Skip to main content

Premium VBA static analyser & compile-safety prechecker for AI-generated VBA

Project description

VBAlidator

Premium VBA static analyser & compile-safety prechecker for AI-generated VBA. Drop-in behind any LLM-VBA generator — get a deterministic 0–100 confidence score and a stable JSON report before the code ever reaches a workbook.

PyPI version CI Docker Docs License: MIT

VBAlidator parses .bas / .cls / .frm files through a real lexer / preprocessor / parser / analyzer pipeline, applies 35 documented rules (see the catalogue), and returns a verdict CI pipelines and AI agents can act on directly.

What it catches

The full list lives at docs/rules/index.md. The high-impact subset that AI generators trip over most often:

  • VBA001 undefined identifiers, VBA002 missing object members
  • VBA006 wrong argument count, VBA007 ByRef type mismatches
  • VBA210 Set on scalar, VBA211 missing Set on object assignment
  • VBA221–VBA224 Property Get/Let/Set arity & semantics
  • VBA230 / VBA231 non-constant Const initialisers
  • VBA240 arithmetic between string and numeric literal
  • VBA300 missing PtrSafe on 64-bit Office Declares
  • VBA320 missing Option Explicit (warning)
  • VBA330 incomplete Implements of an interface
  • VBA340 / VBA341 RaiseEvent without matching Event declaration / wrong arity
  • VBA_LEX001 / VBA_LEX002 unrecognised characters & malformed date literals
  • VBA_RT001 (optional) errors caught by the actual VBE compiler via Office round-trip

Install

pip install vbalidator

…or grab the multi-arch image:

docker pull ghcr.io/twobeass/vbalidator:latest

In one CLI call

vbalidator ./MyModules --host excel
MyModules/Module1.bas:42: ERROR  [VBA001]  Undefined identifier 'tpyo' in 'DoStuff'.
MyModules/Module1.bas:1:  WARNING [VBA320]  Module 'Module1' is missing `Option Explicit`.

Files scanned : 12
Errors        : 1
Warnings      : 1
Confidence    : 77 / 100  (needs fixes)
Report saved  : vba_report.json

Exit code is 1 when the score is below the gate (--score-threshold, default 90) or there is at least one error.

In one Python call

from vbalidator import precheck

result = precheck("Module1.bas", host="excel")

if result.compile_safe:
    deploy(result)
else:
    for err in result.errors:
        print(f"{err['rule_id']}: {err['message']}")

print(f"score = {result.score} / 100")
print(result.json())   # canonical JSON v2 report

PrecheckResult exposes errors, warnings, info, issues, and the canonical .json() report. It's also truthy when compile_safe, so if precheck(...): ... is idiomatic.

See docs/ai-integration.md for full recipes (Anthropic SDK, OpenAI Agents, LangChain, GitHub Actions).

Optional dynamic verification

When a Windows host with Office is available, cross-check the static verdict against the actual VBE compiler:

vbalidator ./MyModules --host excel --roundtrip
result = precheck("Module1.bas", host="excel", roundtrip=True)

Compile errors VBE itself reports come back with severity='compile_verified' and rule_id VBA_RT001. On non-Windows hosts the call degrades gracefully to a single info-level notice instead of crashing.

Bundled host models

--host excel|word|access|outlook auto-loads the matching object model from src/models/. Cover ~600 host-specific globals / classes out of the box — no need to run a COM exporter for the common case.

Documentation

The full site is at https://twobeass.github.io/VBAlidator/ — deployed automatically from main by .github/workflows/docs.yml.

In-repo:

  • Quickstart — install → scan → ship
  • AI pipeline integration — patterns for Claude, OpenAI Agents, LangChain, CI gates
  • Usage — full CLI / Python API reference
  • Configuration — host models, custom models, conditional-compilation defaults, heuristics
  • Architecture — pipeline internals
  • Rule catalogue — all 35 rules with examples
  • CI / CD — workflows, release lifecycle, branch protection
  • UAT walkthrough — section-by-section human validation script for the current branch
  • TODO — open items that need a Windows machine, live PyPI/GHCR/Pages access, or future code work
  • Roadmap — what's done, what's queued

Development

git clone https://github.com/twobeass/VBAlidator
cd VBAlidator
pip install -e ".[dev]"
pytest                              # 148 tests
ruff check src tests                # lint
python tools/generate_rule_docs.py  # refresh docs/rules/ after a rule change

PR titles must follow Conventional Commits (feat: / fix: / docs: …). The release pipeline derives the next semver bump from your commit history.

License

MIT.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

vbalidator-1.3.6.tar.gz (554.3 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

vbalidator-1.3.6-py3-none-any.whl (547.6 kB view details)

Uploaded Python 3

File details

Details for the file vbalidator-1.3.6.tar.gz.

File metadata

  • Download URL: vbalidator-1.3.6.tar.gz
  • Upload date:
  • Size: 554.3 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for vbalidator-1.3.6.tar.gz
Algorithm Hash digest
SHA256 0c1d46222aa1611401c0ae4ada6f0f62e1f90a3af94cf45fdeab9e177e5d2aac
MD5 ab8a8431ba61d870a855d06021226cfd
BLAKE2b-256 4912d8151a8ac7b437120c5690a295ec3874a14839f938e548aa7611362a6004

See more details on using hashes here.

Provenance

The following attestation bundles were made for vbalidator-1.3.6.tar.gz:

Publisher: release.yml on twobeass/VBAlidator

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file vbalidator-1.3.6-py3-none-any.whl.

File metadata

  • Download URL: vbalidator-1.3.6-py3-none-any.whl
  • Upload date:
  • Size: 547.6 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for vbalidator-1.3.6-py3-none-any.whl
Algorithm Hash digest
SHA256 f6ea364ca2335d985f5a54900634e31007c27667fe4e530b0cb139068fb6d1c7
MD5 0f0ca16c3712f88641357ff771a6d183
BLAKE2b-256 25664a248e3e1c03f5367c8b3e45de6af52d8742923208d06ae2b5fb11aa71a8

See more details on using hashes here.

Provenance

The following attestation bundles were made for vbalidator-1.3.6-py3-none-any.whl:

Publisher: release.yml on twobeass/VBAlidator

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page