normal_forms/domain/relational.py, normal_forms/domain/relational_advanced.py).
2. Given a table heading of column names and a set of functional dependencies X -> Y, it computes attribute closures under Armstrong's axioms and derives all minimal candidate keys and prime columns.
3. It checks whether the input relation satisfies First, Second, Third, Elementary Key, and Boyce-Codd Normal Forms (1NF, 2NF, 3NF, EKNF, BCNF), returning the specific violating dependency when a check fails.
4. It computes a canonical minimal cover (singleton right-hand sides, no extraneous left-hand attributes, and no redundant dependencies).
5. Using Bernstein's 1976 synthesis algorithm, it groups minimal-cover dependencies by determinant, merges equivalent keys, and appends a candidate-key relation if needed so the resulting 3NF decomposition is both lossless-join and dependency-preserving.
6. Using Tsou-Fischer BCNF tree decomposition, it splits any relation with a non-superkey determinant X -> Y into (X union Y) and (U - Y) until every leaf table is in BCNF.
7. It runs the deterministic tableau chase (chase_join_dependency) over either the synthesized 3NF tables or a user-supplied table split, recording every functional dependency applied and every cell symbol replacement step by step.
8. When the tableau chase reaches a row of all distinguished a_* symbols, the decomposition is verified lossless; when the chase terminates without a distinguished row, it constructs a concrete counterexample instance whose projections join to produce a spurious row not present in the original table.
9. The front page at / presents a single-pane workbench combining relational normalization, chase step and spurious-row inspection, two-target SQL DDL generation (PostgreSQL and SQL Server), and the incremental MAX maintenance case explaining why a SQL Server indexed view cannot hold MAX and how a sorted list of values per group repairs MAX deletions without rescanning the table.
10. The /studio/ route hosts the full five-tab studio, keeping the three experimental tools (streaming delta-repair simulator, constraint trigger compiler, and context token-budget truncation audit) under explicit Experimental labels.

Baseline

• Desktop front page screenshot (1280 × 800): overnight/baseline/front_1280x800.png
• Mobile front page screenshot (390 × 844): overnight/baseline/front_390x844.png
• Browser console output: overnight/baseline/browser_console.txt
• Time to first result: overnight/baseline/first_result_timing.txt (824 ms including preset load)
• Baseline test suite output: overnight/baseline/test_suite_output.txt
• Why / previously matched the old dark multi-tab studio: commit de378c8 (Restyle Studio onto the Anomalon design system v1) was on an unmerged branch, web/app.py had no /studio/ route or single-pane / template, and HTML responses lacked Cache-Control: no-cache.

Changed

• Record baseline screenshots, browser console log, timing, and test output (7f7d93c) — checked by overnight/baseline/front_1280x800.png, overnight/baseline/front_390x844.png, and overnight/baseline/test_suite_output.txt.
• Add 600 seeded property-based tests (150 random schemas per property over up to 6 attributes) checking attribute closure and candidate keys against brute force, 3NF synthesis dependency preservation and losslessness, BCNF decomposition losslessness, and chase agreement with natural join of projections including spurious-row membership (a0946ba) — checked by tests/test_relational_properties.py.
• Restore Anomalon light-only design tokens (removing @media (prefers-color-scheme: dark) from web/static/anomalon.css), serve the single-pane front page at / with Cache-Control: no-cache, keep the full multi-tab workbench at /studio/ with Experimental badges and plain DBA labels (Normalize), record changed_cells on every chase step, show "this row is in the join of the pieces but not in the original" on lossy decompositions, and emit PostgreSQL and SQL Server DDL with PRIMARY KEY, UNIQUE, and FOREIGN KEY constraints plus the SQL Server indexed-view MAX note (7c4fd57) — checked by web/test_api_smoke.py, tests/test_frontend.cjs, overnight/after_front_1280x800.png, overnight/after_front_390x844.png, overnight/after_studio_1280x800.png, and overnight/after_studio_390x844.png.
• Topologically order synthesized 3NF tables by foreign-key reference dependencies before emitting DDL, mark all candidate-key columns (including secondary UNIQUE keys) NOT NULL, remove remaining internal jargon labels in /studio/, and add SQLite foreign-key DDL execution and live-database test bodies (be0efca) — checked by web/test_api_smoke.py::test_ddl_topological_order_and_sqlite_fk_execution.
• Start live PostgreSQL (postgres:16-alpine), SQL Server (mcr.microsoft.com/mssql/server:2022-latest), and MongoDB (mongo:7) containers in .github/workflows/ci.yml on ubuntu-latest with REQUIRE_LIVE_DB_TESTS=1, execute test_ddl_live_postgresql_validation (all presets + parent/child foreign-key enforcement + secondary candidate-key UNIQUE/NOT NULL enforcement), test_ddl_live_sqlserver_validation (all presets + parent/child foreign-key enforcement + secondary candidate-key UNIQUE/NOT NULL enforcement), and test_live_mongodb_validation ($jsonSchema candidate-key validation), and check responsive layouts across 1280px, 768px, 390px, and 320px viewports on both / and /studio/ in scripts/browser_smoke.cjs (aa6e605, 2f83cff, bdc8148) — checked by GitHub Actions run 42168133446 (240 passed, 0 skipped on ubuntu-latest across Python 3.10 and 3.14; 33/33 checks in browser).

Not done

• None — all unit, property-based (600 seeded cases), frontend, browser (33/33 checks across / and /studio/ at 1280px, 768px, 390px, and 320px), and live database container checks (postgres:16-alpine, mcr.microsoft.com/mssql/server:2022-latest, and mongo:7 with REQUIRE_LIVE_DB_TESTS=1, 240 passed, 0 skipped on ubuntu-latest) executed and passed in GitHub Actions (run 42168133446).

Decisions for the owner

1. Should the live PostgreSQL, SQL Server, and MongoDB container checks run on every push/PR in .github/workflows/ci.yml?
   • Option A (Recommended): Keep postgres:16-alpine, mcr.microsoft.com/mssql/server:2022-latest, and mongo:7 running on ubuntu-latest in .github/workflows/ci.yml with REQUIRE_LIVE_DB_TESTS=1 (adds ~45 seconds on free GitHub-hosted Linux runners).
   • Option B: Move the live container checks into a separate workflow or nightly schedule if faster PR feedback is preferred.
2. Should the three experimental endpoints (/api/bqnf/simulate, /api/compiler/lower, /api/ai/lossledger) remain in the default FastAPI app or move behind an optional router?
   • Option A (Recommended): Keep all three endpoints mounted in web/app.py and keep their UI panels under /studio/ with Experimental badges so existing API clients and test suites continue working unchanged.
   • Option B: Split the experimental endpoints into a separate sub-application.

Found but out of scope

• Untracked working-tree planning notes (normal_forms/planned-tests.md, normal_forms/planned-tests-516-615.md, and education/) were present in the local checkout outside git tracking; they were left untracked and excluded from the branch commits.