Skip to content

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.

refactor(chat): give the three transcripts one set of message readers

codingFeng1014434062764e7
diff --git a/client/shared/chat/message-content.ts b/client/shared/chat/message-content.ts
new file mode 100644
index 0000000..34d2b9d
--- /dev/null
+++ b/client/shared/chat/message-content.ts
@@ -0,0 +1,102 @@
+/**
+ * Reading the text back out of a chat message.
+ *
+ * Every workspace's transcript renders the same underlying shape, and every
+ * one of them appends uploaded file contents to the user's own words before
+ * sending. These are the readers that undo that for display.
+ */
+
+export interface MessagePart {
+ type: string;
+ text?: string;
+ toolCallId?: string;
+ state?: string;
+ input?: unknown;
+ output?: unknown;
+ errorText?: string;
+ [key: string]: unknown;
+}
+
+export interface UIMessage {
+ id: string;
+ role: "user" | "assistant" | "system";
+ content?: string | unknown[];
+ parts?: MessagePart[];
+ [key: string]: unknown;
+}
+
+export interface TextSection {
+ type: "text" | "file";
+ content: string;
+ filename?: string;
+ fileType?: "pdf" | "text";
+ /** Size of the extracted text, shown on the collapsed attachment. */
+ charCount?: number;
+}
+
+/** Matches the `[PDF: name]` / `[File: name]` blocks the chat panels append. */
+const FILE_SECTION = /\[(PDF|File):\s*([^\]]+)\]\n([\s\S]*?)(?=\n\n\[(PDF|File):|$)/g;
+
+/**
+ * Splits a message into what the user typed and the file contents that were
+ * appended to it, so the transcript can show the files as collapsed
+ * attachments instead of walls of extracted text.
+ */
+export function splitTextIntoFileSections(text: string): TextSection[] {
+ const sections: TextSection[] = [];
+ const pattern = new RegExp(FILE_SECTION.source, "g");
+ let lastIndex = 0;
+ let match: RegExpExecArray | null;
+
+ while ((match = pattern.exec(text)) !== null) {
+ const beforeText = text.slice(lastIndex, match.index).trim();
+ if (beforeText) sections.push({ type: "text", content: beforeText });
+
+ const fileContent = match[3].trim();
+ sections.push({
+ type: "file",
+ content: fileContent,
+ filename: match[2].trim(),
+ fileType: match[1].toLowerCase() === "pdf" ? "pdf" : "text",
+ charCount: fileContent.length,
+ });
+ lastIndex = match.index + match[0].length;
+ }
+
+ const remainingText = text.slice(lastIndex).trim();
+ if (remainingText) sections.push({ type: "text", content: remainingText });
+
+ // No file blocks at all: hand back the message unchanged rather than empty.
+ if (sections.length === 0) sections.push({ type: "text", content: text });
+
+ return sections;
+}
+
+/**
+ * The message's text, from whichever of the three shapes it arrived in:
+ * streaming `parts`, an array `content`, or a plain string.
+ */
+export function getMessageTextContent(message: UIMessage): string {
+ if (message.parts && Array.isArray(message.parts)) {
+ return message.parts
+ .filter((part) => part.type === "text")
+ .map((part) => String(part.text ?? ""))
+ .join("\n");
+ }
+
+ if (Array.isArray(message.content)) {
+ return message.content
+ .filter((part): part is MessagePart => (part as MessagePart)?.type === "text")
+ .map((part) => String(part.text || ""))
+ .join("\n");
+ }
+
+ return message.content || "";
+}
+
+/** What the user actually typed, with the appended file contents stripped. */
+export function getUserOriginalText(message: UIMessage): string {
+ return getMessageTextContent(message)
+ .replace(/\n\n\[(PDF|File):\s*[^\]]+\]\n[\s\S]*$/, "")
+ .trim();
+}
diff --git a/client/shared/chat/ppt-message-tags.ts b/client/shared/chat/ppt-message-tags.ts
new file mode 100644
index 0000000..ddc85c0
--- /dev/null
+++ b/client/shared/chat/ppt-message-tags.ts
@@ -0,0 +1,215 @@
+/**
+ * The markers a PPT conversation carries inside otherwise plain message text.
+ *
+ * The chat input writes `[[PPT_SLIDE|n|title|kind]]` when the user attaches a
+ * slide, and the assistant answers slide edits as a `ppt_edit` JSON payload.
+ * The transcript has to pull both back out to render them as chips and diffs
+ * rather than as raw text.
+ */
+
+export interface PptEditSlidePatch {
+ id?: string;
+ title?: string;
+ content?: string[];
+ description?: string;
+ note?: string;
+ layout?: string;
+ [key: string]: unknown;
+}
+
+export interface PptToolPayload {
+ type: "ppt_edit";
+ slides: PptEditSlidePatch[];
+}
+
+export type PptSlideTag = { n: number; title?: string; kind?: "outline" | "slide_image" };
+export type ImageTag = { name?: string; url: string };
+
+export type InlinePptPart =
+ | { type: "text"; text: string }
+ | { type: "ppt"; n: number; title?: string; kind?: "outline" | "slide_image" };
+
+/** Pulls out slide markers that sit alone on their own line. */
+export function extractPptSlideTags(text: string): { tags: PptSlideTag[]; rest: string } {
+ const tags: PptSlideTag[] = [];
+ const restLines: string[] = [];
+
+ for (const line of String(text || "").split(/\r?\n/)) {
+ const tagged = line.match(/^\[\[PPT_SLIDE\|(\d+)\|([^|]*)\|(outline|slide_image)\]\]$/);
+ if (tagged) {
+ const n = Number(tagged[1]);
+ const title = String(tagged[2] || "").trim();
+ if (!Number.isNaN(n)) {
+ tags.push({ n, title: title || undefined, kind: tagged[3] as "outline" | "slide_image" });
+ }
+ continue;
+ }
+
+ // Markers written before `kind` existed default to a slide image.
+ const legacy = line.match(/^\[\[PPT_SLIDE\|(\d+)\|(.*)\]\]$/);
+ if (legacy) {
+ const n = Number(legacy[1]);
+ const title = String(legacy[2] || "").trim();
+ if (!Number.isNaN(n)) tags.push({ n, title: title || undefined, kind: "slide_image" });
+ continue;
+ }
+
+ restLines.push(line);
+ }
+
+ return { tags, rest: restLines.join("\n").replace(/^\s+|\s+$/g, "") };
+}
+
+const SAFE_IMAGE_PREFIXES = ["data:image", "blob:", "http://", "https://"];
+
+/** Pulls out image markers, ignoring any whose URL is not one we will render. */
+export function extractImageTags(text: string): { images: ImageTag[]; rest: string } {
+ const images: ImageTag[] = [];
+ const restLines: string[] = [];
+
+ for (const line of String(text || "").split(/\r?\n/)) {
+ const match = line.match(/^\[\[IMAGE\|([^|]*)\|([\s\S]+)\]\]$/);
+ if (match) {
+ const name = String(match[1] || "").trim();
+ const url = String(match[2] || "").trim();
+ if (SAFE_IMAGE_PREFIXES.some((prefix) => url.startsWith(prefix))) {
+ images.push({ name: name || undefined, url });
+ continue;
+ }
+ }
+ restLines.push(line);
+ }
+
+ return { images, rest: restLines.join("\n").replace(/^\s+|\s+$/g, "") };
+}
+
+/**
+ * Splits a line into text and the slide markers embedded in it, so a slide
+ * referenced mid-sentence renders as a chip in place.
+ */
+export function splitInlinePptTags(text: string): InlinePptPart[] {
+ const raw = String(text || "");
+ const out: InlinePptPart[] = [];
+ const pattern = /\[\[PPT_SLIDE\|(\d+)\|([\s\S]*?)\]\]/g;
+ let lastIndex = 0;
+ let match: RegExpExecArray | null;
+
+ while ((match = pattern.exec(raw))) {
+ const before = raw.slice(lastIndex, match.index);
+ if (before) out.push({ type: "text", text: before });
+
+ const n = Number(match[1]);
+ const title = String(match[2] || "").trim();
+ if (Number.isNaN(n)) {
+ // Not a real marker after all; keep it as the text it looked like.
+ out.push({ type: "text", text: match[0] });
+ } else {
+ let kind: "outline" | "slide_image" = "slide_image";
+ let normalizedTitle = title;
+ // A title may itself contain "|", so only a trailing known kind counts.
+ const parts = title.split("|");
+ if (parts.length >= 2) {
+ const maybeKind = parts[parts.length - 1].trim();
+ if (maybeKind === "outline" || maybeKind === "slide_image") {
+ kind = maybeKind;
+ normalizedTitle = parts.slice(0, -1).join("|").trim();
+ }
+ }
+ out.push({ type: "ppt", n, title: normalizedTitle || undefined, kind });
+ }
+ lastIndex = match.index + match[0].length;
+ }
+
+ const rest = raw.slice(lastIndex);
+ if (rest) out.push({ type: "text", text: rest });
+ if (out.length === 0) out.push({ type: "text", text: raw });
+ return out;
+}
+
+/**
+ * Reads a slide-edit payload out of an assistant reply, whether it arrived
+ * fenced, bare, or as a naked array of slides.
+ */
+export function extractPptToolPayload(text: string): PptToolPayload | null {
+ const raw = String(text || "").trim();
+ if (!raw) return null;
+
+ const fenced = raw.match(/```(?:json)?\s*([\s\S]*?)\s*```/i);
+ let parsed: unknown;
+ try {
+ parsed = JSON.parse((fenced?.[1] ?? raw).trim());
+ } catch {
+ return null;
+ }
+
+ if (Array.isArray(parsed)) {
+ return {
+ type: "ppt_edit",
+ slides: parsed.filter((s): s is PptEditSlidePatch => !!s && typeof s === "object"),
+ };
+ }
+ if (!parsed || typeof parsed !== "object") return null;
+
+ const record = parsed as Record<string, unknown>;
+ const declaredType = String(record.type || "").trim().toLowerCase();
+ if (declaredType && declaredType !== "ppt_edit") return null;
+
+ const slides = Array.isArray(record.slides)
+ ? record.slides.filter((s): s is PptEditSlidePatch => !!s && typeof s === "object")
+ : [];
+ return slides.length > 0 ? { type: "ppt_edit", slides } : null;
+}
+
+/** The slide's 1-based number, from its id where it has one. */
+export function getSlideNumber(slide: PptEditSlidePatch, index: number): number {
+ const match = String(slide?.id || "").match(/slide-(\d+)/i);
+ if (match) {
+ const n = Number(match[1]);
+ if (!Number.isNaN(n)) return n;
+ }
+ return index + 1;
+}
+
+/** The fields shown first, in the order a reader expects them. */
+const KNOWN_FIELDS = ["title", "description", "layout", "note", "content"];
+const EMPTY_VALUE = "(空)";
+
+/** Flattens a patch into displayable rows, keeping unknown fields rather than hiding them. */
+export function slidePatchEntries(slide: PptEditSlidePatch): Array<{ key: string; value: string }> {
+ const rows: Array<{ key: string; value: string }> = [];
+
+ for (const key of KNOWN_FIELDS) {
+ if (!(key in slide)) continue;
+ const raw = slide[key];
+ if (Array.isArray(raw)) {
+ const text = raw.map((x) => String(x || "").trim()).filter(Boolean).join(";");
+ rows.push({ key, value: text || EMPTY_VALUE });
+ } else if (raw && typeof raw === "object") {
+ rows.push({ key, value: JSON.stringify(raw) });
+ } else {
+ rows.push({ key, value: String(raw ?? "").trim() || EMPTY_VALUE });
+ }
+ }
+
+ for (const [key, value] of Object.entries(slide)) {
+ if (key === "id" || KNOWN_FIELDS.includes(key)) continue;
+ if (value === undefined) continue;
+ rows.push({ key, value: typeof value === "string" ? value : JSON.stringify(value) });
+ }
+
+ return rows;
+}
+
+const FIELD_LABELS: Record<string, { zh: string; en: string }> = {
+ title: { zh: "标题", en: "Title" },
+ description: { zh: "画面描述", en: "Description" },
+ layout: { zh: "布局", en: "Layout" },
+ note: { zh: "备注", en: "Note" },
+ content: { zh: "内容", en: "Content" },
+};
+
+/** An unknown field falls back to its own key rather than disappearing. */
+export function slideFieldLabel(key: string, uiLang: "zh" | "en"): string {
+ const hit = FIELD_LABELS[key];
+ return hit ? hit[uiLang] : key;
+}
diff --git a/client/workspaces/cad/chat/ChatMessageDisplay.tsx b/client/workspaces/cad/chat/ChatMessageDisplay.tsx
index 32f5238..a41989b 100644
--- a/client/workspaces/cad/chat/ChatMessageDisplay.tsx
+++ b/client/workspaces/cad/chat/ChatMessageDisplay.tsx
@@ -7,26 +7,26 @@ import { ScrollArea } from "@/shared/ui/scroll-area";
import { CodeBlock } from "@/workspaces/cad/chat/code-block";
import { useUiLanguage } from "@/shared/i18n";
import { cn } from "@/shared/lib/utils";
+import {
+ getMessageTextContent,
+ getUserOriginalText,
+ splitTextIntoFileSections,
+ type MessagePart,
+ type UIMessage,
+} from "@/shared/chat/message-content";
+import {
+ extractImageTags,
+ extractPptSlideTags,
+ extractPptToolPayload,
+ getSlideNumber,
+ slideFieldLabel,
+ slidePatchEntries,
+ splitInlinePptTags,
+} from "@/shared/chat/ppt-message-tags";
+
+export type { MessagePart, UIMessage };
// Types
-export interface MessagePart {
- type: 'text' | 'reasoning' | 'tool-call' | 'tool-result' | 'file';
- text?: string;
- toolName?: string;
- toolCallId?: string;
- state?: string;
- input?: any;
- output?: string;
- url?: string; // for images
-}
-
-export interface UIMessage {
- id: string;
- role: 'user' | 'assistant' | 'system';
- content: string; // Fallback
- parts?: MessagePart[];
-}
-
interface ChatMessageDisplayProps {
messages: UIMessage[];
setInput: (input: string) => void;
@@ -38,24 +38,6 @@ interface ChatMessageDisplayProps {
}
// Helper to split text content into regular text and file sections
-interface TextSection {
- type: "text" | "file";
- content: string;
- filename?: string;
- charCount?: number;
- fileType?: "pdf" | "text";
-}
-
-interface PptEditSlidePatch {
- id?: string;
- [key: string]: any;
-}
-
-interface PptToolPayload {
- type: "ppt_edit";
- slides: PptEditSlidePatch[];
-}
-
interface CadPatchEdit {
search: string;
replace: string;
@@ -104,131 +86,6 @@ function AutoApplyCodeOnce({
return null;
}
-function splitTextIntoFileSections(text: string): TextSection[] {
- const sections: TextSection[] = [];
- const filePattern = /\[(PDF|File):\s*([^\]]+)\]\n([\s\S]*?)(?=\n\n\[(PDF|File):|$)/g;
- let lastIndex = 0;
- let match;
-
- while ((match = filePattern.exec(text)) !== null) {
- const beforeText = text.slice(lastIndex, match.index).trim();
- if (beforeText) {
- sections.push({ type: "text", content: beforeText });
- }
-
- const fileType = match[1].toLowerCase() === "pdf" ? "pdf" : "text";
- const filename = match[2].trim();
- const fileContent = match[3].trim();
- sections.push({
- type: "file",
- content: fileContent,
- filename,
- charCount: fileContent.length,
- fileType,
- });
-
- lastIndex = match.index + match[0].length;
- }
-
- const remainingText = text.slice(lastIndex).trim();
- if (remainingText) {
- sections.push({ type: "text", content: remainingText });
- }
-
- if (sections.length === 0) {
- sections.push({ type: "text", content: text });
- }
-
- return sections;
-}
-
-type PptSlideTag = { n: number; title?: string; kind?: "outline" | "slide_image" };
-
-function extractPptSlideTags(text: string): { tags: PptSlideTag[]; rest: string } {
- const lines = String(text || "").split(/\r?\n/);
- const tags: PptSlideTag[] = [];
- const restLines: string[] = [];
- for (const line of lines) {
- const m = line.match(/^\[\[PPT_SLIDE\|(\d+)\|([^|]*)\|(outline|slide_image)\]\]$/);
- if (m) {
- const n = Number(m[1]);
- const title = String(m[2] || "").trim();
- const kind = m[3] as "outline" | "slide_image";
- if (!Number.isNaN(n)) tags.push({ n, title: title || undefined, kind });
- continue;
- }
- const legacy = line.match(/^\[\[PPT_SLIDE\|(\d+)\|(.*)\]\]$/);
- if (legacy) {
- const n = Number(legacy[1]);
- const title = String(legacy[2] || "").trim();
- if (!Number.isNaN(n)) tags.push({ n, title: title || undefined, kind: "slide_image" });
- continue;
- }
- restLines.push(line);
- }
- const rest = restLines.join("\n").replace(/^\s+|\s+$/g, "");
- return { tags, rest };
-}
-
-type ImageTag = { name?: string; url: string };
-
-function extractImageTags(text: string): { images: ImageTag[]; rest: string } {
- const lines = String(text || "").split(/\r?\n/);
- const images: ImageTag[] = [];
- const restLines: string[] = [];
- for (const line of lines) {
- const m = line.match(/^\[\[IMAGE\|([^|]*)\|([\s\S]+)\]\]$/);
- if (m) {
- const name = String(m[1] || "").trim();
- const url = String(m[2] || "").trim();
- if (url.startsWith("data:image") || url.startsWith("blob:") || url.startsWith("http://") || url.startsWith("https://")) {
- images.push({ name: name || undefined, url });
- continue;
- }
- }
- restLines.push(line);
- }
- const rest = restLines.join("\n").replace(/^\s+|\s+$/g, "");
- return { images, rest };
-}
-
-type InlinePptPart =
- | { type: "text"; text: string }
- | { type: "ppt"; n: number; title?: string; kind?: "outline" | "slide_image" };
-
-function splitInlinePptTags(text: string): InlinePptPart[] {
- const raw = String(text || "");
- const out: InlinePptPart[] = [];
- const re = /\[\[PPT_SLIDE\|(\d+)\|([\s\S]*?)\]\]/g;
- let lastIndex = 0;
- let m: RegExpExecArray | null;
- while ((m = re.exec(raw))) {
- const before = raw.slice(lastIndex, m.index);
- if (before) out.push({ type: "text", text: before });
- const n = Number(m[1]);
- const title = String(m[2] || "").trim();
- if (!Number.isNaN(n)) {
- let kind: "outline" | "slide_image" = "slide_image";
- let normalizedTitle = title;
- const parts = title.split("|");
- if (parts.length >= 2) {
- const maybeKind = parts[parts.length - 1].trim();
- if (maybeKind === "outline" || maybeKind === "slide_image") {
- kind = maybeKind;
- normalizedTitle = parts.slice(0, -1).join("|").trim();
- }
- }
- out.push({ type: "ppt", n, title: normalizedTitle || undefined, kind });
- }
- else out.push({ type: "text", text: m[0] });
- lastIndex = m.index + m[0].length;
- }
- const rest = raw.slice(lastIndex);
- if (rest) out.push({ type: "text", text: rest });
- if (out.length === 0) out.push({ type: "text", text: raw });
- return out;
-}
-
type MarkdownSegment =
| { type: "markdown"; content: string }
| { type: "code"; code: string; language: string };
@@ -289,46 +146,6 @@ function splitMarkdownAndCodeBlocks(text: string): MarkdownSegment[] {
return segments;
}
-const getMessageTextContent = (message: UIMessage): string => {
- if (message.parts && message.parts.length > 0) {
- return message.parts
- .filter((part) => part.type === "text")
- .map((part) => part.text || "")
- .join("\n");
- }
- return message.content || "";
-};
-
-const getUserOriginalText = (message: UIMessage): string => {
- const fullText = getMessageTextContent(message);
- const filePattern = /\n\n\[(PDF|File):\s*[^\]]+\]\n[\s\S]*$/;
- return fullText.replace(filePattern, "").trim();
-};
-
-function extractPptToolPayload(text: string): PptToolPayload | null {
- const raw = String(text || "").trim();
- if (!raw) return null;
- const fenced = raw.match(/```(?:json)?\s*([\s\S]*?)\s*```/i);
- const candidate = (fenced?.[1] ?? raw).trim();
- let parsed: any;
- try {
- parsed = JSON.parse(candidate);
- } catch {
- return null;
- }
- if (Array.isArray(parsed)) {
- return { type: "ppt_edit", slides: parsed.filter((s: any) => s && typeof s === "object") };
- }
- if (!parsed || typeof parsed !== "object") return null;
- const typeRaw = String((parsed as any).type || "").trim().toLowerCase();
- if (typeRaw && typeRaw !== "ppt_edit") return null;
- const type = "ppt_edit" as const;
- const slides = Array.isArray((parsed as any).slides)
- ? (parsed as any).slides.filter((s: any) => s && typeof s === "object")
- : [];
- return slides.length > 0 ? { type, slides } : null;
-}
-
function extractCadPatchToolPayload(text: string): CadPatchToolPayload | null {
const raw = String(text || "").trim();
if (!raw) return null;
@@ -359,54 +176,6 @@ function extractCadPatchToolPayload(text: string): CadPatchToolPayload | null {
return { type: "cad_patch", target: "2d_svg", mode: "patch", edits };
}
-function getSlideNumber(slide: PptEditSlidePatch, index: number): number {
- const id = String(slide?.id || "");
- const m = id.match(/slide-(\d+)/i);
- if (m) {
- const n = Number(m[1]);
- if (!Number.isNaN(n)) return n;
- }
- return index + 1;
-}
-
-function slidePatchEntries(slide: PptEditSlidePatch): Array<{ key: string; value: string }> {
- const keys = ["title", "description", "layout", "note", "content"];
- const rows: Array<{ key: string; value: string }> = [];
- for (const key of keys) {
- if (!(key in slide)) continue;
- const raw = (slide as any)[key];
- if (Array.isArray(raw)) {
- const text = raw.map((x) => String(x || "").trim()).filter(Boolean).join(";");
- rows.push({ key, value: text || "(空)" });
- continue;
- }
- if (raw && typeof raw === "object") {
- rows.push({ key, value: JSON.stringify(raw) });
- continue;
- }
- rows.push({ key, value: String(raw ?? "").trim() || "(空)" });
- }
- for (const [k, v] of Object.entries(slide)) {
- if (k === "id" || keys.includes(k)) continue;
- if (v === undefined) continue;
- rows.push({ key: k, value: typeof v === "string" ? v : JSON.stringify(v) });
- }
- return rows;
-}
-
-function slideFieldLabel(key: string, uiLang: "zh" | "en"): string {
- const dict: Record<string, { zh: string; en: string }> = {
- title: { zh: "标题", en: "Title" },
- description: { zh: "画面描述", en: "Description" },
- layout: { zh: "布局", en: "Layout" },
- note: { zh: "备注", en: "Note" },
- content: { zh: "内容", en: "Content" },
- };
- const hit = dict[key];
- if (hit) return uiLang === "zh" ? hit.zh : hit.en;
- return key;
-}
-
function CadPatchEditsDisplay({ edits, uiLang }: { edits: CadPatchEdit[]; uiLang: "zh" | "en" }) {
const tr = (zh: string, en: string) => (uiLang === "zh" ? zh : en);
return (
diff --git a/client/workspaces/flow/chat/ChatMessageDisplay.tsx b/client/workspaces/flow/chat/ChatMessageDisplay.tsx
index f109e37..7f553ee 100644
--- a/client/workspaces/flow/chat/ChatMessageDisplay.tsx
+++ b/client/workspaces/flow/chat/ChatMessageDisplay.tsx
@@ -1,5 +1,12 @@
"use client"
+import {
+ getMessageTextContent,
+ getUserOriginalText,
+ splitTextIntoFileSections,
+ type TextSection,
+} from "@/shared/chat/message-content"
+
/**
* A message as the chat hook hands it over. `parts` is deliberately loose:
* the AI SDK streams tool calls, files and reasoning blocks whose shapes vary
@@ -92,74 +99,6 @@ function OperationsDisplay({ operations }: { operations: DiagramOperation[] }) {
import { useDiagram } from "@/workspaces/flow/state/diagram-context"
// Helper to split text content into regular text and file sections (PDF or text files)
-interface TextSection {
- type: "text" | "file"
- content: string
- filename?: string
- charCount?: number
- fileType?: "pdf" | "text"
-}
-
-function splitTextIntoFileSections(text: string): TextSection[] {
- const sections: TextSection[] = []
- // Match [PDF: filename] or [File: filename] patterns
- const filePattern =
- /\[(PDF|File):\s*([^\]]+)\]\n([\s\S]*?)(?=\n\n\[(PDF|File):|$)/g
- let lastIndex = 0
- let match
-
- while ((match = filePattern.exec(text)) !== null) {
- // Add text before this file section
- const beforeText = text.slice(lastIndex, match.index).trim()
- if (beforeText) {
- sections.push({ type: "text", content: beforeText })
- }
-
- // Add file section
- const fileType = match[1].toLowerCase() === "pdf" ? "pdf" : "text"
- const filename = match[2].trim()
- const fileContent = match[3].trim()
- sections.push({
- type: "file",
- content: fileContent,
- filename,
- charCount: fileContent.length,
- fileType,
- })
-
- lastIndex = match.index + match[0].length
- }
-
- // Add remaining text after last file section
- const remainingText = text.slice(lastIndex).trim()
- if (remainingText) {
- sections.push({ type: "text", content: remainingText })
- }
-
- // If no file sections found, return original text
- if (sections.length === 0) {
- sections.push({ type: "text", content: text })
- }
-
- return sections
-}
-
-const getMessageTextContent = (message: UIMessage): string => {
- if (message.parts && message.parts.length > 0) {
- return message.parts
- .filter((part) => part.type === "text")
- .map((part) => String(part.text ?? ""))
- .join("\n")
- }
- if (Array.isArray(message.content)) {
- return message.content
- .filter((part: any) => part?.type === "text")
- .map((part: any) => String(part?.text || ""))
- .join("\n")
- }
- return message.content || ""
-}
-
function extractFileSectionsFromText(text: string): TextSection[] {
return splitTextIntoFileSections(text).filter(
(s) => s.type === "file",
@@ -174,13 +113,6 @@ function stripFileSectionsFromText(text: string): string {
}
// Get only the user's original text, excluding appended file content
-const getUserOriginalText = (message: UIMessage): string => {
- const fullText = getMessageTextContent(message)
- // Strip out [PDF: ...] and [File: ...] sections that were appended
- const filePattern = /\n\n\[(PDF|File):\s*[^\]]+\]\n[\s\S]*$/
- return fullText.replace(filePattern, "").trim()
-}
-
interface ChatMessageDisplayProps {
messages: UIMessage[]
setInput: (input: string) => void
diff --git a/client/workspaces/ppt/chat/ChatMessageDisplay.tsx b/client/workspaces/ppt/chat/ChatMessageDisplay.tsx
index 981d890..901d371 100644
--- a/client/workspaces/ppt/chat/ChatMessageDisplay.tsx
+++ b/client/workspaces/ppt/chat/ChatMessageDisplay.tsx
@@ -7,26 +7,26 @@ import { ScrollArea } from "@/shared/ui/scroll-area";
import { CodeBlock } from "@/workspaces/ppt/chat/code-block";
import { useUiLanguage } from "@/shared/i18n";
import { cn } from "@/shared/lib/utils";
+import {
+ getMessageTextContent,
+ getUserOriginalText,
+ splitTextIntoFileSections,
+ type MessagePart,
+ type UIMessage,
+} from "@/shared/chat/message-content";
+import {
+ extractImageTags,
+ extractPptSlideTags,
+ extractPptToolPayload,
+ getSlideNumber,
+ slideFieldLabel,
+ slidePatchEntries,
+ splitInlinePptTags,
+} from "@/shared/chat/ppt-message-tags";
+
+export type { MessagePart, UIMessage };
// Types
-export interface MessagePart {
- type: 'text' | 'reasoning' | 'tool-call' | 'tool-result' | 'file';
- text?: string;
- toolName?: string;
- toolCallId?: string;
- state?: string;
- input?: any;
- output?: string;
- url?: string; // for images
-}
-
-export interface UIMessage {
- id: string;
- role: 'user' | 'assistant' | 'system';
- content: string; // Fallback
- parts?: MessagePart[];
-}
-
interface ChatMessageDisplayProps {
messages: UIMessage[];
setInput: (input: string) => void;
@@ -37,237 +37,6 @@ interface ChatMessageDisplayProps {
}
// Helper to split text content into regular text and file sections
-interface TextSection {
- type: "text" | "file";
- content: string;
- filename?: string;
- charCount?: number;
- fileType?: "pdf" | "text";
-}
-
-interface PptEditSlidePatch {
- id?: string;
- [key: string]: any;
-}
-
-interface PptToolPayload {
- type: "ppt_edit";
- slides: PptEditSlidePatch[];
-}
-
-function splitTextIntoFileSections(text: string): TextSection[] {
- const sections: TextSection[] = [];
- const filePattern = /\[(PDF|File):\s*([^\]]+)\]\n([\s\S]*?)(?=\n\n\[(PDF|File):|$)/g;
- let lastIndex = 0;
- let match;
-
- while ((match = filePattern.exec(text)) !== null) {
- const beforeText = text.slice(lastIndex, match.index).trim();
- if (beforeText) {
- sections.push({ type: "text", content: beforeText });
- }
-
- const fileType = match[1].toLowerCase() === "pdf" ? "pdf" : "text";
- const filename = match[2].trim();
- const fileContent = match[3].trim();
- sections.push({
- type: "file",
- content: fileContent,
- filename,
- charCount: fileContent.length,
- fileType,
- });
-
- lastIndex = match.index + match[0].length;
- }
-
- const remainingText = text.slice(lastIndex).trim();
- if (remainingText) {
- sections.push({ type: "text", content: remainingText });
- }
-
- if (sections.length === 0) {
- sections.push({ type: "text", content: text });
- }
-
- return sections;
-}
-
-type PptSlideTag = { n: number; title?: string; kind?: "outline" | "slide_image" };
-
-function extractPptSlideTags(text: string): { tags: PptSlideTag[]; rest: string } {
- const lines = String(text || "").split(/\r?\n/);
- const tags: PptSlideTag[] = [];
- const restLines: string[] = [];
- for (const line of lines) {
- const m = line.match(/^\[\[PPT_SLIDE\|(\d+)\|([^|]*)\|(outline|slide_image)\]\]$/);
- if (m) {
- const n = Number(m[1]);
- const title = String(m[2] || "").trim();
- const kind = m[3] as "outline" | "slide_image";
- if (!Number.isNaN(n)) tags.push({ n, title: title || undefined, kind });
- continue;
- }
- const legacy = line.match(/^\[\[PPT_SLIDE\|(\d+)\|(.*)\]\]$/);
- if (legacy) {
- const n = Number(legacy[1]);
- const title = String(legacy[2] || "").trim();
- if (!Number.isNaN(n)) tags.push({ n, title: title || undefined, kind: "slide_image" });
- continue;
- }
- restLines.push(line);
- }
- const rest = restLines.join("\n").replace(/^\s+|\s+$/g, "");
- return { tags, rest };
-}
-
-type ImageTag = { name?: string; url: string };
-
-function extractImageTags(text: string): { images: ImageTag[]; rest: string } {
- const lines = String(text || "").split(/\r?\n/);
- const images: ImageTag[] = [];
- const restLines: string[] = [];
- for (const line of lines) {
- const m = line.match(/^\[\[IMAGE\|([^|]*)\|([\s\S]+)\]\]$/);
- if (m) {
- const name = String(m[1] || "").trim();
- const url = String(m[2] || "").trim();
- if (url.startsWith("data:image") || url.startsWith("blob:") || url.startsWith("http://") || url.startsWith("https://")) {
- images.push({ name: name || undefined, url });
- continue;
- }
- }
- restLines.push(line);
- }
- const rest = restLines.join("\n").replace(/^\s+|\s+$/g, "");
- return { images, rest };
-}
-
-type InlinePptPart =
- | { type: "text"; text: string }
- | { type: "ppt"; n: number; title?: string; kind?: "outline" | "slide_image" };
-
-function splitInlinePptTags(text: string): InlinePptPart[] {
- const raw = String(text || "");
- const out: InlinePptPart[] = [];
- const re = /\[\[PPT_SLIDE\|(\d+)\|([\s\S]*?)\]\]/g;
- let lastIndex = 0;
- let m: RegExpExecArray | null;
- while ((m = re.exec(raw))) {
- const before = raw.slice(lastIndex, m.index);
- if (before) out.push({ type: "text", text: before });
- const n = Number(m[1]);
- const title = String(m[2] || "").trim();
- if (!Number.isNaN(n)) {
- let kind: "outline" | "slide_image" = "slide_image";
- let normalizedTitle = title;
- const parts = title.split("|");
- if (parts.length >= 2) {
- const maybeKind = parts[parts.length - 1].trim();
- if (maybeKind === "outline" || maybeKind === "slide_image") {
- kind = maybeKind;
- normalizedTitle = parts.slice(0, -1).join("|").trim();
- }
- }
- out.push({ type: "ppt", n, title: normalizedTitle || undefined, kind });
- }
- else out.push({ type: "text", text: m[0] });
- lastIndex = m.index + m[0].length;
- }
- const rest = raw.slice(lastIndex);
- if (rest) out.push({ type: "text", text: rest });
- if (out.length === 0) out.push({ type: "text", text: raw });
- return out;
-}
-
-const getMessageTextContent = (message: UIMessage): string => {
- if (message.parts && message.parts.length > 0) {
- return message.parts
- .filter((part) => part.type === "text")
- .map((part) => part.text || "")
- .join("\n");
- }
- return message.content || "";
-};
-
-const getUserOriginalText = (message: UIMessage): string => {
- const fullText = getMessageTextContent(message);
- const filePattern = /\n\n\[(PDF|File):\s*[^\]]+\]\n[\s\S]*$/;
- return fullText.replace(filePattern, "").trim();
-};
-
-function extractPptToolPayload(text: string): PptToolPayload | null {
- const raw = String(text || "").trim();
- if (!raw) return null;
- const fenced = raw.match(/```(?:json)?\s*([\s\S]*?)\s*```/i);
- const candidate = (fenced?.[1] ?? raw).trim();
- let parsed: any;
- try {
- parsed = JSON.parse(candidate);
- } catch {
- return null;
- }
- if (Array.isArray(parsed)) {
- return { type: "ppt_edit", slides: parsed.filter((s: any) => s && typeof s === "object") };
- }
- if (!parsed || typeof parsed !== "object") return null;
- const typeRaw = String((parsed as any).type || "").trim().toLowerCase();
- if (typeRaw && typeRaw !== "ppt_edit") return null;
- const type = "ppt_edit" as const;
- const slides = Array.isArray((parsed as any).slides)
- ? (parsed as any).slides.filter((s: any) => s && typeof s === "object")
- : [];
- return slides.length > 0 ? { type, slides } : null;
-}
-
-function getSlideNumber(slide: PptEditSlidePatch, index: number): number {
- const id = String(slide?.id || "");
- const m = id.match(/slide-(\d+)/i);
- if (m) {
- const n = Number(m[1]);
- if (!Number.isNaN(n)) return n;
- }
- return index + 1;
-}
-
-function slidePatchEntries(slide: PptEditSlidePatch): Array<{ key: string; value: string }> {
- const keys = ["title", "description", "layout", "note", "content"];
- const rows: Array<{ key: string; value: string }> = [];
- for (const key of keys) {
- if (!(key in slide)) continue;
- const raw = (slide as any)[key];
- if (Array.isArray(raw)) {
- const text = raw.map((x) => String(x || "").trim()).filter(Boolean).join(";");
- rows.push({ key, value: text || "(空)" });
- continue;
- }
- if (raw && typeof raw === "object") {
- rows.push({ key, value: JSON.stringify(raw) });
- continue;
- }
- rows.push({ key, value: String(raw ?? "").trim() || "(空)" });
- }
- for (const [k, v] of Object.entries(slide)) {
- if (k === "id" || keys.includes(k)) continue;
- if (v === undefined) continue;
- rows.push({ key: k, value: typeof v === "string" ? v : JSON.stringify(v) });
- }
- return rows;
-}
-
-function slideFieldLabel(key: string, uiLang: "zh" | "en"): string {
- const dict: Record<string, { zh: string; en: string }> = {
- title: { zh: "标题", en: "Title" },
- description: { zh: "画面描述", en: "Description" },
- layout: { zh: "布局", en: "Layout" },
- note: { zh: "备注", en: "Note" },
- content: { zh: "内容", en: "Content" },
- };
- const hit = dict[key];
- if (hit) return uiLang === "zh" ? hit.zh : hit.en;
- return key;
-}
-
export function ChatMessageDisplay({
messages,
setInput,
diff --git a/tests/chat-message-content.test.ts b/tests/chat-message-content.test.ts
new file mode 100644
index 0000000..4fed057
--- /dev/null
+++ b/tests/chat-message-content.test.ts
@@ -0,0 +1,205 @@
+import assert from "node:assert/strict";
+import { test } from "node:test";
+import {
+ getMessageTextContent,
+ getUserOriginalText,
+ splitTextIntoFileSections,
+ type UIMessage,
+} from "../client/shared/chat/message-content";
+import {
+ extractImageTags,
+ extractPptSlideTags,
+ extractPptToolPayload,
+ getSlideNumber,
+ slideFieldLabel,
+ slidePatchEntries,
+ splitInlinePptTags,
+} from "../client/shared/chat/ppt-message-tags";
+
+/**
+ * The readers every workspace transcript runs over a message before showing
+ * it. All three chat panels append uploaded file contents to what the user
+ * typed, and the PPT one embeds slide markers in the text; getting either
+ * wrong shows the user raw machinery instead of their own words.
+ */
+
+test("a message with no attachments comes back as one text section", () => {
+ const sections = splitTextIntoFileSections("just a question");
+ assert.deepEqual(sections, [{ type: "text", content: "just a question" }]);
+});
+
+test("an appended file becomes its own section with a character count", () => {
+ const sections = splitTextIntoFileSections("summarise this\n\n[PDF: report.pdf]\nthe extracted text");
+ assert.equal(sections.length, 2);
+ assert.deepEqual(sections[0], { type: "text", content: "summarise this" });
+ assert.equal(sections[1].type, "file");
+ assert.equal(sections[1].filename, "report.pdf");
+ assert.equal(sections[1].fileType, "pdf");
+ assert.equal(sections[1].content, "the extracted text");
+ assert.equal(sections[1].charCount, "the extracted text".length, "the count the header shows");
+});
+
+test("several attachments each get their own section", () => {
+ const sections = splitTextIntoFileSections(
+ "look\n\n[File: a.txt]\nAAA\n\n[File: b.txt]\nBBB",
+ );
+ assert.deepEqual(
+ sections.map((s) => s.filename ?? s.type),
+ ["text", "a.txt", "b.txt"],
+ );
+});
+
+test("a non-pdf attachment is typed as text", () => {
+ const [section] = splitTextIntoFileSections("[File: notes.md]\nhello");
+ assert.equal(section.fileType, "text");
+});
+
+test("an empty message is not turned into nothing", () => {
+ assert.deepEqual(splitTextIntoFileSections(""), [{ type: "text", content: "" }]);
+});
+
+test("text is read from streaming parts", () => {
+ const message: UIMessage = {
+ id: "1",
+ role: "assistant",
+ parts: [
+ { type: "text", text: "hello" },
+ { type: "tool-call", toolCallId: "x" },
+ { type: "text", text: "world" },
+ ],
+ };
+ assert.equal(getMessageTextContent(message), "hello\nworld", "non-text parts are skipped");
+});
+
+test("text is read from a multimodal content array", () => {
+ // The shape a message takes once it has been sent with an image attached.
+ const message: UIMessage = {
+ id: "1",
+ role: "user",
+ content: [
+ { type: "text", text: "what is this" },
+ { type: "image", image: "data:image/png;base64,x" },
+ ] as unknown[],
+ };
+ assert.equal(getMessageTextContent(message), "what is this");
+});
+
+test("text is read from a plain string content", () => {
+ assert.equal(getMessageTextContent({ id: "1", role: "user", content: "plain" }), "plain");
+});
+
+test("a message with nothing in it reads as empty", () => {
+ assert.equal(getMessageTextContent({ id: "1", role: "user" }), "");
+ assert.equal(getMessageTextContent({ id: "1", role: "assistant", parts: [] }), "");
+});
+
+test("the user's own words are recovered from an attachment-laden message", () => {
+ const message: UIMessage = {
+ id: "1",
+ role: "user",
+ content: "summarise this\n\n[PDF: report.pdf]\npages and pages of text",
+ };
+ assert.equal(getUserOriginalText(message), "summarise this", "editing shows what was typed");
+});
+
+test("a message without attachments is returned whole", () => {
+ assert.equal(getUserOriginalText({ id: "1", role: "user", content: " hi " }), "hi");
+});
+
+test("slide markers on their own line are lifted out of the text", () => {
+ const { tags, rest } = extractPptSlideTags("[[PPT_SLIDE|2|Revenue|outline]]\nplease redo this");
+ assert.deepEqual(tags, [{ n: 2, title: "Revenue", kind: "outline" }]);
+ assert.equal(rest, "please redo this");
+});
+
+test("a marker from before kind existed still reads as a slide image", () => {
+ const { tags } = extractPptSlideTags("[[PPT_SLIDE|3|Old Title]]");
+ assert.deepEqual(tags, [{ n: 3, title: "Old Title", kind: "slide_image" }]);
+});
+
+test("a marker with no title keeps the slide number", () => {
+ const { tags } = extractPptSlideTags("[[PPT_SLIDE|4||outline]]");
+ assert.deepEqual(tags, [{ n: 4, title: undefined, kind: "outline" }]);
+});
+
+test("image markers are lifted out, and only for renderable urls", () => {
+ const { images, rest } = extractImageTags(
+ ["[[IMAGE|shot|data:image/png;base64,AAA]]", "[[IMAGE|bad|javascript:alert(1)]]", "text"].join("\n"),
+ );
+ assert.deepEqual(images, [{ name: "shot", url: "data:image/png;base64,AAA" }]);
+ assert.ok(rest.includes("javascript:alert(1)"), "a rejected marker stays as text rather than rendering");
+ assert.ok(rest.includes("text"));
+});
+
+test("a marker mid-sentence splits into text and a chip", () => {
+ const parts = splitInlinePptTags("please fix [[PPT_SLIDE|2|Revenue|outline]] today");
+ assert.deepEqual(parts, [
+ { type: "text", text: "please fix " },
+ { type: "ppt", n: 2, title: "Revenue", kind: "outline" },
+ { type: "text", text: " today" },
+ ]);
+});
+
+test("a title containing a pipe survives the split", () => {
+ // Only a trailing known kind is treated as the kind, so titles keep theirs.
+ const parts = splitInlinePptTags("[[PPT_SLIDE|1|A|B|slide_image]]");
+ assert.deepEqual(parts, [{ type: "ppt", n: 1, title: "A|B", kind: "slide_image" }]);
+});
+
+test("text with no markers comes back as a single run", () => {
+ assert.deepEqual(splitInlinePptTags("nothing here"), [{ type: "text", text: "nothing here" }]);
+});
+
+test("a slide-edit payload is read whether fenced or bare", () => {
+ const slides = '{"type":"ppt_edit","slides":[{"id":"slide-2","title":"New"}]}';
+ for (const raw of [slides, "Sure!\n```json\n" + slides + "\n```"]) {
+ const payload = extractPptToolPayload(raw);
+ assert.equal(payload?.slides[0].title, "New", raw);
+ }
+});
+
+test("a naked array of slides counts as an edit payload", () => {
+ const payload = extractPptToolPayload('[{"id":"slide-1","title":"A"}]');
+ assert.equal(payload?.type, "ppt_edit");
+ assert.equal(payload?.slides.length, 1);
+});
+
+test("anything that is not a slide edit is refused", () => {
+ assert.equal(extractPptToolPayload(""), null);
+ assert.equal(extractPptToolPayload("just prose"), null);
+ assert.equal(extractPptToolPayload('{"type":"cad_patch","edits":[]}'), null, "another tool's payload");
+ assert.equal(extractPptToolPayload('{"type":"ppt_edit","slides":[]}'), null, "nothing to apply");
+});
+
+test("the slide number comes from the id, and falls back to position", () => {
+ assert.equal(getSlideNumber({ id: "slide-7" }, 0), 7);
+ assert.equal(getSlideNumber({ id: "cover" }, 3), 4, "1-based position");
+ assert.equal(getSlideNumber({}, 0), 1);
+});
+
+test("a patch is flattened into rows in a readable order", () => {
+ const rows = slidePatchEntries({ id: "slide-1", note: "n", title: "T", content: ["a", "b"] });
+ assert.deepEqual(
+ rows.map((r) => r.key),
+ ["title", "note", "content"],
+ "known fields come first, in their own order",
+ );
+ assert.equal(rows[2].value, "a;b");
+});
+
+test("an empty field is shown as empty rather than dropped", () => {
+ const rows = slidePatchEntries({ title: "", content: [] });
+ assert.equal(rows.find((r) => r.key === "title")?.value, "(空)");
+ assert.equal(rows.find((r) => r.key === "content")?.value, "(空)");
+});
+
+test("a field the model invented is still shown", () => {
+ const rows = slidePatchEntries({ id: "slide-1", surprise: "value" });
+ assert.deepEqual(rows, [{ key: "surprise", value: "value" }]);
+});
+
+test("field labels follow the language and fall back to the key", () => {
+ assert.equal(slideFieldLabel("title", "zh"), "标题");
+ assert.equal(slideFieldLabel("title", "en"), "Title");
+ assert.equal(slideFieldLabel("surprise", "en"), "surprise");
+});