CanvasAnvil is an open source AI diagram maker and flowchart generator - an infinite-canvas diagram editor and self-hosted drawio alternative that writes editable .drawio files. Text to diagram, then fix nodes, layout and labels in place. Also AI presentations, interior design CAD/BOM, posters and infographics.
TypeScript ai-agentsai-diagramai-presentationbomcadcanvasdiagramdiagram-generatordrawioflowchartinfographicspptppt-generatorpresentation-generatorreactsvgedittext-to-diagramtypescriptworkflow
import assert from "node:assert/strict";
import { test } from "node:test";
import {
materialLabel,
materialToken,
referencedMaterials,
uploadedImageLabel,
} from "../client/workspaces/ppt/canvas/lib/material-tokens";
/**
* Which of a slide's attached pictures are actually sent to the image model.
*
* Attached is not the same as referenced. The prompt names every reference
* image it is given and tells the model where to put it, so a picture the
* user uploaded and never mentioned is one the prompt cannot account for.
*/
const mat = (name: string) => ({ name, dataUrl: `${name}.png` });
test("only the pictures the description mentions come through", () => {
const picked = referencedMaterials(
`左侧放 ${materialToken("第1张")},右侧留白。`,
[mat("第1张"), mat("第2张")],
);
assert.deepEqual(picked.map((m) => m.name), ["第1张"]);
});
test("an upload nobody referenced is not sent", () => {
assert.deepEqual(referencedMaterials("一张普通的示意图,没有引用任何素材。", [mat("第1张")]), []);
});
test("references are ordered by where the description mentions them", () => {
// If a cap ever bites, what survives should be what the description leads
// with, not whatever order the files happened to be uploaded in.
const description = `先看 ${materialToken("第3张")},再看 ${materialToken("第1张")}。`;
const picked = referencedMaterials(description, [mat("第1张"), mat("第2张"), mat("第3张")]);
assert.deepEqual(picked.map((m) => m.name), ["第3张", "第1张"]);
});
test("a description that mentions everything keeps everything", () => {
const description = `${materialToken("第1张")} ${materialToken("第2张")}`;
const picked = referencedMaterials(description, [mat("第1张"), mat("第2张")]);
assert.equal(picked.length, 2);
});
test("an empty description references nothing", () => {
assert.deepEqual(referencedMaterials("", [mat("第1张")]), []);
assert.deepEqual(referencedMaterials(undefined as unknown as string, [mat("第1张")]), []);
});
test("a name that only appears as bare text is not a reference", () => {
// The token is the reference; the words around it are just prose.
assert.deepEqual(referencedMaterials("把第1张放在左边", [mat("第1张")]), []);
});
test("no attachments means nothing to send", () => {
assert.deepEqual(referencedMaterials(`看 ${materialToken("第1张")}`, []), []);
});
/**
* Pictures attached in the chat panel are named separately from a slide's own
* materials, because both can ride along with one slide and the prompt has to
* be able to say which is which.
*/
test("chat pictures and slide materials never share a name", () => {
const slideNames = [1, 2, 3].map((n) => materialLabel(n, "zh"));
const chatNames = [1, 2, 3].map((n) => uploadedImageLabel(n, "zh"));
assert.equal(new Set([...slideNames, ...chatNames]).size, 6);
const slideEn = [1, 2, 3].map((n) => materialLabel(n, "en"));
const chatEn = [1, 2, 3].map((n) => uploadedImageLabel(n, "en"));
assert.equal(new Set([...slideEn, ...chatEn]).size, 6);
});
test("a chat picture's name is stable and numbered from one", () => {
assert.equal(uploadedImageLabel(1, "zh"), "对话图1");
assert.equal(uploadedImageLabel(2, "en"), "Chat image 2");
});
test("a chat picture's name is not mistaken for a description reference", () => {
// The token syntax belongs to slide materials; a chat picture is passed by
// label alone, so naming one in a description must not pull it in here.
const picked = referencedMaterials(`看 ${materialToken(uploadedImageLabel(1, "zh"))}`, [
{ name: materialLabel(1, "zh") },
]);
assert.deepEqual(picked, []);
});