Antripic - Handling the second type of token count exceeded message - "input length and max_tokens exceed context limit..."

Author: peterbanda

anthropic-client/src/main/scala/io/cequence/openaiscala/anthropic/service/HandleAnthropicErrorCodes.scala

@@ -10,6 +10,11 @@ import io.cequence.wsclient.service.WSClient
  */
 trait HandleAnthropicErrorCodes extends WSClient {
 
+  private val TokenCountExceededMessages = Set(
+    "input length and `max_tokens` exceed context limit",
+    "prompt is too long"
+  )
+
   override protected def handleErrorCodes(
     httpCode: Int,
     message: String
@@ -19,7 +24,7 @@ trait HandleAnthropicErrorCodes extends WSClient {
 
       case 400 => {
         // Check if the error message indicates token count exceeded
-        if (message.toLowerCase.contains("prompt is too long") && message.toLowerCase.contains("tokens")) {
+        if (TokenCountExceededMessages.exists(message.contains)) {
           throw new AnthropicScalaTokenCountExceededException(errorMessage)
         } else {
           // 400 - invalid_request_error: There was an issue with the format or content of your request.

openai-count-tokens/src/main/scala/io/cequence/openaiscala/service/OpenAICountTokensHelper.scala

@@ -166,7 +166,7 @@ trait OpenAICountTokensHelper {
    * @param modelType
    * @return
    */
-  protected def countTokens(
+  def countTokens(
     text: String,
     modelType: Option[ModelType] = None
   ) = {

openai-examples/src/main/scala/io/cequence/openaiscala/examples/anthropic/AnthropicCreateChatCompletionWithOpenAIAdapterTokenCountExceeded2.scala

@@ -0,0 +1,34 @@
+package io.cequence.openaiscala.examples.anthropic
+
+import io.cequence.openaiscala.domain.settings.CreateChatCompletionSettings
+import io.cequence.openaiscala.domain.{NonOpenAIModelId, SystemMessage, UserMessage}
+import io.cequence.openaiscala.examples.{ChatCompletionProvider, ExampleBase}
+import io.cequence.openaiscala.service.OpenAIChatCompletionService
+
+import scala.concurrent.Future
+
+// requires `openai-scala-anthropic-client` as a dependency and `ANTHROPIC_API_KEY` environment variable to be set
+object AnthropicCreateChatCompletionWithOpenAIAdapterTokenCountExceeded2
+    extends ExampleBase[OpenAIChatCompletionService] {
+
+  override val service: OpenAIChatCompletionService = ChatCompletionProvider.anthropic()
+
+  private val messages = Seq(
+    SystemMessage("You are a helpful assistant."),
+    UserMessage("What is the weather like in Norway?" * 20000)
+  )
+
+  override protected def run: Future[_] =
+    service
+      .createChatCompletion(
+        messages = messages,
+        settings = CreateChatCompletionSettings(
+          NonOpenAIModelId.claude_3_7_sonnet_20250219,
+          // the second type of "token exceeded" error - input + max_tokens > limit (200000)
+          max_tokens = Some(100000)
+        )
+      )
+      .map { content =>
+        println(content.contentHead)
+      }
+}