AnthropicChatModel support InternalToolExecutionMaxAttempts

Signed-off-by: lambochen <[email protected]>

Author: lambochen

models/spring-ai-anthropic/src/main/java/org/springframework/ai/anthropic/AnthropicChatModel.java

@@ -90,6 +90,7 @@
  * @author Alexandros Pappas
  * @author Jonghoon Park
  * @author Soby Chacko
+ * @author lambochen
  * @since 1.0.0
  */
 public class AnthropicChatModel implements ChatModel {
@@ -170,10 +171,10 @@ public ChatResponse call(Prompt prompt) {
 		// Before moving any further, build the final request Prompt,
 		// merging runtime and default options.
 		Prompt requestPrompt = buildRequestPrompt(prompt);
-		return this.internalCall(requestPrompt, null);
+		return this.internalCall(requestPrompt, null, 1);
 	}
 
-	public ChatResponse internalCall(Prompt prompt, ChatResponse previousChatResponse) {
+	public ChatResponse internalCall(Prompt prompt, ChatResponse previousChatResponse, int attempts) {
 		ChatCompletionRequest request = createRequest(prompt, false);
 
 		ChatModelObservationContext observationContext = ChatModelObservationContext.builder()
@@ -203,7 +204,7 @@ public ChatResponse internalCall(Prompt prompt, ChatResponse previousChatRespons
 				return chatResponse;
 			});
 
-		if (this.toolExecutionEligibilityPredicate.isToolExecutionRequired(prompt.getOptions(), response)) {
+		if (this.toolExecutionEligibilityPredicate.isToolExecutionRequired(prompt.getOptions(), response, attempts)) {
 			var toolExecutionResult = this.toolCallingManager.executeToolCalls(prompt, response);
 			if (toolExecutionResult.returnDirect()) {
 				// Return tool execution result directly to the client.
@@ -215,7 +216,7 @@ public ChatResponse internalCall(Prompt prompt, ChatResponse previousChatRespons
 			else {
 				// Send the tool execution result back to the model.
 				return this.internalCall(new Prompt(toolExecutionResult.conversationHistory(), prompt.getOptions()),
-						response);
+						response, attempts + 1);
 			}
 		}
 
@@ -232,10 +233,10 @@ public Flux<ChatResponse> stream(Prompt prompt) {
 		// Before moving any further, build the final request Prompt,
 		// merging runtime and default options.
 		Prompt requestPrompt = buildRequestPrompt(prompt);
-		return this.internalStream(requestPrompt, null);
+		return this.internalStream(requestPrompt, null, 1);
 	}
 
-	public Flux<ChatResponse> internalStream(Prompt prompt, ChatResponse previousChatResponse) {
+	public Flux<ChatResponse> internalStream(Prompt prompt, ChatResponse previousChatResponse, int attempts) {
 		return Flux.deferContextual(contextView -> {
 			ChatCompletionRequest request = createRequest(prompt, true);
 
@@ -260,7 +261,8 @@ public Flux<ChatResponse> internalStream(Prompt prompt, ChatResponse previousCha
 				Usage accumulatedUsage = UsageCalculator.getCumulativeUsage(currentChatResponseUsage, previousChatResponse);
 				ChatResponse chatResponse = toChatResponse(chatCompletionResponse, accumulatedUsage);
 
-				if (this.toolExecutionEligibilityPredicate.isToolExecutionRequired(prompt.getOptions(), chatResponse) && chatResponse.hasFinishReasons(Set.of("tool_use"))) {
+				if (this.toolExecutionEligibilityPredicate.isToolExecutionRequired(prompt.getOptions(), chatResponse, attempts)
+						&& chatResponse.hasFinishReasons(Set.of("tool_use"))) {
 					// FIXME: bounded elastic needs to be used since tool calling
 					//  is currently only synchronous
 					return Flux.defer(() -> {
@@ -274,7 +276,7 @@ public Flux<ChatResponse> internalStream(Prompt prompt, ChatResponse previousCha
 						else {
 							// Send the tool execution result back to the model.
 							return this.internalStream(new Prompt(toolExecutionResult.conversationHistory(), prompt.getOptions()),
-									chatResponse);
+									chatResponse, attempts + 1);
 						}
 					}).subscribeOn(Schedulers.boundedElastic());
 				}
@@ -437,6 +439,11 @@ Prompt buildRequestPrompt(Prompt prompt) {
 			requestOptions.setInternalToolExecutionEnabled(
 					ModelOptionsUtils.mergeOption(runtimeOptions.getInternalToolExecutionEnabled(),
 							this.defaultOptions.getInternalToolExecutionEnabled()));
+			requestOptions.setInternalToolExecutionMaxAttempts(
+					ModelOptionsUtils.mergeOption(
+							runtimeOptions.getInternalToolExecutionMaxAttempts(),
+							defaultOptions.getInternalToolExecutionMaxAttempts())
+			);
 			requestOptions.setToolNames(ToolCallingChatOptions.mergeToolNames(runtimeOptions.getToolNames(),
 					this.defaultOptions.getToolNames()));
 			requestOptions.setToolCallbacks(ToolCallingChatOptions.mergeToolCallbacks(runtimeOptions.getToolCallbacks(),
@@ -447,6 +454,7 @@ Prompt buildRequestPrompt(Prompt prompt) {
 		else {
 			requestOptions.setHttpHeaders(this.defaultOptions.getHttpHeaders());
 			requestOptions.setInternalToolExecutionEnabled(this.defaultOptions.getInternalToolExecutionEnabled());
+			requestOptions.setInternalToolExecutionMaxAttempts(this.defaultOptions.getInternalToolExecutionMaxAttempts());
 			requestOptions.setToolNames(this.defaultOptions.getToolNames());
 			requestOptions.setToolCallbacks(this.defaultOptions.getToolCallbacks());
 			requestOptions.setToolContext(this.defaultOptions.getToolContext());