Releases

FlowState 1.5.3 Latest

Released 28 October 2026

Highlights

• Slow models solve in the background. The window keeps responding while a large model solves: the status bar shows Solving… and the previous results are drawn dimmed until the new ones arrive. Opening the lithium example used to freeze the window for about 2.6 s; the longest frame is now under 50 ms. Exports that need results wait for them, with a Cancel button.
• Drag and drop on Wayland. .flow, .spdx and .spf files can be dropped onto the window under Wayland as well as X11, Windows and macOS.
• Narrow windows. At small window widths or large UI scales the canvas toolbar moves its less-used controls into a ⋯ menu, and the status bar no longer overlaps the flowsheet. The ribbon drops its labels in steps so it fits with any platform's fonts.

Security

• The automation API requires a session token by default. Each user gets a random 256-bit token, stored in api-token in the FlowState settings folder and readable only by that user. Requests without it get 401. flowstate-cli mcp --connect, excel-link and the Python example client read the token on their own, so AI-agent setups keep working unchanged. Your own scripts get the token from Tools ▸ Copy API token or flowstate-cli api-token. Web-query files written by excel-link before this release must be written again. To turn the token off, untick Tools ▸ Require API token or use serve --no-token.
• Store, snap, Flatpak and Mac App Store installs keep their token in their own settings folder; FlowState's tools also try those files, so a flowstate-cli from another package still connects.
• Pages from other web sites can no longer send requests to the automation API (requests with Origin: null are refused).
• Proposals sent through the API can no longer replace the whole document or rewrite its change history.
• The assistant refuses to send an API key over plain HTTP to a remote server.

Fixes

• Saving and files:
  • Saves are atomic: a full disk or a crash during a save leaves the previous file intact.
  • Numbers keep their exact value through save and load.
  • Files with a byte-order mark open; CSV files from Excel and from locales with a decimal comma import correctly, and 1.234,5 is read as 1234.5.
  • Descriptions are no longer renamed every time a file opens; a stale ID counter in a file no longer hands out IDs already in use.
  • Malformed SimaPro CSV or recipe text no longer crashes or hangs the app.
  • One invalid setting no longer resets all settings; Flatpak builds keep settings and crash recovery.
• Results:
  • The same project gives the same results on every solve, and opening a file no longer starts from the previous file's results.
  • Edits made through scenarios, studies or renames no longer change physical results.
  • A maximum equipment size of 0 is treated as no limit; the evaporator condensate temperature is correct; a diverged distillation column reports an error instead of passing invalid numbers on.
  • Kinetic reactions that do not close the mass balance give a warning.
• Economics and environment:
  • A construction schedule of all zeros no longer drops the capital cost.
  • Projects that end before production starts no longer recover working capital or a loan they never had.
  • The cost-only study shortcut follows changes to the cost index.
  • Life-cycle assessment credits power you generate instead of counting it as a burden.
  • With cost evaluation off, exports contain no money columns.
• Interface and performance:
  • An API edit or a late study, estimation or LCA result can no longer land in a different file you have since opened.
  • The app no longer hangs on exit when a local client stalls.
  • Undo history has a memory limit.
  • Building large flowsheets and redrawing large tables is much faster.
  • Dropping a model-tree item onto itself no longer moves it; light-theme text has better contrast; Ctrl+K ranks non-ASCII names correctly.

FlowState 1.5.2

Released 25 October 2026

Changes

• Documentation:
  • The README and the in-app reference lists were revised.
  • The Equations section cites the published textbook literature for scheduling and cost conventions.
• MCP server: the documentation it offers to AI agents is now the user guide only.

FlowState 1.5.1

Released 25 October 2026

A maintenance release with fixes found while packaging 1.5 for Linux.

Fixes

• File dialogs no longer freeze the window. Open, Save and every import and export dialog run off the window thread, so FlowState keeps responding while a dialog is open. This stops the "not responding" warnings on GNOME.
• A file dialog that never appears is handled. If the system file dialog (the desktop portal on Linux) does not answer within 30 s, or fails straight away, FlowState says so and lets you type or paste the file path.
• Symbols render on Linux and macOS. Arrows, triangles and check boxes (→ ▸ ☐) showed as empty boxes, because only Windows loaded a symbol font. FlowState now uses DejaVu Sans, Noto Sans Symbols 2 or Apple Symbols, including the host fonts inside snaps and Flatpaks.
• Slow exports run in the background:
  • scenario comparison export, which solves each scenario;
  • the HTML report while a scenario is being edited;
  • LCA "Add upstream project".
• Typing in large models is smoother. Models that take more than 100 ms to solve re-solve at most every 0.4–1.5 s while you type, not on every keystroke.
• The waterfall chart in Scenarios floats each step with value labels and readable axis labels, and the comparison table uses the full width.
• Section frames after auto-layout: a section split across rows is labelled "(cont.)" and kept together where possible.
• Streams tab: the per batch / per h / per yr toggles no longer overlap "Mass Sankey" in a narrow window.

Packaging

• Snap (Ubuntu and other Linux distributions): sudo snap install flowstate --beta.
• Website: asset URLs carry a content hash, so screenshots and scripts update immediately after a release. A "Support FlowState" section has been added.

FlowState 1.5.0

Released 24 October 2026

FlowState 1.5 follows 1.0.0 directly (there was no 1.1). Highlights:

• SuperPro Designer v15 parity: the new v15 unit procedures, operation options, resources and costing, and an .spf importer for v10–v15 files (all 194 SuperPro v15 examples import).
• Plain-language names for every unit procedure and operation; search still finds the old names.
• A COMSOL-inspired interface: Model tree, Settings pane with the model equations, contextual toolbar ("ribbon-lite"), bottom dock, line-art PFD drawing style and an illustrated icon family.
• AI and automation: an MCP server for AI agents, reviewable changesets for every AI or API edit, and an in-app assistant with the provider of your choice (offline by default).
• Analysis: study engine v2, TEA presets (NREL DCFROR, MACRS, IRR / minimum selling price), a scenario workspace, quick GWP and LCA additions, parameter estimation with D-optimal design, stiff dynamics with rigorous batch distillation, site fit and recipe scaling, and an ISA-88 recipe text view. Existing .flow files open unchanged and give the same results; new data is left out of files while unused.

Modelling

• New unit procedures: heat pump (Carnot-grade COP), standalone steam ejector, gas membrane (cross-flow, permeance in GPU), cooling tower (Merkel, blowdown, drift), and a shared combustion core for rotary kiln, gas turbine, fuel cell and gasifier; PEM electrolyser and electrowinning presets. A new example uses the utility units (39 examples).
• Operation upgrades: pH targets in Charge to target and set-point mixing, advanced split grids (S % / w %) with a degrees-of-freedom check, and the throughput analysis (utilisation indices, batch-size potential).
• Dynamics: a stiff BDF/DAE integrator inside operations, with rigorous batch distillation on top of it. The kinetic integrator default is Auto (non-stiff runs stay explicit).
• Thermodynamics: per-component enthalpy reference (normal boiling point and reference phase), reaction heat from formation enthalpies, LHV from ΔHf, and a "vapour data unknown" flag.
• Materials and properties: ChemSep v8.3 databank (431 compounds, searchable by name, CAS or formula); DIPPR / ChemSep correlations with validity ranges and a Property data window; transport properties on every stream; NREL biomass and biologics component sets. Opt-in, off by default: estimated transport properties in sizing and kLa, Laliberté salt solutions, temperature-dependent Henry constants with salting-out, ChemSep binary parameters. ChemSep © Harry Kooijman and Ross Taylor, Artistic License 2.0.
• Resources and costing: stream identity (regular, heat-transfer agent, or withdrawn from storage); agents costed as use-and-return or use-and-dispose, with regeneration, losses, penalty, waste treatment and credits; auxiliary equipment types (transfer panels, vacuum pumps, user types) with an AUX report; user cost models with several size ranges; zero mass / flow thresholds; dry-solids components (dry solids, dry solids % and moisture in stream tables and reports).
• Parameter estimation from CSV experiments (Levenberg–Marquardt with sensitivities, confidence intervals, correlation matrix, parity and residual plots; fitted values written back with provenance) and D-optimal design of experiments.
• Site fit and recipe scaling: a facility database (vessels, minimum stir volumes, materials of construction, ratings, utilities); fit problems in the Problems pane; scale to the largest vessel, to a target amount or by ×N.
• ISA-88 recipe text view, generated from the model; edits are parsed back and can be undone.
• Plain equipment names: procedures are named after their equipment in US spelling ("In a Fermentor" → Fermenter (stirred, aerated), "Generic Box" → Custom Unit (black box)); operations read as verbs ("Pull In" → Charge to target); 14 palette categories showing mode and default recipe; P&ID tag letters for new equipment; search knows SuperPro, Aspen and DWSIM names. Stored keys and file formats are unchanged.

Interface

• Model tree (left dock, next to the Library): the whole model with problem badges, a filter, right-click commands, drag to reorder operations, and F2 rename. Renaming in the tree or the Settings pane uses the same rename as the CLI and API, so references to the name follow it (one undo step).
• Settings pane: breadcrumb, Copy API path, Show advanced, and an Equations section with literature references; numeric fields show their valid range and a definition.
• Contextual toolbar whose groups follow the tab and selection, and a bottom dock with Problems, Log, Progress and Results. Panes can be shown, hidden and resized, and the layout is remembered.
• Command palette (Ctrl+K) over commands, procedures, streams, components, examples and variable paths.
• Flowsheet drawing: right-angle stream routes in lanes, readable labels, grid snapping with alignment guides, live overlays (mass-flow Sankey, temperature, cost, CO₂e, utilisation with the bottleneck), a line-art PFD style (used by default for exports) and an illustrated icon family.
• Flowsheet editing: marquee and Shift/Ctrl multi-select, group move, arrow-key nudge, copy / cut / paste with the streams in between (also into another process), align and distribute, auto-layout left to right (one undo step), drop a procedure on a stream to insert it, collapsible section frames, a minimap above 20 procedures, and Space+drag panning.
• Tables: paste ranges from Excel (unit-aware, one undo step) into the stream and equipment tables and a new Material prices table; per batch / per h / per yr and amounts / mass / mole fraction views of the stream table; sortable headers, right-aligned numbers, Ctrl+C to Excel.
• Inputs: type values with units (3 barg, 98.6 F), hover for other units, unit sets (Metric, SI, US, Biotech), and Alt-drag a number to re-solve live.
• Problems pane for all errors and warnings, with why / how-to-fix explanations and one-click fixes.
• Never lose work: "Save changes?" before a document is closed or replaced, viewing no longer marks a file as changed, and crash-safe autosave with recovery on start-up.
• Look and feel: WCAG-contrast colours, Dark / Light / Follow system, a UI scale setting, a Learn panel on the welcome screen, Sankey diagrams for mass, cost and impact, and a four-step first-run tour (Help ▸ Take the tour).

Automation and AI

• Command registry behind menus, toolbar, palette, HTTP API, MCP and the assistant. Workflow commands: switch a procedure within its family, Insert file (merge projects), and rename / replace across the model from the CLI, API and MCP.
• MCP server (flowstate-cli mcp; --connect works on the document open in the app) with read-only and editing tools and correct hints.
• Reviewable changesets: edits from MCP (review mode by default) or POST /changeset/propose wait in the Proposed changes panel with current → proposed values, rationale and a shadow-solve preview of the key figures. Accept a selection as one undo step; accepted edits show "set by AI". New endpoints GET /changeset, POST /changeset/accept|reject; MCP tools propose_changes, get_changeset.
• In-app assistant (View ▸ Assistant): Anthropic, any OpenAI-compatible server (including local Ollama, LM Studio and vLLM) or Azure OpenAI. Nothing is sent until a provider is set up and you press Send; context is attached only with right-click ▸ Send to assistant. It runs the MCP tools in-process and can only propose edits. API keys are kept in the OS credential store, never in .flow files.

Economics and environment

• Study engine v2: Latin hypercube, Sobol and Morris sampling, correlated inputs (Iman–Conover), Spearman tornado, a cost-only fast path, fan charts and density plots.
• TEA conventions: NREL / Humbird DCFROR preset (MACRS, construction schedule, start-up fractions, working capital), IRR and minimum-selling-price solve, and a cost-item table.
• Scenario workspace (Tasks ▸ Scenarios…): named cases of changed inputs saved in the file; edit in a scenario with the base values shown; Run all with a comparison table; a waterfall per key figure; what-if chips ("titer ×2", "price +10 %"); promote to base; CSV / Excel export and a Scenarios section in reports. The economics named overrides are now scenarios, and the studies window is Tornado & scale-up.
• Materials cost split into raw material, cleaning agent and heat-transfer agent.
• Quick GWP report (kg CO₂e per functional unit without an LCA dataset) and an energy report per section, utility and operation with an enthalpy balance per procedure.
• LCA additions: energy allocation, displacement credits, characterisation factors per utility agent, construction and transport items.
• Emissions: EPA drying and vacuum-drying models, an EPA-MACT section, and emission paths through control devices (condenser, oxidizer, …) in the Environment tab and the HTML, Excel and Word reports.
• Verified coefficients: AP-42 boiler and gas-turbine factors, the Seider ejector citation, Schumpe HCO₃⁻, the PMO pasteurisation preset (C. burnetii, z = 4.34 °C), and sourced cost laws for the gas turbine, fuel cell, heat pump, gas membrane and gasifier.

Import and export

• SuperPro .spf import for v10–v15: the schema stamp is read from the header and fields are gated by version, streams and ports are re-decoded, and every SuperPro icon class is mapped. A file that cannot be read reports an error instead of producing an empty model. The import guide has a "Renamed from SuperPro" table.
• Reports: GWP, energy, EPA and emission paths, resources (AUX equipment, dry solids), materials-cost split and scenario sections in HTML, Excel and Word, with a PFD-style flowsheet.

Fixes

• The status-bar solve time is correct, including the first solve.
• Gasifier Biomass LHV on the N₂ basis; the large-boiler AP-42 row is labelled FGR and uses the uncontrolled N₂O factor.
• SuperPro imports without cost data no longer report meaningless costs (economics are off for them; balances are unchanged).
• Labels: the Tornado & scale-up window title matches its command, the Pasteurizer note names the Thermal lethality section, and the Stream identity hint names the Resources section; the dry-solid flag has a field in Pure Components.

Known limitations

• The rotary-kiln cost law has no published source yet; the Schumpe ion parameters for H₂PO₄⁻, HPO₄²⁻ and PO₄³⁻ are unverified.
• Operation parameters are not decoded from v15 .spf files, so imported models use calibrated splits. PulpAndPaper_v15's recycle does not converge within the iteration limit, and VitaminC_v15's BMED and crystallisation train is not reproduced.
• Adding or removing items while editing a scenario is not recorded in the scenario (a warning says so).
• The assistant does not stream replies; answer quality depends on the provider and model you choose.

FlowState 1.0.0

Released 21 October 2026

First release of FlowState: a desktop application and command-line tool for process simulation, scheduling and techno-economic evaluation.

Modelling

• Flowsheet and unit procedures. 98 unit procedures for batch, continuous and hybrid plants. Recipes are built from operations, with material and energy balances solved sequentially; recycle loops use automatic tear selection and Wegstein acceleration.
• Thermodynamics. Ideal, NRTL, Wilson and UNIQUAC activity models, Peng–Robinson and SRK equations of state, three-phase VLLE flash and IF97 steam tables.
• Reactions. Stoichiometric reactions on a mass or molar basis, with extents set by conversion, final concentration, residual amount or equilibrium. Also kinetics, CSTR and plug-flow reactors, fed-batch, perfusion, oxygen transfer, and intracellular products.
• Separations.
  • Columns and evaporation: rigorous MESH columns, FUG shortcut design, batch distillation, and multi-effect evaporation with MVR/TVR.
  • Liquid and solid separations: extraction, crystallisation from solubility data, and dryers that dry to a target moisture.
  • Membranes and adsorption: membranes, reverse osmosis, ion exchange, carbon beds, and chromatography including PCC and SMB.
  • Other: electrodialysis, electrolysis and PSA.
• Utilities and environment. Gas compressors, fans, vacuum pumps and air-supply trains. Heat exchangers, pinch analysis, boilers, turbines and combined heat and power. Wastewater treatment and pH control.
• Scheduling. Cycle-time analysis, campaigns, shared and staggered equipment, CIP/SIP, inventories and resource limits.
• Economics. Equipment sizing and costing; capital and operating costs; consumables and single-use equipment; QC; credits; financing and depreciation; NPV, IRR and minimum selling price; tornado charts and scenario comparison.
• Environmental assessment. Emissions, process mass intensity, and life-cycle inventory and impacts, with SimaPro CSV import and export.

Examples and interoperability

• Examples Library. 38 models, including reproductions of published SuperPro Designer case studies with their comparison documents.
• SuperPro Designer import. Imports SuperPro Designer .spf files.
• Reports. Excel, Word, RTF, SVG and HTML reports, plus electronic batch sheets.
• Automation. flowstate-cli (simulate, variable get/set, sweeps, Monte Carlo, reports, import, LCA) and a local HTTP automation API that works with Excel, Power Query and Python.