fix(enhance): include note images in summaries

apps/desktop/src/settings/ai/shared/model-capabilities.test.ts

@@ -0,0 +1,33 @@
+import { describe, expect, it } from "vitest";
+
+import { modelSupportsImageInput } from "./model-capabilities";
+
+describe("modelSupportsImageInput", () => {
+  it("allows known multimodal hosted models", () => {
+    expect(modelSupportsImageInput("hyprnote", "Auto")).toBe(true);
+    expect(modelSupportsImageInput("openai", "gpt-4o")).toBe(true);
+    expect(modelSupportsImageInput("anthropic", "claude-3-5-sonnet")).toBe(
+      true,
+    );
+    expect(
+      modelSupportsImageInput("google_generative_ai", "gemini-2.5-pro"),
+    ).toBe(true);
+  });
+
+  it("blocks known text-only or non-chat models", () => {
+    expect(modelSupportsImageInput("openai", "gpt-3.5-turbo")).toBe(false);
+    expect(modelSupportsImageInput("openai", "gpt-4")).toBe(false);
+    expect(modelSupportsImageInput("anthropic", "claude-2.1")).toBe(false);
+    expect(modelSupportsImageInput("anthropic", "custom-text-model")).toBe(
+      false,
+    );
+    expect(modelSupportsImageInput("openai", "text-embedding-3-large")).toBe(
+      false,
+    );
+  });
+
+  it("requires a vision-like model name for unknown local providers", () => {
+    expect(modelSupportsImageInput("ollama", "llava:latest")).toBe(true);
+    expect(modelSupportsImageInput("custom", "llama-3.1-8b")).toBe(false);
+  });
+});

apps/desktop/src/settings/ai/shared/model-capabilities.ts

@@ -0,0 +1,28 @@
+const TEXT_ONLY_MODEL_RE =
+  /(?:^|[/:\-.])(?:gpt-3\.5|claude-2|claude-instant|davinci|babbage|curie|ada|dall-e|sora|gpt-image|image-generation|embed|embedding|whisper|tts|transcribe|moderation|realtime|computer)(?:$|[/:\-.])/i;
+
+const IMAGE_INPUT_MODEL_RE =
+  /(?:gpt-4o|gpt-4\.1|gpt-5|claude-3|claude-sonnet|claude-opus|claude-haiku|gemini|pixtral|vision|vl|llava|llama-3\.2-vision|llama3\.2-vision|moondream|minicpm-v|internvl|qwen(?:2|2\.5|3)?-vl|gemma-3|gemma3)/i;
+
+export function modelSupportsImageInput(
+  providerId: string | undefined,
+  modelId: string | undefined,
+): boolean {
+  if (!providerId || !modelId) {
+    return false;
+  }
+
+  if (TEXT_ONLY_MODEL_RE.test(modelId)) {
+    return false;
+  }
+
+  if (providerId === "hyprnote" && modelId === "Auto") {
+    return true;
+  }
+
+  if (IMAGE_INPUT_MODEL_RE.test(modelId)) {
+    return true;
+  }
+
+  return false;
+}

apps/desktop/src/store/zustand/ai-task/task-configs/enhance-images.test.ts

@@ -0,0 +1,121 @@
+import { beforeEach, describe, expect, it, vi } from "vitest";
+
+import {
+  collectEnhanceImageContext,
+  collectImageReferences,
+  getBase64ByteLength,
+} from "./enhance-images";
+
+const fsSyncMocks = vi.hoisted(() => ({
+  attachmentList: vi.fn(),
+  attachmentRead: vi.fn(),
+}));
+
+vi.mock("@hypr/plugin-fs-sync", () => ({
+  commands: fsSyncMocks,
+}));
+
+describe("enhance image context", () => {
+  beforeEach(() => {
+    vi.clearAllMocks();
+    fsSyncMocks.attachmentList.mockResolvedValue({
+      status: "ok",
+      data: [
+        {
+          attachmentId: "diagram.png",
+          path: "/vault/sessions/session-1/attachments/diagram.png",
+          extension: "png",
+          modifiedAt: "",
+        },
+        {
+          attachmentId: "stale.png",
+          path: "/vault/sessions/session-1/attachments/stale.png",
+          extension: "png",
+          modifiedAt: "",
+        },
+      ],
+    });
+    fsSyncMocks.attachmentRead.mockResolvedValue({
+      status: "ok",
+      data: [104, 101, 108, 108, 111],
+    });
+  });
+
+  it("reads only image attachments referenced by note JSON", async () => {
+    const rawContent = JSON.stringify({
+      type: "doc",
+      content: [
+        {
+          type: "image",
+          attrs: {
+            src: "asset://localhost/%2Fvault%2Fsessions%2Fsession-1%2Fattachments%2Fdiagram.png",
+            attachmentId: "diagram.png",
+          },
+        },
+      ],
+    });
+
+    const images = await collectEnhanceImageContext("session-1", rawContent);
+
+    expect(images).toEqual([
+      {
+        base64: "aGVsbG8=",
+        mimeType: "image/png",
+        filename: "diagram.png",
+      },
+    ]);
+    expect(fsSyncMocks.attachmentRead).toHaveBeenCalledWith(
+      "session-1",
+      "diagram.png",
+    );
+  });
+
+  it("extracts markdown image filenames from asset URLs", () => {
+    expect(
+      collectImageReferences(
+        "![diagram](asset://localhost/%2Fvault%2Fsessions%2Fsession-1%2Fattachments%2Fdiagram.png)",
+      ),
+    ).toEqual([{ filename: "diagram.png" }]);
+  });
+
+  it("does not treat remote markdown images as local attachments", () => {
+    expect(
+      collectImageReferences("![diagram](https://example.com/diagram.png)"),
+    ).toEqual([{ filename: undefined }]);
+  });
+
+  it("keeps base64 data URL images without reading attachments", async () => {
+    const images = await collectEnhanceImageContext(
+      "session-1",
+      "![pasted](data:image/png;base64,abc123)",
+    );
+
+    expect(images).toEqual([{ base64: "abc123", mimeType: "image/png" }]);
+    expect(fsSyncMocks.attachmentList).not.toHaveBeenCalled();
+  });
+
+  it("does not load an attachment again for a node that already has a data URL", async () => {
+    const rawContent = JSON.stringify({
+      type: "doc",
+      content: [
+        {
+          type: "image",
+          attrs: {
+            src: "data:image/png;base64,abc123",
+            attachmentId: "diagram.png",
+          },
+        },
+      ],
+    });
+
+    const images = await collectEnhanceImageContext("session-1", rawContent);
+
+    expect(images).toEqual([{ base64: "abc123", mimeType: "image/png" }]);
+    expect(fsSyncMocks.attachmentList).not.toHaveBeenCalled();
+  });
+
+  it("computes decoded base64 byte length before applying the data URL cap", () => {
+    expect(getBase64ByteLength("aGVsbG8=")).toBe(5);
+    expect(getBase64ByteLength("YW55IGNhcm5hbCBwbGVhcw==")).toBe(16);
+  });
+});