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, 72 insertions, 3 deletions

diff --git a/packages/core/tests/models/catalog.test.ts b/packages/core/tests/models/catalog.test.ts
index 51043e6..f4bddc2 100644
--- a/packages/core/tests/models/catalog.test.ts
+++ b/packages/core/tests/models/catalog.test.ts
@@ -4,6 +4,7 @@ import {
 	__resetCatalogCacheForTests,
 	getModelsCatalog,
 	resolveContextLimit,
+	resolveModelCapabilities,
 } from "../../src/models/catalog.js";
 
 const CACHE_PATH = "/tmp/dispatch/models-dev.json";
@@ -13,14 +14,30 @@ const CATALOG = {
 	anthropic: {
 		id: "anthropic",
 		models: {
-			"claude-sonnet-4-5": { limit: { context: 200000, output: 64000 } },
-			"claude-sonnet-4-6": { limit: { context: 1000000, output: 64000 } },
+			"claude-sonnet-4-5": {
+				limit: { context: 200000, output: 64000 },
+				modalities: { input: ["text", "image", "pdf"], output: ["text"] },
+			},
+			"claude-sonnet-4-6": {
+				limit: { context: 1000000, output: 64000 },
+				modalities: { input: ["text", "image", "pdf"], output: ["text"] },
+			},
+			// A text-only model: definitively no image/pdf input.
+			"text-only-model": {
+				limit: { context: 100000, output: 8192 },
+				modalities: { input: ["text"], output: ["text"] },
+			},
+			// An entry predating the modalities field → capability unknown.
+			"legacy-model": { limit: { context: 100000, output: 8192 } },
 		},
 	},
 	opencode: {
 		id: "opencode",
 		models: {
-			"glm-4-6": { limit: { context: 131072, output: 8192 } },
+			"glm-4-6": {
+				limit: { context: 131072, output: 8192 },
+				modalities: { input: ["text", "image"], output: ["text"] },
+			},
 		},
 	},
 };
@@ -156,3 +173,55 @@ describe("getModelsCatalog caching", () => {
 		warn.mockRestore();
 	});
 });
+
+describe("resolveModelCapabilities", () => {
+	it("reports image + pdf for a vision model", async () => {
+		mockFetchOnce(CATALOG);
+		expect(await resolveModelCapabilities("anthropic", "claude-sonnet-4-5")).toEqual({
+			image: true,
+			pdf: true,
+		});
+	});
+
+	it("reports image-only for a model whose modalities omit pdf", async () => {
+		mockFetchOnce(CATALOG);
+		// glm-4-6 lists image but not pdf (resolved via the opencode fallback).
+		expect(await resolveModelCapabilities("opencode-anthropic", "glm-4-6")).toEqual({
+			image: true,
+			pdf: false,
+		});
+	});
+
+	it("reports a definitive no for a text-only model", async () => {
+		mockFetchOnce(CATALOG);
+		expect(await resolveModelCapabilities("anthropic", "text-only-model")).toEqual({
+			image: false,
+			pdf: false,
+		});
+	});
+
+	it("returns null (unknown) for an entry without modalities", async () => {
+		mockFetchOnce(CATALOG);
+		expect(await resolveModelCapabilities("anthropic", "legacy-model")).toBeNull();
+	});
+
+	it("returns null (unknown) for an unknown model id", async () => {
+		mockFetchOnce(CATALOG);
+		expect(await resolveModelCapabilities("anthropic", "no-such-model")).toBeNull();
+	});
+
+	it("returns null for an unsupported provider without hitting the network", async () => {
+		const fetchFn = mockFetchOnce(CATALOG);
+		expect(await resolveModelCapabilities("google", "gemini-2.5-pro")).toBeNull();
+		expect(await resolveModelCapabilities("anthropic", "")).toBeNull();
+		expect(fetchFn).not.toHaveBeenCalled();
+	});
+
+	it("returns null (unknown) when the catalog is offline with no cache", async () => {
+		const fetchFn = vi.fn(() => Promise.reject(new Error("offline")));
+		vi.stubGlobal("fetch", fetchFn);
+		const warn = vi.spyOn(console, "warn").mockImplementation(() => {});
+		expect(await resolveModelCapabilities("anthropic", "claude-sonnet-4-5")).toBeNull();
+		warn.mockRestore();
+	});
+});