feat(chat): paste-to-attach images/PDFs with model capability check

Add multimodal image/PDF input to the chat box via clipboard paste, gated by a
graceful per-model capability check.

UX: a pasted image/PDF inserts an inline token (【image:…】 / 【pdf:…】) into the
draft, so attachments have ORDER relative to typed text and can be referenced
positionally. The token is the only handle — deleting it (atomic Backspace/
Delete, or selection overlap) detaches the file; an input-reconciliation safety
net detaches any attachment whose token is no longer intact. No preview strip.

Capability check: resolveModelCapabilities reads models.dev modalities.input
(new GET /models/capabilities, mirrors /context-limit). The input blocks Send
(no tokens spent) only on a definitive 'no'; unknown capability (catalog offline
/ unmapped provider) stays permissive. Attachments require a fresh turn — Send is
blocked while generating and /chat rejects content mid-turn (409).

Attachments are EPHEMERAL: forwarded to the model for the turn via ordered AI SDK
ImagePart/FilePart content, but never persisted (history keeps the text with
[image]/[pdf] markers). Text-only turns serialize byte-identically to before.

Limits (Anthropic-aligned, enforced at paste + re-validated server-side):
PNG/JPEG/WebP/GIF/PDF; image ≤5MB, PDF ≤32MB, ≤20 attachments, ≤32MB total.

core: UserContentPart types, models/attachments validator, capability resolver,
agent.run+toModelMessages thread ordered content. api: /chat content validation +
passthrough. frontend: attachment-tokens helper, ChatInput paste/token/gating,
per-tab staged attachments, App.svelte capability fetch. +44 tests.

1 files changed, 98 insertions, 0 deletions

diff --git a/packages/core/tests/agent/agent.test.ts b/packages/core/tests/agent/agent.test.ts
index d8edec7..f4b33cc 100644
--- a/packages/core/tests/agent/agent.test.ts
+++ b/packages/core/tests/agent/agent.test.ts
@@ -1544,4 +1544,102 @@ describe("anthropicThinkingProviderOptions — adaptive-thinking model detection
 			effort: "xhigh",
 		});
 	});
+
+	describe("multimodal user content", () => {
+		it("emits ordered text + image parts to the model when content is provided", async () => {
+			vi.mocked(streamText).mockReturnValue(
+				makeMockStreamResult([{ type: "text-delta", id: "t0", text: "ok" }, finishStop]),
+			);
+
+			const agent = new Agent(makeConfig());
+			for await (const _ of agent.run("here is image A: [image]", {
+				content: [
+					{ type: "text", text: "here is image A: " },
+					{ type: "attachment", mediaType: "image/png", data: "QQ==" },
+				],
+			})) {
+				// consume
+			}
+
+			const callArgs = vi.mocked(streamText).mock.calls.at(-1)?.[0];
+			const messages = callArgs?.messages as Array<{ role: string; content: unknown }>;
+			const userMsg = messages.find((m) => m.role === "user");
+			expect(userMsg).toBeDefined();
+			// Multimodal turn → content is an ordered parts array, not a string.
+			expect(Array.isArray(userMsg?.content)).toBe(true);
+			const parts = userMsg?.content as Array<Record<string, unknown>>;
+			expect(parts[0]).toMatchObject({ type: "text", text: "here is image A: " });
+			expect(parts[1]).toMatchObject({ type: "image", mediaType: "image/png" });
+			expect(String(parts[1]?.image)).toBe("data:image/png;base64,QQ==");
+		});
+
+		it("emits a FilePart for a PDF attachment with its filename", async () => {
+			vi.mocked(streamText).mockReturnValue(
+				makeMockStreamResult([{ type: "text-delta", id: "t0", text: "ok" }, finishStop]),
+			);
+
+			const agent = new Agent(makeConfig());
+			for await (const _ of agent.run("see [pdf]", {
+				content: [
+					{ type: "text", text: "see " },
+					{ type: "attachment", mediaType: "application/pdf", data: "QQ==", name: "doc.pdf" },
+				],
+			})) {
+				// consume
+			}
+
+			const callArgs = vi.mocked(streamText).mock.calls.at(-1)?.[0];
+			const messages = callArgs?.messages as Array<{ role: string; content: unknown }>;
+			const userMsg = messages.find((m) => m.role === "user");
+			const parts = userMsg?.content as Array<Record<string, unknown>>;
+			const filePart = parts.find((p) => p.type === "file");
+			expect(filePart).toMatchObject({
+				type: "file",
+				mediaType: "application/pdf",
+				filename: "doc.pdf",
+			});
+			expect(String(filePart?.data)).toBe("data:application/pdf;base64,QQ==");
+		});
+
+		it("persists the user turn as text only (no content) for history", async () => {
+			vi.mocked(streamText).mockReturnValue(
+				makeMockStreamResult([{ type: "text-delta", id: "t0", text: "ok" }, finishStop]),
+			);
+
+			const agent = new Agent(makeConfig());
+			for await (const _ of agent.run("look: [image]", {
+				content: [
+					{ type: "text", text: "look: " },
+					{ type: "attachment", mediaType: "image/png", data: "QQ==" },
+				],
+			})) {
+				// consume
+			}
+
+			// The in-memory user message keeps the text chunk for the render/persist
+			// path; the ephemeral `content` rides alongside it but isn't a chunk.
+			const userMsg = agent.messages.find((m) => m.role === "user");
+			expect(userMsg?.chunks).toEqual([{ type: "text", text: "look: [image]" }]);
+		});
+
+		it("falls back to a plain string when content has no attachment", async () => {
+			vi.mocked(streamText).mockReturnValue(
+				makeMockStreamResult([{ type: "text-delta", id: "t0", text: "ok" }, finishStop]),
+			);
+
+			const agent = new Agent(makeConfig());
+			for await (const _ of agent.run("plain text", {
+				content: [{ type: "text", text: "plain text" }],
+			})) {
+				// consume
+			}
+
+			const callArgs = vi.mocked(streamText).mock.calls.at(-1)?.[0];
+			const messages = callArgs?.messages as Array<{ role: string; content: unknown }>;
+			const userMsg = messages.find((m) => m.role === "user");
+			// No attachment → plain string content (byte-identical to text-only path).
+			expect(typeof userMsg?.content).toBe("string");
+			expect(userMsg?.content).toBe("plain text");
+		});
+	});
 });