Mudyla - Multimodal Dynamic Launcher

A script orchestrator: define graphs of Python/Bash/etc actions in Markdown files and run them in parallel under Nix environments.

Totally Claude'd.

Based on some ideas from DIStage Dependency Injection, Grandmaster Meta Build System and ix package manager.

Successor of mobala

If you use Scala and SBT, Mudyla works well with Squish.

An example of a real project using this gloomy tool: Baboon.

Terminal interfaces

The default pure logger presents one interactive Actions: dependency graph, with status, action, context, elapsed time and the latest log line. Scroll up for compact context summaries and the same run information printed before execution. The graph keeps execution-plan order; the cursor sits to its left. Supported terminals add a subtle selection background derived from their theme, preserving text colors. Arrow keys select and reveal any action. Both pure and table support action selection, logs, metadata, outputs, source, and action input. Table uses content-sized live columns, with counts below the rows and the same subtle theme-aware selection. --plan table selects the full static plan table. Pure and table display inline by default, leaving the mouse wheel to scroll terminal history. Shrinking the terminal during inline updates can leave previous frame fragments in history. --fullscreen uses the alternate screen during execution and exits automatically. --it / --interactive opens fullscreen and keeps the completed view open. Opening details from an inline view temporarily enters fullscreen; returning restores the same action selection. --plan auto|tree|dag|table selects the plan layout in any logger. The default, auto, uses DAG for fewer than 30 planned actions and table for 30 or more, or when DAG layout reaches its 1-second deadline. Explicit tree, dag, and table selections always use that layout. DAG uses --plan-dag-solver grid-auto by default. --plan-minimize true|false controls redundant connections in tree and DAG views (default: true); table plans retain all dependency information. Press i in the Actions list or stdout view to send input to a selected running action:

mdl :build
mdl --logger table :build
mdl --plan tree :build          # Separate Plan tree and flat Actions list
mdl --plan table :build         # Full static Plan table and flat Actions list
mdl --fullscreen :build         # Fullscreen view; exit after completion
mdl --it :build                 # Fullscreen view; keep open after completion
mdl --logger simple :build      # Append-only progress and compact Plan
mdl --logger verbose :build     # Commands and immediate action output
mdl --logger github :build      # GitHub Actions groups and streaming
mdl --logger teamcity :build    # TeamCity blocks; --par adds isolated action flows

--simple-log, --verbose, --github-actions, and --teamcity are aliases for those modes; --logger raw remains an alias for simple. Simple, verbose, GitHub and TeamCity share the compact Run info, Plan, Result and typed outputs, with plain redirected output. Verbose, GitHub and TeamCity execute sequentially by default; --par enables parallel execution. Parallel verbose logs carry action@context: prefixes; sequential verbose and GitHub preserve unprefixed child output. Capture files remain unchanged.

The action definitions, arguments, and execution workflow stay the same in every view. See the logger and keyboard guide.

Try the local 20-second demo in this repository:

.venv/bin/python -m mudyla --without-nix --par --it :demo-interactive

Use j/k to select an action, Enter for logs, q to return, and q again to stop execution or close the completed view. Mouse wheel and page keys scroll; Home shows the run information. Add --logger table to compare. Pure metadata and outputs use compact labelled fields; v toggles their original JSON. Contexts use the same @name throughout the plan, checklist and details.

Views

Pure DAG (light / dark) · Table · Pure tree · Simple

Details

Logs · Outputs · Metadata · Source

Documentation

📚 Read the Full Documentation

• Installation
• Getting Started
• Core Concepts (Actions, Dependencies, Contexts)
• Reference (CLI, Syntax, API)

Demo

• Parallel build
• Checkpoint recovery
• Weak dependencies
• Context reduction

Features

• Markdown-based action definitions: Define actions in readable Markdown files
• Multi-language support: Write actions in Bash or Python
• Dependency graph execution: Automatic dependency resolution and parallel execution
• Multi-version actions: Different implementations based on axis values (e.g., build-mode)
• Multi-context execution: Run the same action multiple times with different configurations
• Axis wildcards: Use * and prefix* patterns to run actions across multiple axis values
• Nix integration: All actions run in Nix development environment (optional on Windows)
• Checkpoint recovery: Resume from previous runs with --continue flag
• Rich CLI output: Beautiful tables, execution plans, and progress tracking

Quick Install

# Install with pipx (recommended)
pipx install mudyla

# Or run with Nix
nix run github:7mind/mudyla -- --help

See Installation Guide for more details.

Quick Start

1. Create .mdl/defs/actions.md:

   # action: hello-world
   ```bash
   echo "Hello, World!"
   ret message:string=Hello
   ```
   

2. Run:

   mdl :hello-world
   

See Getting Started for a full tutorial.

Testing

Mudyla uses pytest. Run ./run-tests.sh to execute the suite. See TESTING.md for details.

License

MIT