Update models.json

models.json

@@ -1,80 +1,64 @@
 {
-  "providers": {
-    "zuccaro": {
-      "baseUrl": "https://ai.zuccaro.me/api",
-      "api": "openai-completions",
-      "models": [
-        {
-          "id": "DeepSeek-V4-Flash-0731",
+  "provider": {
+    "zuccaro-oui": {
+      "npm": "@ai-sdk/openai-compatible",
+      "name": "Zuccaro AI (OUI)",
+      "options": {
+        "baseURL": "https://ai.zuccaro.me/api",
+        "apiKey": "sk-<YOUR-OUI-API-KEY>"
+      },
+      "models": {
+        "bifrost.DeepSeek-V4-Flash-0731": {
           "name": "DeepSeek V4 Flash 0731",
           "reasoning": true,
-          "input": ["text"],
-          "contextWindow": 524288,
-          "maxTokens": 384000,
-          "cost": {
-            "input": 0.44,
-            "output": 1.32,
-            "cacheRead": 0.0028,
-            "cacheWrite": 0
+          "interleaved": { "field": "reasoning_content" },
+          "limit": { "context": 786432, "output": 384000 },
+          "cost": { "input": 0.44, "output": 1.32, "cache_read": 0.0028, "cache_write": 0 },
+          "modalities": { "input": ["text"], "output": ["text"] }
         },
-          "thinkingLevelMap": {
-            "max": "max"
-          }
-        },
-        {
-          "id": "DeepSeek-V4-Flash-Vision-Exp",
+        "bifrost.DeepSeek-V4-Flash-Vision-Exp": {
           "name": "DeepSeek V4 Flash Vision Exp",
           "reasoning": true,
-          "input": ["text", "image"],
-          "contextWindow": 524288,
-          "maxTokens": 384000,
-          "thinkingLevelMap": {
-            "max": "max"
-          }
+          "interleaved": { "field": "reasoning_content" },
+          "limit": { "context": 524288, "output": 384000 },
+          "modalities": { "input": ["text", "image"], "output": ["text"] }
         },
-        {
-          "id": "Qwen3.8-27B-NVFP4",
+        "bifrost.Qwen3.8-27B-NVFP4": {
           "name": "Qwen3.8 27B NVFP4",
           "reasoning": true,
-          "input": ["text"],
-          "contextWindow": 217088,
-          "maxTokens": 32768,
-          "thinkingLevelMap": {
-            "xhigh": "xhigh"
+          "interleaved": { "field": "reasoning_content" },
+          "limit": { "context": 217088, "output": 32768 },
+          "modalities": { "input": ["text"], "output": ["text"] },
+          "variants": {
+            "xhigh": { "reasoningEffort": "xhigh" },
+            "medium": { "reasoningEffort": "medium" },
+            "low": { "reasoningEffort": "low" },
+            "no-thinking": { "reasoningEffort": "none" }
           }
         },
-        {
-          "id": "GLM-5.3-Flash-SGLang",
+        "bifrost.GLM-5.3-Flash-SGLang": {
           "name": "GLM 5.3 Flash SGLang",
           "reasoning": true,
-          "input": ["text", "image"],
-          "contextWindow": 524288,
-          "maxTokens": 131072,
-          "cost": {
-            "input": 0.15,
-            "output": 0.5,
-            "cacheRead": 0.03,
-            "cacheWrite": 0
+          "interleaved": { "field": "reasoning_content" },
+          "limit": { "context": 524288, "output": 131072 },
+          "cost": { "input": 0.15, "output": 0.5, "cache_read": 0.03, "cache_write": 0 },
+          "modalities": { "input": ["text", "image"], "output": ["text"] }
         },
-          "thinkingLevelMap": {
-            "max": "max"
-          }
-        },
-        {
-          "id": "Blackwell-Auto",
+        "bifrost.Blackwell-Auto": {
           "name": "Blackwell Auto",
           "reasoning": true,
-          "input": ["text"]
+          "interleaved": { "field": "reasoning_content" },
+          "limit": { "context": 524288, "output": 131072 },
+          "cost": { "input": 0.15, "output": 0.5, "cache_read": 0.03, "cache_write": 0 },
+          "modalities": { "input": ["text", "image"], "output": ["text"] }
         },
-        {
-          "id": "Qwen3-4B-Instruct-2507-GGUF-Q4_K_XL",
+        "Qwen3-4B-Instruct-2507-GGUF:Q4_K_XL": {
           "name": "Qwen3 4B",
           "reasoning": false,
-          "input": ["text"],
-          "contextWindow": 32768,
-          "maxTokens": 32768
+          "limit": { "context": 32768, "output": 32768 },
+          "modalities": { "input": ["text"], "output": ["text"] }
         }
-      ]
+      }
     }
   }
 }