natolambert/colloquium

A markdown native slides tool for academics building with agents.

What it solves

Colloquium simplifies the creation of research presentations by allowing users to write slides in Markdown. It eliminates the need for manual slide design in GUI tools, providing a Git-friendly workflow that produces a single, self-contained HTML file for presenting.

How it works

Users write their content in a Markdown file with a YAML frontmatter block for global configuration (like fonts, themes, and bibliography). Slides are separated by --- delimiters. The tool converts this Markdown into a formatted HTML presentation using a build pipeline that supports advanced layouts, math rendering via KaTeX, and code highlighting via highlight.js. It also includes a CLI for building, serving with live reload, and exporting to PDF or PPTX.

Who it’s for

Researchers and academics who want a version-controlled, programmatic way to create high-quality talks, especially those needing to include complex math, citations from .bib files, and technical diagrams.

Highlights

  • Research-centric features: Built-in support for BibTeX citations, automatic reference slides, and KaTeX math.
  • AI-ready output: A capture command that exports slides as individual PNGs, specifically designed for AI agents to visually review presentations.
  • Specialized content blocks: Custom YAML-based blocks for rendering LLM-style chat conversations, data charts (via Chart.js), and callout boxes.
  • Flexible Layouts: Support for multi-column and multi-row grids with arbitrary numeric ratios, and various slide layouts (e.g., section breaks, image-left/right).
  • Developer-friendly: Git-compatible source files and a dev server with live reload.

Related

  • Project
    javifalces/HFTFramework

    A high-frequency trading framework in Java and Python that enables L2 tick-data backtesting and live market execution using a unified codebase.

  • Project
    facebookresearch/atlas-lean

    ATLAS is a research project that uses large language models to automatically translate textbook mathematics into formal Lean 4 libraries, providing reusable, machine‑checkable mathematics for theorem proving and AI research.

  • Project
    netdur/llama_cpp_dart

    A Dart FFI binding for llama.cpp that enables running LLMs, vision, and audio models locally within Flutter apps on iOS and Android.

  • Project
    openxla/tokamax

    A library of custom accelerator kernels for NVIDIA GPUs and Google TPUs built on JAX and Pallas, providing optimized implementations of common AI operations and autotuning tools.