Merge pull request #4112 from janhq/enhancement/max-token-limit-claude

enhancement: increased max-token limit claude-3.5

extensions/inference-anthropic-extension/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@janhq/inference-anthropic-extension",
   "productName": "Anthropic Inference Engine",
-  "version": "1.0.2",
+  "version": "1.0.3",
   "description": "This extension enables Anthropic chat completion API calls",
   "main": "dist/index.js",
   "module": "dist/module.js",

extensions/inference-anthropic-extension/resources/models.json

@@ -85,7 +85,7 @@
     "format": "api",
     "settings": {},
     "parameters": {
-      "max_tokens": 4096,
+      "max_tokens": 8192,
       "temperature": 0.7,
       "stream": true
     },