Revert "Request less `max_new_tokens` on tools and reduce `truncate` on CMD R+ (#1292)"

chart/env/prod.yaml

@@ -47,7 +47,7 @@ envVars:
         "tools": true,
         "parameters": {
           "stop": ["<|END_OF_TURN_TOKEN|>"],
-          "truncate" : 12288,
+          "truncate" : 28672,
           "max_new_tokens" : 2048,
           "temperature" : 0.3
         },

src/lib/server/textGeneration/tools.ts

@@ -131,7 +131,7 @@ export async function* runTools(
 	for await (const output of await endpoint({
 		messages: messagesWithFilesPrompt,
 		preprompt,
-		generateSettings: { ...assistant?.generateSettings, max_new_tokens: 512 },
+		generateSettings: assistant?.generateSettings,
 		tools,
 	})) {
 		// model natively supports tool calls