Roadmap
Every capability this documentation describes and where it stands — shipped, partial, planned, deferred, or deliberately not planned. One table per product, generated from the status registry.
This is a roadmap that starts with what is already built, because that is the part most people are actually asking about. These products are under active development, and this documentation describes some things that are built and some that are designed — this page is the authority on which is which, in both directions.
Read down a table and the shipped rows come first, then the gaps, then what is coming, then what has been deliberately put off to a named release, then what is not coming at all.
How to read it#
| State | Means |
|---|---|
| Shipped | Built and in the product today. Every shipped claim is backed by a named file in the product's source, and a script re-checks that it is still there |
| Partial | Built, with a stated gap. The gap is named in the notes column rather than left for you to discover |
| Partial · by design | The same, where the gap is a boundary rather than unfinished work — something that will not be closed because closing it would be the wrong thing. Read the note for what the boundary is |
| Planned | Designed but not built. Documented so the shape is settled, and badged everywhere it appears |
| Deferred | Built or specified, and deliberately not in this release. The state carries the release it is deferred to, because "deferred" without a destination is just "planned" in a better font |
| Not planned | A deliberate refusal, with its reason. Documented so nobody spends three weeks expecting it |
Everywhere else on this site, a capability that is not shipped carries a badge saying so — inline, next to the sentence that describes it. Nothing on these pages describes a capability that is not listed below, because the site does not build if it does.
Versions are the products' own. None of the three has had a public release yet, so every version below is a pre-release, and the dates are when the registry was last checked against the source.
Amalgative Draft#
Version 0.5.0 · beta. Last checked against the source on 2026-08-18. Not yet generally available. The 2D core and most of the 3D chapter are built.
| Capability | State | Notes |
|---|---|---|
| Command line, aliases and keyword prompts | Shipped | |
| Arithmetic in a length field | Shipped | |
| Draw commands | Shipped | |
| Modify commands | Shipped | |
| Object snap, ortho and polar | Shipped | |
| Extension snap with endpoint acquisition | Shipped | |
| Isometric drafting (ISOPLANE) | Shipped | |
| Snapping and ortho in three dimensions | Shipped | |
| Layers, linetypes and lineweights | Shipped | |
| Blocks, attributes and WBLOCK | Shipped | |
| External references | Shipped | |
| Hatching | Shipped | |
| Text, dimensions, leaders and tables | Shipped | |
| Layouts, viewports and plotting | Shipped | |
| Plot to PDF at true paper size | Shipped | |
| The 3D pane, solids and booleans | Shipped | |
| 3D draw commands | Shipped | |
| The .amdr native format | Shipped | |
| DXF read and write | Shipped | |
| DWG read through an external converter | Shipped | Reads only, and only where a supported external converter is installed — Draft bundles none of them. That is the shipped design rather than a shortfall: converters are detected when present, and a refusal names every route it tried with the licence each one carries. DWG write is not planned, so there is nothing to write with. |
| settings.json and keybindings.json | Shipped | |
| Drawing version log and merge | Shipped | |
| Parametric node graph | Shipped | Deferred past 1.0 on 2026-08-19 and built anyway, shipping 2026-09-01. Both scopes: the drawing's own graph, the editor as a third viewport mode with every gesture routed through a registry command, and component graphs end to end. Eleven graph.* commands; the activity-rail panel has been stage "shipped" since 2026-09-04. This entry read "deferred" with the note "Its panel is listed in the rail, disabled" until 2026-09-10 — by then every clause of that sentence was false. |
| Draft collaboration client | Deferred · 1.1 | Deferred to 1.1 by decision. The wire protocol, the server and the client-side three-way merge all exist — the server appends and orders, and the merge runs on the CLIENT. One limit is real: a single shared drawing per application instance, and a second SHARE is refused until UNSHARE. Real merge semantics is correctness-critical — getting it wrong LOSES a drafter’s work rather than degrading gracefully — so it is not being rushed into the run to 1.0. |
| Plugin API | Deferred · past 1.0 | Deferred past 1.0 by decision. The manifest, lifecycle and sandbox are specified and those specifications stand; there is no plugin host in the product. |
| System definition documents (.amsys) — parametric geometry Draft authors and Build consumes | Deferred · past 1.0 | Specified by Build (“Draft — System Definition Spec”, 2026-08-28) and answered by Draft in MESSAGE_TO_BUILD.md, which resolves all six of the spec's open questions against source. Deferred because it is six slices of NEW SCOPE and the ladder puts “no new scope” at W4 and “feature complete” at 0.9.0; Draft's remaining gate to 1.0 is signing and the updater, which is infrastructure rather than features. Its two prerequisites are the size of the job. The formula evaluator was one of them and its ARITHMETIC half now exists (engine/units/formula.ts, 2026-08-29 — a drafter can type =1200/3 into a length field); what .amsys needs on top is a symbol table, so that a formula can NAME a parameter, and that does not exist. The second prerequisite is untouched: there is NO constraint solver — Draft's associativity runs geometry → dimension, and .amsys needs parameter → dimension → geometry, which is a different machine rather than a fuller version of the same one. What DOES exist and shortens it: engine/deps/graph.ts already recomputes transitively in dependency order and RETURNS cycles rather than truncating, which closes the spec's §3.2 and §6.2 cycle requirements outright; arrayEntity.ts is already a live parametric entity whose named parameters drive derived geometry; the .amdr container and the headless command bus are both reusable as they are. ⚠ Draft passes the spec's §1 opacity test today — no building sense of wall, door, schedule or storey appears in code. |
| Precise B-rep solids and NURBS surfaces | Not planned | Draft models with parametric solids and meshes. Curvature is faceted, and the consequences are stated rather than hidden. |
| Edge fillets and chamfers on solids | Not planned | Requires B-rep. The most conspicuous absence, deliberately visible. |
| Surface bodies | Not planned | No surface representation exists. Thickness covers the simple cases; open profiles are refused with the reason. |
| Materials, lights and ray tracing | Not planned | A separate product with a separate audience. Shading uses one fixed light. |
| STEP and IGES | Not planned | B-rep interchange formats, with no B-rep to write. |
| DWG write | Not planned | The value of DWG is read. Everything reads DXF, which Draft writes. |
| Proprietary BIM and sketch-model geometry | Not planned | Such files are identified and described — format, version, worksharing state — never converted. There is no permissively-licensed reader for their geometry. |
| Classic TRIM and EXTEND boundary selection | Not planned | Quick mode only — every visible entity is an implicit boundary. Added only if users ask. |
Amalgative Redline#
Version 0.7.0 · beta. Last checked against the source on 2026-08-18. Not yet generally available. The editor core, tools, measurement and PDF round trip are built.
| Capability | State | Notes |
|---|---|---|
| Three-canvas viewport, pan and zoom | Shipped | |
| Draw, annotate and navigate tools | Shipped | |
| Command bus with undo and redo | Shipped | |
| Measurement and takeoff | Shipped | |
| Calibration, units and the scale cascade | Shipped | |
| Dynamic Fill | Shipped | |
| Area cutouts | Shipped | |
| Dynamic takeoff legends | Shipped | |
| Spaces | Shipped | |
| Page operations | Shipped | |
| Bookmarks | Shipped | |
| Sets — the virtual binder | Shipped | |
| Batch operations | Shipped | |
| Export as real PDF annotations with appearance streams | Shipped | |
| Import a PDF's own annotations as editable markups | Shipped | |
| XFDF import and export | Shipped | |
| Hyperlinks and batch links | Shipped | |
| Flags | Shipped | |
| Form fields | Shipped | |
| Text and visual search | Shipped | |
| Document compare | Shipped | |
| OCR | Shipped | |
| Rights-managed document detection | Shipped | |
| The conductor — one schema-validated action surface | Shipped | |
| MCP bridge | Shipped | |
| Local single-user version history | Shipped | |
| Signature fields | Shipped | |
| Action Wizard macro automation | Shipped | Record a sequence of conductor actions as a named macro, inspect it, and replay it — five actions on the command surface: start and finish recording, recording status, inspect and run. |
| Printing, with a composed-bytes preview and a 1:1 default | Partial | Windows spooling does not honour driver options — the shell print verb takes none — and the dialog says so. The composed bytes are proved 1:1 exactly: a 24x36 ARCH D source page measures 1728 x 2592 pt and composing at actual size returns the same, delta 0.0 pt. So if paper disagrees, the fault is the printer driver rather than the composition. What remains outstanding is the physical-ruler check against real paper. |
| AI command centre | Partial | The desktop build reaches any provider. The browser build serves only the keyless local runtimes — Ollama and LM Studio — because §H1.6 gate 5 keeps API keys out of the webview and a page has no vault to hold them in. Those runtimes must also be configured to allow the page's origin: LM Studio's CORS toggle, or OLLAMA_ORIGINS. Without that the browser sends a preflight that the runtime rejects, because a JSON content type makes the preflight compulsory. The gateway, adapters and agent loop are tested at their parts and seams. |
| Digital IDs — signing with your own certificate | Planned | Needs a cryptographic stack that does not exist in the application yet. Signature FIELDS are a different, shipped feature. |
| Quantity Link — live takeoff totals in a spreadsheet | Planned | The local query endpoint and the spreadsheet add-in are designed, not built. |
| Collaboration client | Deferred · 1.1 | Deferred to 1.1 by decision, not outstanding. The servers are built and tested and the Collaboration panel is the UI seam; nothing in the Redline app dials them, and nothing will before 1.0. |
| 3D PDF (PRC/U3D) rendering | Not planned | PRC is effectively undocumented and has no open parser. The model is PRESERVED through a round trip; it is not displayed. |
| Enterprise DRM connector | Not planned | Needs a cloud tenant registration and a platform-specific SDK that breaks the Linux-primary rule. The DETECTOR is shipped. |
| CAD virtual printer driver | Not planned | A per-OS kernel-level component, out of scope for the desktop application. |
Amalgative Sequence#
Version 0.5.0 · beta. Last checked against the source on 2026-08-18. Beta. Scheduling, baselines, resources, cost, look-ahead, scenarios and impact/recovery are built. Documents, Automation and schedule export are not.
| Capability | State | Notes |
|---|---|---|
| Activities, logic and the critical path | Shipped | |
| Driving path and longest path | Shipped | |
| Schedule quality checks | Shipped | |
| Baselines and variance | Shipped | |
| Resource pool, assignments and capacity | Shipped | |
| Resource levelling | Shipped | Desktop only, and it proposes rather than applies — you review the moves before they land. |
| Cost, accruals and cost codes | Shipped | |
| Earned value and forecasting | Shipped | |
| Look-ahead windows | Shipped | |
| Scenarios: fork, diff and merge | Shipped | |
| Compression and recovery | Shipped | |
| Saved views and filters | Shipped | |
| Conditional formatting rules | Shipped | |
| Working calendars, exceptions and non-working time | Shipped | |
| Hour precision: sub-day durations and shift windows | Shipped | |
| Threaded comments and mentions | Shipped | |
| Notifications and alert thresholds | Shipped | |
| Activity history | Shipped | |
| The .amsq project file | Shipped | |
| Delimited text import | Shipped | |
| Output templates, and the two shipped families | Shipped | |
| Page setup, fit and tiling | Shipped | |
| Bar and milestone styles | Shipped | |
| Palettes, with print-safe and monochrome variants | Shipped | |
| EULA acceptance and beta expiry | Shipped | |
| Permissions | Partial · by design | ⚠ A schedule with no room — or a copy of the file — carries no restrictions at all: anyone holding it has the whole thing. Roles, scope precedence, dynamic views with per-column grants, row filters and grant reporting are all authored here; evaluation is the server's, per reader, per row, per column, when somebody joins the room, and a client-side filter would be a UI convenience wearing a permission's clothes. Group membership comes from identity and is not held here. No seat or tenant administration, and pushing rules replaces the whole set — two admins pushing at once means last write wins, silently. |
| Plot output — PDF, PNG and SVG | Partial | All three emitters ship, and the PDF is vector and multi-page. ⚠ Its text is NO LONGER Latin-1 (2026-08-19): Liberation Sans is embedded as a Type0/CIDFontType2 with Identity-H and a /ToUnicode map, so Greek, Cyrillic and Central European letters print AND remain selectable and searchable — Polish, Turkish and Czech checked against the emitter and extracted back out with pdftotext. ⚠⚠ CJK still does not print, and it is NOT a missing font: the shipped face carries no CJK glyphs, and every CJK face on a typical machine is a ttcf collection of OTTO (CFF) fonts that this pipeline cannot read either. Closing it needs a second emitter path (/FontFile3, CIDFontType0) — deferred past 1.0, see deferredGap. ⚠ A brand typeface now PLOTS: output/fontLibrary.ts resolves the template's family stack and output/pdf.ts keys embedded faces by family, so one document may carry two typefaces; the desktop shell enumerates installed families (fonts_list). ⚠ The font is SUBSET before embedding (subsetFont, called from output/pdf.ts) — a plot uses ~75 of Liberation's 2,620 glyphs, so a deliverable is ~44 KB rather than ~415 KB. |
| Output timescale, period bands and emphasis | Partial | Three tiers, period bands, non-working shading, the data-date line, the progress line, and curtain and highlight all draw. A template's own presentational bands are authored too — they are additional to the sheet's, which stay on the sheet so the Gantt and the page cannot disagree about where a phase runs. ⚠ The one shortfall is media underlays: they need a media layer, and there is no asset storage anywhere in the product for one to be built on, so this is a dependency rather than an oversight. |
| The annotation layer, legend and title block | Partial | Text, notes, callouts with leaders, arrows, lines, rectangles and ellipses, anchored to the page, to a row or to a date, plus the generated legend, the title block, a north point with an authored bearing, and a page-anchored LOGO carried as template data. ⚠⚠ GENERAL IMAGERY IS A BOUNDARY, not unfinished work: D13 resolved document storage to a distinct product, Amalgative Storage, and plan.md A8 puts blob storage there — "Sequence does document control, the server does document storage". Building an asset store here would be the bug, the same shape as permissions' server-side enforcement. A logo is excluded from that boundary by ruling, and the exclusion is CHECKABLE rather than merely stated: a stored image is capped at 512 KB, normalised to the one PNG form a PDF can embed before it is ever stored, and read back through the plotter's own reader before it is accepted. ⚠ Transparency is flattened onto the page colour at authoring time. ⚠ No QR code to a live view — the one clause that is deferred rather than bounded; it needs the live-artifact path sequence.exportSchedule's feed also waits on, since §14.3's published links serve a frozen artifact. |
| Schedule data export | Partial | CSV, iCalendar and .xlsx all ship. The workbook carries the hierarchy as a real outline level and a frozen bold header. ⚠ The one shortfall against the specification is the calendar: §12.11 asks for a subscribable FEED — a URL the calendar re-fetches — and what ships is a FILE, saved once, which goes stale silently while continuing to look current. That is a difference of kind rather than a missing wire: published links exist, but they serve a FROZEN artifact that does not change when the schedule does, so there is no live-artifact path in the product for a feed to be built on. ⚠ Separately and by decision, the workbook carries NO FORMULAS: durations, dates and rollups are computed by the scheduling engine from logic and calendars, and no Excel expression reproduces them, so a formula would recalculate into disagreeing with the schedule. That is a refusal on the merits, not an omission. |
| Report packs | Planned | An ordered set of outputs produced in one action as one document — cover, contents, narrative, the schedule, the look-ahead, the S-curve, appendices — with section and page numbering running across the whole pack. The output pipeline produces one artifact at a time today. Specified in full; not started. |
| Schedule-interchange import and export | Planned | Not in the beta, and deferred rather than refused. spec.md §17 specifies it in full and §17.2 argues for it on principle — "export is not a retention hostage; a customer who wants to leave can take a working schedule with them". The menu entries stay out of a beta build because they dispatch to nothing, not because the capability is unwanted. ⚠ NAME COLLISION: the menu ACTION file.exportSchedule is THIS feature, the §17 interchange — not sequence.exportSchedule, which is the CSV/ICS data export and is a different capability at a different state. Grepping exportSchedule lands you in whichever you did not want. |
| Documents | Planned | Specified, not built. |
| Automation | Planned | Specified, not built. |
| Editable PPTX export | Deferred · past 1.0 | spec.md §12.11 marks this [OPEN] — "deferred, and possibly refused" — and it was ruled DEFERRED rather than refused on 2026-08-19. The argument against is already written: slide-shaped PDF and PNG cover the real need, the artifact lands in the deck and stays correct, and a genuinely editable deck invites someone to edit the numbers — the failure mode the whole of §12 exists to remove. It stays available as a decision rather than being closed off. |
| The ledger, the invoice and the payment run | Not planned | spec.md §8 carries a dated scope ruling: Sequence's scope ENDS AT A CERTIFIED CLAIM. It owns the schedule of values, the periodic application, its certification and its retention, in both directions — and it does not own the ledger, the invoice, the tax treatment or the payment run, and never becomes the system of record for money that has moved. §8.8 puts the reason plainly: "a product that stores what was paid will be asked to reconcile it, and reconciliation is the whole of the thing this section refuses to become". ⚠ Export is a boundary rather than an integration — certified values, retention and cost-code actuals leave in a documented shape, and certified amounts can come back so Committed and Actual stay honest. Direct posting into a finance system is a separate question and is still open, not refused here. |
| Document storage — a folder tree, a sync client, a DMS | Not planned | spec.md §10.5: "Sequence does document CONTROL. It does not do document STORAGE." Sequence owns which revision is current, who holds it, who approved it, what it supersedes and what is outstanding; the bytes and their durability belong to the collaboration server, and viewing, marking up and comparing revisions belong to Redline. There is no folder tree, no sync client and no attempt at SharePoint — a folder hierarchy is a worse register than a register: it holds one axis of classification, enforces nothing, and answers none of those questions. ⚠ plan.md's risk table names this one to "say no early", and D13 resolved the storage itself to a distinct product, Amalgative Storage. The same boundary is what sequence.annotations rests on for imagery. |
| Automatic target-date solving | Not planned | Ruled 2026-08-19 on spec.md §5.6's own argument: "probably should not exist — automatic compression produces programmes nobody believes". ⚠ There is a deliberate alternative rather than a hole: §5.6's COMPRESSION ASSISTANT lists candidate rows on the critical path with their duration, float, assigned resources and, where costing is on, the cost to crash. It RANKS AND PROPOSES; it does not act. That is the shape invariant 4 requires of anything that would move a schedule — a proposed diff a person accepts whole, in part, or not at all. |