feat(llm): auto-detect max images per request for LLM providers

Add binary-search probe endpoint POST /providers/{id}/probe-max-images
that discovers the per-request image limit by sending incrementally
more 1x1 JPEGs via the provider's own API. Result is cached in the
new max_images column on the provider config.

- LLMProviderConfig: add max_images: int | None
- LLMProvider (SQLAlchemy): add max_images column
- Migration ed28d34edde7: clean ADD COLUMN
- LLMProviderService: create/update/set_max_images
- POST /providers/{id}/probe-max-images: binary search + error parsing
- ProviderConfig.svelte: 'Detect' button in edit modal + HelpTooltip
- i18n (en/ru): 11 new keys for probe UI

frontend/src/components/llm/ProviderConfig.svelte

@@ -13,6 +13,7 @@
   import { t } from "../../lib/i18n";
   import { requestApi } from "../../lib/api";
   import { addToast } from "../../lib/toasts.js";
+  import HelpTooltip from "$lib/ui/HelpTooltip.svelte";
 
   let {
     providers = [],
@@ -37,6 +38,7 @@
     default_model: "gpt-4o",
     is_active: true,
     is_multimodal: false,
+    max_images: null,
   });
 
   let testStatus = $state({ type: "", message: "" });
@@ -51,6 +53,11 @@
 
   let fetchDebounceTimer = $state(null);
 
+  // Probing state
+  let isProbing = $state(false);
+  let probeResult = $state(null); // { max_images, method } or null
+  let probeError = $state("");
+
   function isMaskedKey(key) {
     if (!key) return false;
     return key === "********" || key.includes("...");
@@ -75,11 +82,14 @@
       default_model: "gpt-4o",
       is_active: true,
       is_multimodal: false,
+      max_images: null,
     };
     editingProvider = null;
     testStatus = { type: "", message: "" };
     availableModels = [];
     modelsLoadError = "";
+    probeResult = null;
+    probeError = "";
   }
 
   function handleEdit(provider) {
@@ -97,10 +107,13 @@
       default_model: provider?.default_model ?? "gpt-4o",
       is_active: Boolean(provider?.is_active),
       is_multimodal: Boolean(provider?.is_multimodal),
+      max_images: provider?.max_images ?? null,
     };
     testStatus = { type: "", message: "" };
     availableModels = [];
     modelsLoadError = "";
+    probeResult = null;
+    probeError = "";
     showForm = true;
     // Auto-fetch models when editing an existing provider
     if (provider?.base_url) {
@@ -205,6 +218,44 @@
     }
   }
 
+  async function detectMaxImages() {
+    const providerId = editingProvider?.id;
+    if (!providerId) {
+      addToast($t.llm.save_before_probe, "warning");
+      return;
+    }
+
+    console.log("[ProviderConfig][Action] Detecting max images for", providerId);
+    isProbing = true;
+    probeError = "";
+    probeResult = null;
+
+    try {
+      const result = await requestApi(`/llm/providers/${providerId}/probe-max-images`, "POST");
+      console.log("[ProviderConfig] probe result:", JSON.stringify(result));
+      probeResult = result;
+
+      if (result.max_images != null) {
+        addToast(
+          $t.llm.max_images_detected.replace("{n}", result.max_images),
+          "success"
+        );
+        formData.max_images = result.max_images;
+      } else {
+        addToast(
+          $t.llm.no_max_images_limit,
+          "info"
+        );
+      }
+    } catch (err) {
+      probeError = err.message || "Failed to detect max images";
+      console.warn("[ProviderConfig] Probe failed:", err.message);
+      addToast(probeError, "error");
+    } finally {
+      isProbing = false;
+    }
+  }
+
   async function handleSubmit() {
     console.log("[ProviderConfig][Action] Submitting provider config");
     const method = editingProvider ? "PUT" : "POST";
@@ -480,6 +531,48 @@
               {$t.llm.multimodal_checkbox}
             </label>
           </div>
+
+          <div class="border-t pt-3">
+            <div class="flex items-center justify-between">
+              <div class="flex items-center gap-1">
+                <span class="text-sm font-medium text-gray-700">
+                  {$t.llm.max_images_label}
+                </span>
+                <HelpTooltip text={$t.llm.help_max_images} />
+                {#if formData.max_images != null}
+                  <span class="ml-1 text-sm font-mono text-gray-600 bg-gray-100 px-2 py-0.5 rounded">
+                    {formData.max_images}
+                  </span>
+                {:else}
+                  <span class="ml-1 text-xs text-gray-400">
+                    {$t.llm.not_detected}
+                  </span>
+                {/if}
+              </div>
+              <button
+                type="button"
+                class="text-xs px-3 py-1.5 border rounded-md hover:bg-gray-50 disabled:opacity-50"
+                onclick={detectMaxImages}
+                disabled={isProbing || !editingProvider}
+                title={editingProvider ? $t.llm.detect_max_images : $t.llm.save_before_probe}
+              >
+                {#if isProbing}
+                  <span class="inline-block w-3 h-3 border-2 border-gray-400 border-t-transparent rounded-full animate-spin mr-1"></span>
+                  {$t.llm.detecting}
+                {:else}
+                  {$t.llm.detect_max_images}
+                {/if}
+              </button>
+            </div>
+            {#if probeError}
+              <div class="mt-1 text-xs text-red-600">{probeError}</div>
+            {/if}
+            {#if probeResult && probeResult.max_images != null}
+              <div class="mt-1 text-xs text-green-600">
+                {$t.llm.detected_via.replace("{method}", probeResult.method === "parse_error" ? $t.llm.detection_method_error : $t.llm.detection_method_binary)}
+              </div>
+            {/if}
+          </div>
         </div>
 
         {#if testStatus.message}

frontend/src/lib/i18n/locales/en/llm.json

@@ -33,5 +33,16 @@
   "models_load_failed": "Failed to load models: {error}",
   "select_model": "Select a model...",
   "refresh_models": "Refresh models",
-  "multimodal_checkbox": "Multimodal (supports images)"
+  "multimodal_checkbox": "Multimodal (supports images)",
+  "max_images_label": "Max images per request",
+  "not_detected": "Not detected",
+  "detect_max_images": "Detect",
+  "detecting": "Detecting...",
+  "detected_via": "Detected via: {method}",
+  "detection_method_error": "error parsing",
+  "detection_method_binary": "binary search",
+  "save_before_probe": "Save the provider first to detect max images",
+  "max_images_detected": "Max images: {n}",
+  "no_max_images_limit": "No image limit detected",
+  "help_max_images": "Auto-detect the maximum number of images the LLM provider accepts per request. Some providers (e.g. Nvidia NeMo) limit to 8-10 images."
 }

frontend/src/lib/i18n/locales/ru/llm.json

@@ -33,5 +33,16 @@
   "models_load_failed": "Не удалось загрузить модели: {error}",
   "select_model": "Выберите модель...",
   "refresh_models": "Обновить модели",
-  "multimodal_checkbox": "Мультимодальная (поддержка изображений)"
+  "multimodal_checkbox": "Мультимодальная (поддержка изображений)",
+  "max_images_label": "Макс. изображений за запрос",
+  "not_detected": "Не определен",
+  "detect_max_images": "Определить",
+  "detecting": "Определение...",
+  "detected_via": "Определено через: {method}",
+  "detection_method_error": "парсинг ошибки",
+  "detection_method_binary": "бинарный поиск",
+  "save_before_probe": "Сначала сохраните провайдера",
+  "max_images_detected": "Макс. изображений: {n}",
+  "no_max_images_limit": "Лимит изображений не обнаружен",
+  "help_max_images": "Автоматическое определение максимального количества изображений, которое LLM-провайдер принимает за один запрос. Некоторые провайдеры (например Nvidia NeMo) ограничивают до 8-10 изображений."
 }