0.8.0

RAPTOR Studio 0.8.0: what's new

RAPTOR Studio is the first public release of a port of Raptor 7, the reliability block diagram (RBD) and Monte Carlo simulator written at ARINC and released as open source by Booz Allen Hamilton. Raptor 7 was built on a proprietary runtime and ran only on Windows. RAPTOR Studio is written in Python with a Qt desktop interface, so it runs on Windows, macOS and Linux. It also comes as a command-line tool, a scripting API and an MCP server, so a model can be built, simulated and queried by a script or an AI assistant as well as by hand.

Like Raptor 7, it is released under the Booz Allen Public License (BAPL): anyone may use, modify and share it, but a commercial entity may use it only for its internal business and for services to government and non-profit organisations, and nobody may charge a fee for it without Booz Allen Hamilton's permission. New code in the port is Apache-2.0; see the README's "Licence & credits".

The aim is fidelity first. The simulation engine was ported from Raptor 7's source and checked against what the real Raptor 7 writes. More than 200 captured models are compared: event logs line by line, report files, the Single Point Failure file, the Failure Effects View log and every random number drawn. Every comparison matches, including the order of events that happen at the same instant, which was worked out with debugger traces of Raptor 7. The latest scan compared the event logs of 4,318 runs and found none different.

Compared with Raptor 7

	Raptor 7	RAPTOR Studio
Platforms	Windows	Windows, macOS and Linux (installers, or pip install)
Model files	.rbd	.rbd opens and saves as before; .rbds is the same model as readable JSON, and Differences (raptor diff) says in words what changed between two versions
Results	Report files and on-screen tables	The same Raptor 7-format report files, plus one readable tab per section with plain names and units; export to PDF (cover, contents, diagrams, model data, results), HTML, spreadsheet (with a Run Info sheet), CSV or JSON
Which part matters most	Weak link colouring and the Single Point Failure file	Both kept, plus importance measures: Birnbaum, RAW, RRW and Vesely-Fussell (IEC 61078 Annex D), each with its sampling error, an Importance tab and diagram colouring
Running a batch	One process	Split across your cores (-j N); reproducible for a given core count, and one core reproduces Raptor 7's own random sequence
Stepping through a run	Step and Jump	Step event by event, jump to the next run, pause on a condition, and see what each step changed
Scripting and automation	—	A command line (run, sweep, compare, diff, estimate, validate, spf, fev, step, extract, render) with machine-readable results, and a Python API
AI assistants	—	An MCP server with 64 tools to build, edit, validate, simulate, compare and explain models
Documentation	Manual and help file	A 23-chapter tutorial book with a model for each chapter, a field-by-field reference manual and a formulas page; F1 opens the right section

What stays the same: the numbers. The engine reproduces Raptor 7's results, including its random-number generator and its surprising behaviours, and models made in Raptor 7 open here.

Also new

• Undo and redo, mass edit, block defaults, auto layout, SVG diagram export, Extract Hierarchy (run one hierarchy on its own) and named, reusable rates.
• Run-time estimate before you start, in the app, on the command line and from an assistant, including the extra batches an importance analysis needs.
• Inputs checked before a run with Raptor 7's own refusals, plus checks for model shapes that would crash Raptor 7 when opened.
• Reproducible runs: the seed, software version and model digest are saved with each result set.
• Parameter sweeps and comparisons from the app, the command line or an assistant.
• Each model in its own window, with its own undo history and results; light and dark themes, annotations and key tips on every ribbon button (press Alt).
• Worked examples on the start page, from a three-block lantern to a hospital's emergency power, a water-treatment works, a data centre and a phased airliner flight.

Highlights

• Importance measures. Switch on Run > Analysis > Importance measures (or raptor run --importance both, or importance on the MCP simulate tool) to rank blocks. The default ranking is Vesely-Fussell, because Birnbaum cannot separate the blocks of a series system. Importance is estimated by re-simulating the model with each block made unable to fail and with it dead, so it costs one or two extra batches per block.
• Faster. The scheduler, layout, validation and file loading have been tuned so that models with thousands of blocks are practical to build, open and run, and batches use several cores. Event logs and reports are unchanged by the speed-ups.
• MCP improvements. Long runs go to the background (start_run, run_status, cancel_run) and a finished run is read back in one call. Tool outputs have schemas and unknown arguments are refused. The server confines file access to its workspace and has size and concurrency limits; over HTTP it needs a bearer token and refuses plain HTTP from other machines. New tools manage pools, resources and triggers, compare models and export tables.
• IEC 61078. docs/IEC61078.md states clause by clause what is and is not implemented. It is a statement of capability, not a certification.

Known limitations

• Importance is per block only. A hierarchy is not yet an element of the analysis; to judge a subsystem, read the blocks inside it or run it on its own (raptor extract).
• Importance is a Monte Carlo estimate, not an analytical factor. The tab reports the sampling error, and warns that the ranking means little when the baseline availability is above 99.9%; raise the number of runs or the run length.
• The macOS and Windows builds are unsigned. Expect Gatekeeper or SmartScreen to ask for confirmation; the install notes below say how to proceed.
• The desktop installers carry the application only. The raptor command line and the MCP server come from the pip package.
• Not implemented: minimal cut and tie sets, Boolean, BDD and analytical evaluation, repeated blocks and non-coherent diagrams.

Install

• Desktop: download the installer for your platform below (Windows setup or portable zip, macOS Apple Silicon or Intel .dmg, Linux .deb or AppImage) and check it against SHA256SUMS.txt.
• Python (application and command line): pip install raptor-studio (Python 3.12 or later).
• MCP server for AI assistants: pip install "raptor-studio[mcp]", or run it without installing: uvx raptor-studio-mcp --workspace ~/raptor-models.