Merge origin/main into plugin/host-model-api

Conflicts were additive: main's AddLayer/ModifyLayer (LayerConfig, OptHandle)
alongside this branch's v7 system-variable, transaction, selection and solid
operations. Main's variants keep their positions in the IPC enums; the v7
additions follow them. src/io/mod.rs takes main's reworded Shape-rotation note.

Co-Authored-By: Claude Sonnet 5 <[email protected]>
This commit is contained in:
Felix RiestraandClaude Sonnet 5 committed 2026-09-21 15:12:58 +02:00
commit 6b8056cfeb
28 files changed
+1425 -97

No files matched your search

+27 -1
View File
@@ -17,6 +17,10 @@ The server provides four tools:
- `ocs_execute` performs one operation, an atomic record update, or a sequential batch against the real editor.
- `ocs_capture` returns a bounded PNG of the drawing viewport or complete window.
Native builds also accept a second, headless entry point (`OpenCADStudio --serve`, one
JSON request per line) for scripts and CI; the two native channels, their startup dialogs
and their limits are described in [native.md](native.md).
The normal flow is to call `ocs_sessions`, pass its returned `session_id` as `ocs_session_id` to the other tools, read the chosen session with `ocs_read`, then pass the returned document state into `ocs_execute`. For example, an undo request has this shape:
```json
@@ -138,7 +142,27 @@ Execute responses use `response_detail: "compact"` by default and return only th
For unfamiliar or conditional commands, use `start`, then inspect `state.command` in every response. Its `accepts` array gives the valid MCP input kinds, `options` gives the current tokens, and `input_example` gives the next request shape. Add the current state fields and a new `request_id` to each step.
Every `ocs_execute` request requires a caller-generated `request_id`. Reuse that ID only to retry the identical request after a timeout, together with the same session ID, document ID, expected revision, and selection. Commands report `waiting_input` while more input is required and asynchronous work remains `running` until its real callback finishes. Geometry stays in OpenCADStudio and its geometry kernel.
Every `ocs_execute` request requires a caller-generated `request_id`. Reuse that ID only to retry the identical request after a timeout, together with the same session ID, document ID, expected revision, and selection. Commands report `waiting_input` while more input is required and asynchronous work remains `running` until its real callback finishes.
A `run` reply also carries two fields that describe what happened to the line:
* `blocked_by` — `null` when the command finished, otherwise what *this line*
left open and waiting: `"command"` (an interactive command), `"text_editor"`
(the in-place editor opened by the `TEXT` content step), `"mtext_editor"`, or
`"modal:<Kind>"` for a dialog. A surface that was already open when the line
started (an earlier line's editor, another tab, a startup dialog) is not
attributed to this line, so a caller polling for `completed` is never stuck
waiting on someone else's editor. `blocked_by` is set whenever `status` is
`waiting_input`.
* `unconsumed` — the tokens of the line that no prompt ever asked for, in order.
A fully consumed line reports `[]`; `CIRCLE 0,0 5 9` reports `["9"]` because
the radius step ends the command. Over-typed input is reported instead of
being dropped on the floor.
When a line ends at the `TEXT` content step (`TEXT 0,0 5 0 hello`) the text is
typed into the in-place editor and committed, so the line behaves exactly as the
same input typed into the GUI command line would: one `Text` entity, nothing
unconsumed, and the command closed rather than left armed for a second line. Geometry stays in OpenCADStudio and its geometry kernel.
`ocs_capture` defaults to the drawing viewport and a longest edge of 1600 pixels. Set `scope` to `window` for the full interface or change `max_dimension` between 256 and 4096.
@@ -156,3 +180,5 @@ The repeatable live-editor evaluation draws three isolated entities in one batch
```sh
python3 docs/automation/mcp_eval.py target/debug/OpenCADStudio
```
The web build is driven through two JavaScript functions instead of a process; see [web.md](web.md).
+131
View File
@@ -0,0 +1,131 @@
# Driving the native builds
Every native build has three entry points besides the editor window: the MCP
server described in the [README](README.md) (`OpenCADStudio --mcp`), a headless
JSON-line channel for scripts and CI (`OpenCADStudio --serve`), and a one-shot
converter (`OpenCADStudio --export IN OUT`). The browser entry is described in
[web.md](web.md).
## Choosing a channel
| | `--mcp` | `--serve` |
|---|---|---|
| Transport | MCP (JSON-RPC over stdio) with the four tools of the README | one JSON request per line on stdin, one reply per line on stdout; `--port N` listens on `127.0.0.1:N` instead |
| Editor | attaches to a running desktop editor, or starts one when there is none (`ocs_sessions`, `launch_if_none`) | no window; a document session of its own that lives as long as the process |
| Needs a display | yes (`xvfb-run` works on a headless machine; software rendering may raise the GPU warning dialog) | no |
| Needs a writable `HOME` / `XDG_CONFIG_HOME` | yes — the session descriptors live in the user configuration directory; without one `ocs_sessions` answers `No user configuration directory` | no |
| Startup dialogs | those of the editor (below) | none |
| `capture` | `ocs_capture` | not available (`gui_required`): there is no window to capture |
| Good for | an MCP client driving a real editor | batch jobs, CI, agents without a display |
## `--serve`
The first line written is a readiness banner, before any request is read:
```json
{"ok":true,"ready":true,"version":"2026.38"}
```
Over `--port N` the banner opens every connection; the listener serves one
client at a time and the document session persists across reconnects.
Two request styles share the channel, told apart by the `protocol` field:
| Style | Shape | Operations |
|---|---|---|
| protocol 1 | `{"protocol":1,"request_id":"…","document_id":N,"op":"…"}` — the same requests and replies as `ocs_execute` / `ocs_read`, described in the README | `new`, `open`, `save`, `activate`, `select`, `run`, `start`, `input`, `cancel`, `stop`, `action`, `property`, `set_properties`, `embed_image`, `undo`, `redo`; reads `state`, `capabilities`, `query`, `entities`, `records`, `record_schema`, `layers`, `header`, `properties`, `measure`, `history`, `commands`, `events`, `operation` |
| legacy | `{"op":"…"}` without `protocol` | `new`, `open`, `run`, `entities`, `query`, `records`, `record_schema`, `capabilities`, `layers`, `header`, `select`, `save`, `undo`, `redo` |
The two styles differ in how a waiting command is treated. A protocol-1 `run`
or `start` while a command is still active is refused with `command_busy`:
continue it with `input` or drop it with `cancel`. A legacy `run` line ends the
waiting command and starts the new one.
### The legacy `run` line
`{"op":"run","cmd":"LINE 0,0 10,10"}` feeds the tokens after the command name
to the prompts in order and then presses Enter. The reply is
```json
{"ok":true,"cmd":"CIRCLE 0,0 5 9","status":"completed","blocked_by":null,"entities":1,"added":1,"unconsumed":["9"]}
```
| Field | Meaning |
|---|---|
| `status` | `completed`, or `waiting_input` when the line left something open |
| `blocked_by` | what *this line* left open: `"command"`, `"text_editor"`, `"mtext_editor"`, `"modal:<Kind>"`, or `null`. A surface that was already open when the line started is not attributed to it |
| `entities`, `added` | entity count after the line, and the difference to before |
| `unconsumed` | tokens no prompt asked for, in order (`["9"]` above: the radius step ended the command); `[]` when everything was consumed |
`waiting_input` means input is still required, not that nothing happened:
`SPLINE` and `ARC` keep accepting input after the entity exists, and
`zoom_extents` reports it after the view has already moved.
`TEXT 0,0 5 0 hello world` creates one `Text` entity holding `hello world`: the
tail of the line is typed into the in-place editor, committed, and the command
is closed so it does not stay armed for a second line. A `TEXT` line without
content answers `waiting_input` with `blocked_by: "text_editor"`.
Known gap (#1390): `MTEXT`, `HATCH` and `POLYGON` answer `completed` with
`added: 0` on this path — they need interactive surfaces (the rich-text editor,
a pattern choice, a boundary pick) that a single line has no tokens to fill.
## `--mcp`
The MCP server is a client of the desktop editor: `ocs_sessions` lists the live
editor sessions found in the configuration directory and starts one when none
is running, so several MCP connections can share one editor and its state.
A fresh profile opens two dialogs before any drawing exists, and both block
document creation:
| Step | `ocs_sessions` reports |
|---|---|
| right after launch | `modal: "AssocPrompt"` |
| after `close_modal` | `modal: "DonationPrompt"` |
| after a second `close_modal` | `modal: null` |
While a dialog is open, `{"op":"new"}` answers `ok` but no tab is created.
Close the dialogs first:
```python
while (info := sessions())["modal"]:
execute({"op": "action", "name": "close_modal", "document_id": info["document_id"]})
```
The donation prompt appears once per application version per profile, so an
automated profile meets it once per version rather than on every run.
`new` makes the new drawing the active document. Requests carry `document_id`;
take it from the reply's state and pass it on the following requests, as the
README describes.
## `--export`
`OpenCADStudio --export IN OUT` loads `IN` and writes `OUT` without a window or
a session. The output format follows the extension: `.dxf` writes DXF, any other
extension writes DWG at the document's own version. The exit code is 0 on
success and 1 when the input cannot be read or the output cannot be written.
## Interactive steps
For commands whose answers cannot be written on one line, use `start` and then
`input`. `start` replies with `state.command`, whose `accepts`, `options` and
`input_example` say what the current step wants:
| Step | Request | Note |
|---|---|---|
| a point | `{"op":"input","kind":"point","point":[x,y,0],"space":"wcs"}` | the point is an array; `space` is `wcs` (default), `ucs` or `relative` |
| a typed value (height, rotation) | `{"op":"input","kind":"text","text":"5"}` | digits go through `text`, not `token`; `{"op":"input","kind":"enter"}` accepts the default |
| a keyword (`J`, `ST`, `C`) | `{"op":"input","kind":"token","text":"C"}` | the payload field is named `text` |
| an object | `{"op":"input","kind":"entity","handle":"HANDLE","point":[x,y,0]}` | `kind: "structure"` for a sub-entity pick, `kind: "selection"` to hand over the current selection |
| the content of `TEXT` | `{"op":"action","name":"text_input","value":"hello"}` then `{"op":"action","name":"text_commit"}` | the command has already ended; the in-place editor owns the input |
| the content of `MTEXT` | `{"op":"action","name":"mtext_insert","value":"hello"}` then `{"op":"action","name":"mtext_commit"}` | the rich-text editor has its own actions; the `TEXT` ones answer `editor_closed` |
## Native-only operations
| Operation | `--serve` (protocol 1) | `--mcp` |
|---|---|---|
| measure | `{"protocol":1,"op":"measure","handles":["2A"]}` — kernel-computed length, area, bounds and mass properties; no window needed | `ocs_read` with `op: "measure"` and `parameters.handles` |
| capture | `{"protocol":1,"op":"capture","path":"out.png"}` — needs the editor window, so it answers `gui_required` here | `ocs_capture` with `scope` (`viewport` or `window`) and `max_dimension` |
| `set_properties` | requires `collection`; colours are the serialised enum — `{"Index": 1}` or `"ByLayer"`; a colour name such as `"Red"` is rejected with `invalid_value` | same request through `ocs_execute` |
+68
View File
@@ -0,0 +1,68 @@
# Driving the web build
The web build exposes the same control core as `--serve` and `--mcp`, through two
functions on the WebAssembly module. This page covers only what is specific to the
browser; the operations, request envelope and replies are the ones in the
[README](README.md).
## Reaching the control channel
The Trunk loader publishes the wasm-bindgen module as `window.wasmBindings`. The
control channel is ready once `window.wasmBindings?.ocs_control_submit` is defined
(the [save smoke test](web_save_smoke.cjs) waits for exactly that).
| Function | Behaviour |
|---|---|
| `ocs_control_submit(request: string) → string` | Queues one request (a JSON string) and returns a ticket immediately. |
| `ocs_control_take(ticket: string) → string \| undefined` | Returns the reply (a JSON string) once it is ready, `undefined` before that. A reply is removed when taken. |
Poll `ocs_control_take` until it returns a reply, as the smoke test does:
```js
async function control(request) {
const api = window.wasmBindings;
const ticket = api.ocs_control_submit(JSON.stringify(request));
for (let i = 0; i < 600; i++) {
const reply = api.ocs_control_take(ticket);
if (reply) return JSON.parse(reply);
await new Promise((resolve) => setTimeout(resolve, 50));
}
throw new Error('control timed out');
}
```
Limits: at most 64 requests wait in the queue (a further `submit` is answered with the
`busy` error), and only the newest 64 replies are kept, so take a reply soon after it is
ready. A request that is not valid JSON is answered with `op: "invalid"` and a
`parse_error`.
## Opening a drawing from bytes
The web build has no file system, so `open` takes the file's bytes instead of a path:
```json
{"op":"open","name":"plan.dwg","data_base64":"…"}
```
`name` is reduced to its file name and labels the tab and the recent-files entry;
`data_base64` must be valid base64 and non-empty (`invalid_request` otherwise). While an
earlier drawing is still opening the request is answered with `busy`; retry it once
`state` shows the tab. A `path` is resolved against the browser's "recent files" store
rather than a disk, so use `data_base64` for a file the page has not seen.
`data_base64` is rejected on native builds; use `path` there.
## Startup dialogs
A dialog can be open when the page is ready. The donation prompt appears once per
application version in each browser profile, so a fresh or disposable profile always
sees it. While a modal is open `state.modal` is set, and `{"op":"new"}` returns `ok`
without creating a tab until the modal is closed. Close it first:
```js
const state = await control({ op: 'state' });
if (state.modal) await control({ op: 'action', name: 'close_modal' });
await control({ op: 'new' });
```
Native builds can show a second dialog at launch (the file-association prompt), so check
`state.modal` again after closing one.