Merge pull request #226 from ollama4j/fix-docs-build

Fix: setting tools in `generate()` method and issue #227

docs/blog/2025-03-08-blog/index.md

@@ -209,7 +209,6 @@ To download/pull the model into your Ollama server, run the following command in
 
 ```shell
 ollama pull mistral
-
 ```
 
 You can list the models available on your model server by running the following command in your terminal.

docs/docs/apis-generate/generate-with-tools.md

@@ -31,7 +31,19 @@ You could do that with ease with the `function calling` capabilities of the mode
 
 ### Create Tools/Functions
 
-We can create static functions as our tools.
+There are two ways to create and register your tools:
+
+1. **Define static or regular methods and register them explicitly as tools.**
+   You can create standalone functions (static or instance methods) and manually associate them with your tool specifications.
+
+2. **Use annotation-based tool discovery for automatic registration.**
+   By annotating your tool methods, you can leverage `registerAnnotatedTools()` to automatically scan your classpath, find all annotated tool functions, and register them without extra boilerplate.
+
+Learn more about annotation-based tool registration [here](/apis-generate/chat-with-tools#annotation-based-tool-registration).
+
+Choose the approach that best fits your project—manual for precise control, or annotation-based for easier scaling.
+
+Let's start by exploring the first approach: manually defining and registering your tools/functions.
 
 This function takes the arguments `location` and `fuelType` and performs an operation with these arguments and returns
 fuel price value.

src/main/java/io/github/ollama4j/Ollama.java

@@ -804,9 +804,21 @@ public class Ollama {
         ocm.setResponse(request.getPrompt());
         chatRequest.setMessages(msgs);
         msgs.add(ocm);
+
+        // Merge request's tools and globally registered tools into a new list to avoid mutating the
+        // original request
+        List<Tools.Tool> allTools = new ArrayList<>();
+        if (request.getTools() != null) {
+            allTools.addAll(request.getTools());
+        }
+        List<Tools.Tool> registeredTools = this.getRegisteredTools();
+        if (registeredTools != null) {
+            allTools.addAll(registeredTools);
+        }
+
         OllamaChatTokenHandler hdlr = null;
         chatRequest.setUseTools(true);
-        chatRequest.setTools(request.getTools());
+        chatRequest.setTools(allTools);
         if (streamObserver != null) {
             chatRequest.setStream(true);
             if (streamObserver.getResponseStreamHandler() != null) {

src/main/java/io/github/ollama4j/models/request/OllamaChatEndpointCaller.java

@@ -44,10 +44,13 @@ public class OllamaChatEndpointCaller extends OllamaEndpointCaller {
 
     /**
      * Parses streamed Response line from ollama chat. Using {@link
-     * com.fasterxml.jackson.databind.ObjectMapper#readValue(String, TypeReference)} should throw
+     * com.fasterxml.jackson.databind.ObjectMapper#readValue(String, TypeReference)}
+     * should throw
      * {@link IllegalArgumentException} in case of null line or {@link
-     * com.fasterxml.jackson.core.JsonParseException} in case the JSON Object cannot be parsed to a
-     * {@link OllamaChatResponseModel}. Thus, the ResponseModel should never be null.
+     * com.fasterxml.jackson.core.JsonParseException} in case the JSON Object cannot
+     * be parsed to a
+     * {@link OllamaChatResponseModel}. Thus, the ResponseModel should never be
+     * null.
      *
      * @param line           streamed line of ollama stream response
      * @param responseBuffer Stringbuffer to add latest response message part to
@@ -59,9 +62,11 @@ public class OllamaChatEndpointCaller extends OllamaEndpointCaller {
         try {
             OllamaChatResponseModel ollamaResponseModel =
                     Utils.getObjectMapper().readValue(line, OllamaChatResponseModel.class);
-            // It seems that under heavy load Ollama responds with an empty chat message part in the
+            // It seems that under heavy load Ollama responds with an empty chat message
+            // part in the
             // streamed response.
-            // Thus, we null check the message and hope that the next streamed response has some
+            // Thus, we null check the message and hope that the next streamed response has
+            // some
             // message content again.
             OllamaChatMessage message = ollamaResponseModel.getMessage();
             if (message != null) {
@@ -118,7 +123,9 @@ public class OllamaChatEndpointCaller extends OllamaEndpointCaller {
                         parseResponseAndAddToBuffer(line, responseBuffer, thinkingBuffer);
                 ollamaChatResponseModel =
                         Utils.getObjectMapper().readValue(line, OllamaChatResponseModel.class);
-                if (body.stream && ollamaChatResponseModel.getMessage().getToolCalls() != null) {
+                if (body.stream
+                        && ollamaChatResponseModel.getMessage() != null
+                        && ollamaChatResponseModel.getMessage().getToolCalls() != null) {
                     wantedToolsForStream = ollamaChatResponseModel.getMessage().getToolCalls();
                 }
                 if (finished && body.stream) {
@@ -153,7 +160,8 @@ public class OllamaChatEndpointCaller extends OllamaEndpointCaller {
     }
 
     /**
-     * Handles error status codes and appends error messages to the response buffer. Returns true if
+     * Handles error status codes and appends error messages to the response buffer.
+     * Returns true if
      * an error was handled, false otherwise.
      */
     private boolean handleErrorStatus(int statusCode, String line, StringBuilder responseBuffer)