Add inline image previews in assistant messages

2026-04-12 22:41:54 +07:00
parent ec445e606c
commit 4547d597ef
9 changed files with 738 additions and 11 deletions
@@ -0,0 +1,58 @@
+import { beforeEach, describe, expect, it } from "vitest";
+import {
+  clearAssistantImageMetadataCache,
+  estimateAssistantMessageHeightFromCache,
+  extractAssistantImageSources,
+  getAssistantImageMetadata,
+  setAssistantImageMetadata,
+} from "./assistant-image-metadata";
+
+describe("assistant image metadata", () => {
+  beforeEach(() => {
+    clearAssistantImageMetadataCache();
+  });
+
+  it("extracts markdown image sources", () => {
+    expect(
+      extractAssistantImageSources(
+        'Before\n\n![local](/tmp/paseo.png)\n\n![remote](https://example.com/test.png "Remote")',
+      ),
+    ).toEqual(["/tmp/paseo.png", "https://example.com/test.png"]);
+  });
+
+  it("reuses cached metadata across canonical and raw source keys", () => {
+    setAssistantImageMetadata(
+      {
+        source: "/tmp/paseo-codex-screenshot.png",
+        workspaceRoot: "/Users/moboudra/dev/paseo",
+        serverId: "server-1",
+      },
+      { width: 1200, height: 800 },
+    );
+
+    expect(
+      getAssistantImageMetadata({
+        source: "/tmp/paseo-codex-screenshot.png",
+      }),
+    ).toEqual({
+      width: 1200,
+      height: 800,
+      aspectRatio: 1.5,
+    });
+  });
+
+  it("estimates assistant message height from cached image metadata", () => {
+    setAssistantImageMetadata(
+      {
+        source: "https://example.com/landscape.png",
+      },
+      { width: 1200, height: 800 },
+    );
+
+    expect(
+      estimateAssistantMessageHeightFromCache(
+        "Here is the screenshot\n\n![Screenshot](https://example.com/landscape.png)",
+      ),
+    ).toBeGreaterThan(220);
+  });
+});
@@ -0,0 +1,213 @@
+import { MAX_CONTENT_WIDTH } from "@/constants/layout";
+import { resolveAssistantImageSource } from "@/utils/assistant-image-source";
+
+export interface AssistantImageMetadata {
+  width: number;
+  height: number;
+  aspectRatio: number;
+}
+
+const assistantImageMetadataCache = new Map<string, AssistantImageMetadata>();
+const assistantImageParseCache = new Map<string, { sources: string[]; hasNonImageText: boolean }>();
+const ASSISTANT_IMAGE_METADATA_CACHE_LIMIT = 500;
+const ASSISTANT_IMAGE_PARSE_CACHE_LIMIT = 500;
+
+const MARKDOWN_IMAGE_PATTERN = /!\[[^\]]*]\((<[^>]+>|[^)\n]+)\)/g;
+const ASSISTANT_IMAGE_ESTIMATE_WIDTH = MAX_CONTENT_WIDTH - 8;
+const ASSISTANT_IMAGE_MIN_HEIGHT = 160;
+const ASSISTANT_IMAGE_BLOCK_GAP = 24;
+const ASSISTANT_MESSAGE_BASE_HEIGHT = 96;
+const ASSISTANT_MESSAGE_MIN_HEIGHT = 220;
+const ASSISTANT_MESSAGE_IMAGE_ONLY_BASE_HEIGHT = 40;
+
+function touchCacheEntry<K, V>(cache: Map<K, V>, key: K, value: V, limit: number): void {
+  cache.delete(key);
+  cache.set(key, value);
+  if (cache.size <= limit) {
+    return;
+  }
+  const oldestKey = cache.keys().next().value;
+  if (oldestKey !== undefined) {
+    cache.delete(oldestKey);
+  }
+}
+
+function normalizeAssistantImageSourceToken(value: string): string | null {
+  const trimmed = value.trim();
+  if (!trimmed) {
+    return null;
+  }
+
+  if (trimmed.startsWith("<") && trimmed.endsWith(">")) {
+    const inner = trimmed.slice(1, -1).trim();
+    return inner || null;
+  }
+
+  const titleMatch = /^(.*?)(?:\s+(['"]).*?\2)?$/.exec(trimmed);
+  const source = titleMatch?.[1]?.trim() ?? trimmed;
+  return source || null;
+}
+
+function createSourceAliasKey(source: string): string {
+  return `source:${source}`;
+}
+
+function createResolutionKey(input: {
+  source: string;
+  workspaceRoot?: string;
+  serverId?: string;
+}): string | null {
+  const resolution = resolveAssistantImageSource({
+    source: input.source,
+    workspaceRoot: input.workspaceRoot,
+  });
+  if (!resolution) {
+    return null;
+  }
+  if (resolution.kind === "direct") {
+    return `direct:${resolution.uri}`;
+  }
+  return `file:${input.serverId ?? "unknown-server"}:${resolution.cwd}:${resolution.path}`;
+}
+
+function getAssistantImageMetadataKeys(input: {
+  source: string;
+  workspaceRoot?: string;
+  serverId?: string;
+}): string[] {
+  const source = input.source.trim();
+  if (!source) {
+    return [];
+  }
+
+  const keys = [createSourceAliasKey(source)];
+  const resolutionKey = createResolutionKey(input);
+  if (resolutionKey) {
+    keys.unshift(resolutionKey);
+  }
+  return [...new Set(keys)];
+}
+
+export function getAssistantImageMetadata(input: {
+  source: string;
+  workspaceRoot?: string;
+  serverId?: string;
+}): AssistantImageMetadata | null {
+  for (const key of getAssistantImageMetadataKeys(input)) {
+    const metadata = assistantImageMetadataCache.get(key);
+    if (metadata) {
+      touchCacheEntry(
+        assistantImageMetadataCache,
+        key,
+        metadata,
+        ASSISTANT_IMAGE_METADATA_CACHE_LIMIT,
+      );
+      return metadata;
+    }
+  }
+  return null;
+}
+
+export function setAssistantImageMetadata(
+  input: {
+    source: string;
+    workspaceRoot?: string;
+    serverId?: string;
+  },
+  dimensions: { width: number; height: number },
+): AssistantImageMetadata | null {
+  const { width, height } = dimensions;
+  if (!Number.isFinite(width) || !Number.isFinite(height) || width <= 0 || height <= 0) {
+    return null;
+  }
+
+  const metadata: AssistantImageMetadata = {
+    width,
+    height,
+    aspectRatio: width / height,
+  };
+
+  for (const key of getAssistantImageMetadataKeys(input)) {
+    touchCacheEntry(
+      assistantImageMetadataCache,
+      key,
+      metadata,
+      ASSISTANT_IMAGE_METADATA_CACHE_LIMIT,
+    );
+  }
+
+  return metadata;
+}
+
+export function extractAssistantImageSources(markdown: string): string[] {
+  const cachedParse = assistantImageParseCache.get(markdown);
+  if (cachedParse) {
+    touchCacheEntry(
+      assistantImageParseCache,
+      markdown,
+      cachedParse,
+      ASSISTANT_IMAGE_PARSE_CACHE_LIMIT,
+    );
+    return cachedParse.sources;
+  }
+
+  const sources: string[] = [];
+  for (const match of markdown.matchAll(MARKDOWN_IMAGE_PATTERN)) {
+    const normalized = normalizeAssistantImageSourceToken(match[1] ?? "");
+    if (normalized) {
+      sources.push(normalized);
+    }
+  }
+  const hasNonImageText = markdown.replace(MARKDOWN_IMAGE_PATTERN, "").trim().length > 0;
+  touchCacheEntry(
+    assistantImageParseCache,
+    markdown,
+    { sources, hasNonImageText },
+    ASSISTANT_IMAGE_PARSE_CACHE_LIMIT,
+  );
+  return sources;
+}
+
+export function estimateAssistantMessageHeightFromCache(markdown: string): number | null {
+  const cachedParse = assistantImageParseCache.get(markdown);
+  const parsed =
+    cachedParse ??
+    (() => {
+      const sources = extractAssistantImageSources(markdown);
+      const nextParsed = assistantImageParseCache.get(markdown);
+      return nextParsed ?? { sources, hasNonImageText: true };
+    })();
+  if (parsed.sources.length === 0) {
+    return null;
+  }
+
+  const knownHeights = parsed.sources
+    .map((source) => getAssistantImageMetadata({ source }))
+    .filter((metadata): metadata is AssistantImageMetadata => metadata !== null)
+    .map((metadata) =>
+      Math.max(
+        ASSISTANT_IMAGE_MIN_HEIGHT,
+        Math.round(ASSISTANT_IMAGE_ESTIMATE_WIDTH / metadata.aspectRatio),
+      ),
+    );
+
+  if (knownHeights.length === 0) {
+    return null;
+  }
+
+  const baseHeight = parsed.hasNonImageText
+    ? ASSISTANT_MESSAGE_BASE_HEIGHT
+    : ASSISTANT_MESSAGE_IMAGE_ONLY_BASE_HEIGHT;
+
+  const estimatedHeight =
+    baseHeight +
+    knownHeights.reduce((sum, height) => sum + height, 0) +
+    ASSISTANT_IMAGE_BLOCK_GAP * knownHeights.length;
+
+  return Math.max(ASSISTANT_MESSAGE_MIN_HEIGHT, estimatedHeight);
+}
+
+export function clearAssistantImageMetadataCache(): void {
+  assistantImageMetadataCache.clear();
+  assistantImageParseCache.clear();
+}
@@ -0,0 +1,81 @@
+import { describe, expect, it } from "vitest";
+
+import { resolveAssistantImageSource } from "./assistant-image-source";
+
+describe("resolveAssistantImageSource", () => {
+  it("passes through direct image URIs", () => {
+    expect(resolveAssistantImageSource({ source: "https://example.com/image.png" })).toEqual({
+      kind: "direct",
+      uri: "https://example.com/image.png",
+    });
+    expect(resolveAssistantImageSource({ source: "data:image/png;base64,abc" })).toEqual({
+      kind: "direct",
+      uri: "data:image/png;base64,abc",
+    });
+  });
+
+  it("uses the workspace root for relative paths", () => {
+    expect(
+      resolveAssistantImageSource({
+        source: "screenshots/output.png",
+        workspaceRoot: "/Users/test/project",
+      }),
+    ).toEqual({
+      kind: "file_rpc",
+      cwd: "/Users/test/project",
+      path: "screenshots/output.png",
+    });
+  });
+
+  it("uses the workspace root for absolute paths inside the workspace", () => {
+    expect(
+      resolveAssistantImageSource({
+        source: "/Users/test/project/screenshots/output.png",
+        workspaceRoot: "/Users/test/project",
+      }),
+    ).toEqual({
+      kind: "file_rpc",
+      cwd: "/Users/test/project",
+      path: "/Users/test/project/screenshots/output.png",
+    });
+  });
+
+  it("falls back to filesystem root for absolute paths outside the workspace", () => {
+    expect(
+      resolveAssistantImageSource({
+        source: "/tmp/paseo-codex-screenshot.png",
+        workspaceRoot: "/Users/test/project",
+      }),
+    ).toEqual({
+      kind: "file_rpc",
+      cwd: "/",
+      path: "/tmp/paseo-codex-screenshot.png",
+    });
+  });
+
+  it("normalizes file URIs into file RPC requests", () => {
+    expect(
+      resolveAssistantImageSource({
+        source: "file:///tmp/paseo-codex-screenshot.png",
+        workspaceRoot: "/Users/test/project",
+      }),
+    ).toEqual({
+      kind: "file_rpc",
+      cwd: "/",
+      path: "/tmp/paseo-codex-screenshot.png",
+    });
+  });
+
+  it("falls back to the drive root for Windows absolute paths", () => {
+    expect(
+      resolveAssistantImageSource({
+        source: "C:/Users/test/Desktop/screenshot.png",
+        workspaceRoot: "D:/repo",
+      }),
+    ).toEqual({
+      kind: "file_rpc",
+      cwd: "C:/",
+      path: "C:/Users/test/Desktop/screenshot.png",
+    });
+  });
+});
@@ -0,0 +1,109 @@
+import { fileUriToPath } from "@/attachments/utils";
+import { isAbsolutePath } from "@/utils/path";
+
+export type AssistantImageSourceResolution =
+  | { kind: "direct"; uri: string }
+  | { kind: "file_rpc"; cwd: string; path: string };
+
+function trimTrailingSeparators(value: string): string {
+  if (value === "/" || /^[A-Za-z]:[\\/]?$/.test(value)) {
+    return value.replace(/\\/g, "/");
+  }
+  return value.replace(/[\\/]+$/, "");
+}
+
+function normalizeForPathComparison(value: string): string {
+  const normalized = trimTrailingSeparators(value.replace(/\\/g, "/"));
+  if (/^[A-Za-z]:\//.test(normalized)) {
+    return `${normalized.slice(0, 1).toUpperCase()}${normalized.slice(1)}`;
+  }
+  return normalized;
+}
+
+function isPathWithinRoot(candidatePath: string, rootPath: string): boolean {
+  const candidate = normalizeForPathComparison(candidatePath);
+  const root = normalizeForPathComparison(rootPath);
+  if (!candidate || !root) {
+    return false;
+  }
+  if (root === "/") {
+    return candidate.startsWith("/");
+  }
+  if (candidate === root) {
+    return true;
+  }
+  return candidate.startsWith(`${root}/`);
+}
+
+function deriveFallbackRootFromAbsolutePath(value: string): string | null {
+  if (value.startsWith("/")) {
+    return "/";
+  }
+
+  const driveMatch = /^([A-Za-z]:)[\\/]/.exec(value);
+  if (driveMatch?.[1]) {
+    return `${driveMatch[1]}/`;
+  }
+
+  const uncMatch = /^(\\\\[^\\]+\\[^\\]+)/.exec(value);
+  if (uncMatch?.[1]) {
+    return uncMatch[1];
+  }
+
+  return null;
+}
+
+export function resolveAssistantImageSource(input: {
+  source: string;
+  workspaceRoot?: string;
+}): AssistantImageSourceResolution | null {
+  const source = input.source.trim();
+  if (!source) {
+    return null;
+  }
+
+  if (/^(https?:|data:|blob:)/i.test(source)) {
+    return { kind: "direct", uri: source };
+  }
+
+  const sourcePath = source.startsWith("file://") ? fileUriToPath(source) : source;
+  if (!sourcePath) {
+    return null;
+  }
+
+  if (!isAbsolutePath(sourcePath)) {
+    const workspaceRoot = input.workspaceRoot?.trim();
+    if (!workspaceRoot || !isAbsolutePath(workspaceRoot)) {
+      return null;
+    }
+    return {
+      kind: "file_rpc",
+      cwd: workspaceRoot,
+      path: sourcePath,
+    };
+  }
+
+  const workspaceRoot = input.workspaceRoot?.trim();
+  if (
+    workspaceRoot &&
+    isAbsolutePath(workspaceRoot) &&
+    isPathWithinRoot(sourcePath, workspaceRoot)
+  ) {
+    return {
+      kind: "file_rpc",
+      cwd: workspaceRoot,
+      path: sourcePath,
+    };
+  }
+
+  const fallbackRoot = deriveFallbackRootFromAbsolutePath(sourcePath);
+  if (!fallbackRoot) {
+    return null;
+  }
+
+  return {
+    kind: "file_rpc",
+    cwd: fallbackRoot,
+    path: sourcePath,
+  };
+}