Featured Project

Python Package · PyPI

DBWarden

loading

The SQL-first database toolkit for SQLAlchemy.

DBWarden is a SQL-first migration system for SQLAlchemy that replaces Python-based migration workflows with explicit, reviewable SQL generated directly from your models. Unlike script-based migration tools, DBWarden does not introduce a migration runtime or require executing generated Python migration code. It produces plain SQL files that can be reviewed, versioned, and executed directly against any environment.

SQL-first

Migrations are plain SQL. No runtime required.

Rollback included

Every migration carries upgrade + rollback SQL.

Schema snapshots

Checksummed JSON after every migration for deterministic diffs.

Column-level diffing

Type, nullability, default, and comment changes.

Impact analysis

AST-based scan to detect breaking changes before deploy.

Offline mode

Export model state, generate migrations anywhere, no DB needed.

FastAPI integration

Startup validation, health checks, runtime lifecycle.

5 DB dialects

PostgreSQL, MySQL, MariaDB, SQLite, ClickHouse.

12stars
1,131tests
51KLOC
5DB dialects
15+CLI commands
Python Package

PyPI

seoslug

loading

Deterministic SEO payload generation for Python.

seoslug is a deterministic, pure-function SEO toolkit for Python that generates complete SEO payloads from two typed inputs. Define your site once with SEOConfig (canonical host, site name, title template, base URL) and your page with SEOEntity (entity type, title, excerpt, image, timestamps). The single build_seo_payload() function merges them into a structured dict that maps directly to canonical URLs, Open Graph tags, Twitter Cards, and JSON-LD structured data.

Same inputs always produce the same output - no side effects, no randomness, no hidden state. This makes seoslug ideal for static site generators, SSR frameworks, and any build pipeline where SEO metadata must be testable, cacheable, and reproducible across builds. Zero external dependencies at runtime.

PythonSEOJSON-LD

Typed inputs

SEOConfig + SEOEntity with validated fields. No raw dicts, no guessing.

Deterministic by design

Same inputs always produce the same payload. Pure function, zero side effects.

Multi-format output

Canonical, OG, Twitter Cards, and JSON-LD from a single call.

SSG-first

Pre-generate SEO at build time. Works with Astro, Next.js, Hugo, Jekyll.

Zero runtime deps

Standard library only. No external packages required at runtime.

Fully typed

Complete type hints, MyPy and Pyright compatible out of the box.

Local-first Tools

heard

v0.1.0

Hold left Shift, speak, release. Speech-to-text via on-device Whisper → intent resolved by a Needle 26M function-call model → dispatched to a tool handler. Tools: launch apps, media control, volume, window actions, workspace switching, system queries. Fully local-first, no cloud dependency.

AI-powered Python Whisper Linux evdev

trustsight

loading

Run it before yay -Syu. Clones AUR repos, generates structured PKGBUILD diffs between versions, and applies 11 detection rules (curl pipe bash, checksum disabled, base64 decode, etc.). Scores packages 0–100 with configurable severity weights, tracks novelty via SQLite, and optionally produces an LLM verdict via OpenAI or Ollama. Published on PyPI.

AI-powered Python Arch Linux Security

PyPI Utilities

schemap

Python

Automatic Pydantic v2 schemas from SQLAlchemy 2.0 ORM models.

detrack

Python

Strip tracking parameters from URLs. Deterministically. Zero dependencies. 60+ default patterns, case-insensitive, pure functions, metadata returned.

acrresolv

Python

Acronym resolver for Python. Converts natural language phrases to acronyms. Zero dependencies, CLI and library usage. Being expanded with ML.

More Projects

crxml

Rust

PyPI package

Fast streaming parser for Crystal Reports XML exports with Rust acceleration. Streams row-by-row without loading the full document into memory, handling files of any size. Built-in pipeline stages: rename, cast types, filter, and drop fields, all chained with pipe operators for composable transformations. Parallel mode distributes batches across worker processes using Python's multiprocessing, achieving near-linear speedups on multi-core systems. Outputs DataFrames via pandas, CSV files, or lists of dicts. The pandas import is lazy, only loaded when you actually convert, keeping the CLI lightweight.

PythonRustXMLpandas

vigil

Prefect

Prefect flow

Personal GitHub activity monitoring workflow built with Prefect. Fetches events, commits, and PR activity via PyGithub, then stores structured records in ClickHouse through dbwarden's migration framework. Designed for automated periodic execution: the flow runs on a cron schedule, collects the latest activity since the last run, and inserts it into the warehouse. Structured logging captures each pipeline stage, and errors surface through Prefect's native notification system instead of failing silently. A simple dashboard query gives you a timeline of your open source contribution activity without relying on GitHub's own graphs.

PythonPrefectPyGithubClickHouse

mapcreator

React

Web application

Tile-based building map editor built with React and Canvas. Draw multi-floor building layouts with a grid-based editor; place walls, designate rooms, and connect them with doors, stairs, and elevators. Each floor is a separate layer, and the A* pathfinding system can route from any room to any other across floors, automatically selecting stairs or elevators. The canvas renders at 60fps with virtual scrolling for large maps. State is managed entirely through React hooks, making it easy to undo, redo, or export the map data as JSON for use in other tools. No external mapping libraries or APIs required.

ReactTypeScriptCanvasA* Pathfinding

clickvault

Go

Vault plugin

HashiCorp Vault database secrets engine plugin for ClickHouse. Generates ephemeral, short-lived ClickHouse users on demand (dynamic secrets) and rotates static role passwords on a schedule. Implements the Vault database plugin SDK, integrating natively with Vault's database secrets engine. All generated ClickHouse DDL is isolated in a dedicated SQL layer, and clustered deployments get ON CLUSTER appended automatically - the rest of the plugin never branches on topology.

GoVaultClickHousePlugin