Skip to content
@MCR-Project

MCR Project

Personal software projects by @MCR-3 and @alexinsc: developer tools, data apps and open-source utilities.

MCR Project

MCR Project is a small engineering organization maintained by @MCR-3 and @alexinsc. It gathers the software we design and build together: developer tools, data applications and desktop utilities.

We favor small, well-scoped projects with clear boundaries between components, automated builds and tests, and documentation that explains not only how to run a project but why it is structured the way it is.

Projects

An ETF correlation dashboard and portfolio simulator.

Market Analyser helps investors understand how the holdings of an ETF move together and how a portfolio built from them would have performed historically. It provides an interactive correlation matrix with hierarchical clustering, a network view of holding relationships, price charts (line or candlestick), a table of per-holding risk measurements, and a backtesting engine (buy-and-hold or periodic rebalancing, recurring contributions and withdrawals, standard performance and risk metrics). Portfolios are stored in the browser only, so no account is required, and they can be shared through links or exported as backups.
Live demo

Architecture

  • Frontend: React + Vite (TypeScript). A single storage layer manages portfolios in localStorage, and view state (matrix ordering, selected metrics) lives in the URL so any view is shareable.
  • Backend: FastAPI. Stateless services handle portfolio simulation and shared statistics; a versioned REST API exposes simulation, risk, correlation, fund and price data, with per-client rate limiting and explicit error semantics (400 / 404 / 429 / 503).
  • Extensibility: holdings-table columns and portfolio summary tiles are implemented as self-describing plugins, each declaring its inputs, computation and documentation.
  • Data: Supabase (PostgreSQL) stores funds, holdings, adjusted prices, dividends, splits and the risk-free rate. Prices are tiered by age (daily, weekly, monthly buckets) to keep storage efficient.
  • Pipeline: Playwright-based scrapers collect holdings from six ETF providers, yfinance provides market data, and a scheduled GitHub Actions job refreshes the data every weekday. A freshness endpoint reports whether the data is on schedule.
  • Delivery: Docker for local development and testing, pytest test suite, deployment configured for Render.

A code visualization tool that shows how the functions of a codebase call each other.

Function Viewer renders functions as draggable cards connected by wires that follow the real call relationships, in the style of an Unreal Engine Blueprint graph. It lets developers navigate a codebase by following calls up (callers) or down (callees) instead of jumping between files, with search, a file explorer, and two auto-layout modes (by call flow, or grouped by file). It currently supports Python, Rust, TypeScript and JavaScript.
Live demo

Architecture

  • Backend: FastAPI. A dispatcher groups the loaded files by extension and hands each group to a language plugin. Each plugin extracts function definitions and resolves calls between them; Python uses the standard ast module, and Rust and TypeScript/JavaScript use tree-sitter. Results are merged into one graph using path-based identifiers to avoid collisions, and adding a language means writing one new plugin.
  • Frontend: React 19 with React Flow, and a single Zustand store holding the graph, the active functions and the layout mode. Layouts are computed with dagre.
  • Delivery: GitHub Actions builds the frontend and runs health checks on the backend. The backend is deployed on Render (Docker) and the frontend on GitHub Pages.

A native Linux client for Google Tasks, with a desktop app and a terminal dashboard.

gtasks is an offline-first client: tasks are cached locally and synchronized in the background, with hierarchical tasks and natural-language dates ("Buy milk tomorrow"). All communication goes directly to Google's API, and credentials never leave the machine.

Architecture

  • Rust workspace with three crates: gtasks_core (business logic, synchronization, date parsing, persistence), gtasks_tui (terminal interface with Ratatui) and gtasks_gui (GTK4 / Libadwaita desktop app with Relm4). Both interfaces share the same core.
  • Offline-first sync: local changes are stored in SQLite and flagged as pending, then pushed to Google Tasks over REST when a connection is available, with delta synchronization in the background (Tokio async runtime).
  • Security: OAuth 2.0 with PKCE and a local loopback listener; tokens are stored in the operating system keyring.

Contact

MCR-3 : GitHub · cariat.mathieu@gmail.com

Popular repositories Loading

  1. google-tasks-gui google-tasks-gui Public

    Rust 2 1

  2. function-viewer function-viewer Public

    Visualize a codebase's function call graph as draggable Blueprint-style cards. Python, Rust, TypeScript, and JavaScript supported.

    Python 2

  3. Market-Analyser Market-Analyser Public

    ETF correlation dashboard and portfolio simulator: React + FastAPI, Supabase, yfinance market data.

    Python 1 1

  4. informations informations Public

    JavaScript

  5. .github .github Public

Repositories

Showing 5 of 5 repositories

Top languages

Loading…

Most used topics

Loading…