fix(catalog/umans): treated supports_vision sentinels as text-only

`umans-glm-5.1` / `umans-glm-5.2` advertise themselves on the Umans
`models/info` endpoint with `supports_vision: "via-handoff"`. That
sentinel means image inputs are routed through a separate vision
handoff pre-analysis step; the GLM endpoint itself rejects raw image
blocks with `400 This model does not support image inputs`.

`umansSupportsVision` was returning `true` for any non-empty string,
so dynamic discovery mapped the GLM models to `input: ["text",
"image"]` and the agent sent images straight to GLM. The bundled
`umans-glm-5.1` / `umans-glm-5.2` rows in `models.json` carried the
same stale `["text","image"]` from a previous regen.

- Tighten `umansSupportsVision` to `value === true`; document the
  sentinel contract.
- Correct the two bundled rows to `input: ["text"]` so the vision
  handoff path runs.
- Add resolver- and bundle-level regression tests that cover
  `supports_vision: "via-handoff"` alongside the native-vision
  `umans-coder` case.

Fixes #3184

packages/catalog/CHANGELOG.md

@@ -2,6 +2,10 @@
 
 ## [Unreleased]
 
+### Fixed
+
+- Fixed Umans `umans-glm-5.1` / `umans-glm-5.2` advertising native image input. The `models/info` endpoint reports `supports_vision: "via-handoff"` for the GLM models, meaning vision routes through a separate handoff pre-analysis step instead of accepting raw image blocks; `umansSupportsVision` treated any non-empty string as native vision support, so image prompts went directly to GLM and were rejected with `400 This model does not support image inputs`. The helper now requires `supports_vision === true`, and the bundled GLM 5.1/5.2 rows are corrected to text-only so the vision-handoff path runs. ([#3184](https://github.com/can1357/oh-my-pi/issues/3184))
+
 ## [16.1.9] - 2026-06-21
 
 ### Fixed

packages/catalog/src/models.json

@@ -69284,8 +69284,7 @@
 			"baseUrl": "https://api.code.umans.ai",
 			"reasoning": true,
 			"input": [
-				"text",
-				"image"
+				"text"
 			],
 			"cost": {
 				"input": 0,
@@ -69327,8 +69326,7 @@
 				]
 			},
 			"input": [
-				"text",
-				"image"
+				"text"
 			],
 			"cost": {
 				"input": 0,

packages/catalog/src/provider-models/openai-compat.ts

@@ -586,8 +586,18 @@ function normalizeUmansBaseUrl(baseUrl: string | undefined): string {
 	return normalized.endsWith("/v1") ? normalized.slice(0, -3) : normalized;
 }
 
+/**
+ * Umans `models/info` reports `supports_vision: true` for natively
+ * vision-capable models and a non-empty string sentinel (e.g.
+ * `"via-handoff"`) for models that route image inputs through a vision
+ * handoff pre-analysis step instead of accepting raw image blocks. Only
+ * `true` means the model accepts image content directly; sentinel values
+ * MUST map to text-only so the agent's vision-handoff path runs instead
+ * of triggering an upstream HTTP 400 (`This model does not support image
+ * inputs`).
+ */
 function umansSupportsVision(value: unknown): boolean {
-	return value === true || (typeof value === "string" && value.length > 0);
+	return value === true;
 }
 
 function umansReasoningSupported(value: unknown): boolean {

packages/catalog/test/umans-provider.test.ts

@@ -101,6 +101,56 @@ describe("umans provider catalog", () => {
 		await expect(fetchDynamicModels()).rejects.toThrow("Failed to fetch Umans models info");
 	});
 
+	it('maps supports_vision sentinel values like "via-handoff" to text-only input', async () => {
+		const fetchImpl: FetchImpl = async () =>
+			new Response(
+				JSON.stringify({
+					"umans-glm-5.2": {
+						display_name: "Umans GLM 5.2",
+						capabilities: {
+							context_window: 405_504,
+							max_completion_tokens: 131_071,
+							recommended_max_tokens: 131_071,
+							supports_vision: "via-handoff",
+							supports_tools: true,
+							reasoning: { supported: true, can_disable: true, default_level: "medium" },
+						},
+					},
+					"umans-coder": {
+						display_name: "Umans Coder",
+						capabilities: {
+							context_window: 262_144,
+							max_completion_tokens: 262_144,
+							recommended_max_tokens: 32_768,
+							supports_vision: true,
+							supports_tools: true,
+							reasoning: { supported: true, can_disable: true, default_level: "medium" },
+						},
+					},
+				}),
+				{ status: 200, headers: { "Content-Type": "application/json" } },
+			);
+
+		const fetchDynamicModels = umansModelManagerOptions({ fetch: fetchImpl }).fetchDynamicModels;
+		if (!fetchDynamicModels) throw new Error("Umans dynamic discovery is not configured");
+
+		const models = await fetchDynamicModels();
+		const glm = models?.find(item => item.id === "umans-glm-5.2");
+		const coder = models?.find(item => item.id === "umans-coder");
+
+		expect(glm?.input).toEqual(["text"]);
+		expect(coder?.input).toEqual(["text", "image"]);
+	});
+
+	it("bundles Umans GLM via-handoff models as text-only", () => {
+		const providers = modelsJson as Record<string, Record<string, BundledModel>>;
+		for (const id of ["umans-glm-5.1", "umans-glm-5.2"] as const) {
+			const model = providers.umans?.[id];
+			expect(model, `${id} should be bundled`).toBeDefined();
+			expect(model.input, `${id} input should be text-only`).toEqual(["text"]);
+		}
+	});
+
 	it("maps the models.dev Umans provider to the Anthropic endpoint", () => {
 		const models = mapModelsDevToModels(
 			{