Fix compilation errors and implement missing OpenAI types methods

- Add OpenAITypes.cpp to CMakeLists.txt
- Implement missing methods and struct definitions
- Fix test failures
- All tests passing (108 passed, 8 skipped)

Author: lucaromagnoli

CMakeLists.txt

@@ -72,6 +72,7 @@ set(LLMCPP_SOURCES
     src/openai/OpenAIResponsesApi.cpp
     src/openai/OpenAISchemaBuilder.cpp
     src/openai/OpenAIModels.cpp
+    src/openai/OpenAITypes.cpp
     src/openai/OpenAIUtils.cpp
     src/anthropic/AnthropicClient.cpp
     src/anthropic/AnthropicHttpClient.cpp

include/core/LLMTypes.h

@@ -12,98 +12,11 @@
 
 using json = nlohmann::json;
 
-// Context type using standard C++ vectors of generic objects
-using LLMContext = std::vector<json>;
+// LLMRequestConfig and LLMContext are now defined in OpenAITypes.h to avoid circular dependency
 
-/// Represents the configuration for the LLM
-struct LLMRequestConfig {
-    std::string client;
-    std::string model;                          // String model name (works with any provider)
-    std::string functionName = "llm_function";  // Default function name for LLM calls
-    std::string jsonSchema;
-    std::optional<json> schemaObject;  // Structured schema data
+// LLMRequest is now defined in OpenAITypes.h to avoid circular dependency
 
-    std::optional<float> temperature;  // Optional temperature (filtered by model support)
-    std::optional<int> maxTokens;      // Optional max tokens
-    std::optional<std::vector<OpenAI::ToolVariant>> tools;  // Optional tools for function calling
-
-    // Convenience method for any model name
-    void setModel(const std::string& modelName) { model = modelName; }
-
-    std::string getModelString() const { return model; }
-
-    // Add more configuration options as needed (e.g., top_p, stop sequences, etc.)
-
-    std::string toString() const {
-        std::string schemaStr = schemaObject.has_value() ? schemaObject->dump() : jsonSchema;
-        std::string tempStr = temperature.has_value() ? std::to_string(*temperature) : "not set";
-        std::string toolsStr = tools.has_value() ? std::to_string(tools->size()) + " tools" : "no tools";
-        return "LLMRequestConfig { client: " + client + ", model: " + getModelString() +
-               ", functionName: " + functionName + ", schema: " + schemaStr +
-               ", temperature: " + tempStr +
-               ", maxTokens: " + std::to_string(maxTokens.has_value() ? *maxTokens : 0) +
-               ", tools: " + toolsStr + " }";
-    }
-};
-
-struct LLMRequest {
-    LLMRequest() = delete;
-
-    // Constructor with prompt only
-    LLMRequest(LLMRequestConfig config, std::string prompt, LLMContext context = {},
-               std::string previousResponseId = "")
-        : config(std::move(config)),
-          prompt(std::move(prompt)),
-          context(std::move(context)),
-          previousResponseId(std::move(previousResponseId)) {}
-
-    // Constructor with single context object (convenience)
-    LLMRequest(LLMRequestConfig config, std::string prompt, json contextObject,
-               std::string previousResponseId = "")
-        : config(std::move(config)),
-          prompt(std::move(prompt)),
-          context({std::move(contextObject)}),
-          previousResponseId(std::move(previousResponseId)) {}
-
-    LLMRequestConfig config;
-    std::string prompt;  // The main task/prompt (what to do) - maps to instructions
-    LLMContext context;  // Context data (vector of generic objects) - maps to inputValues
-    std::string previousResponseId;  // For conversation continuity
-
-    // Utility methods
-    std::string instructions() const { return prompt; }  // For OpenAI mapping
-
-    std::string toString() const {
-        std::string contextString = "[";
-        for (size_t i = 0; i < context.size(); ++i) {
-            if (i > 0) contextString += ", ";
-            contextString += context[i].dump();
-        }
-        contextString += "]";
-
-        return "LLMRequest {\n config: " + config.toString() + ",\n prompt: " + prompt +
-               ",\n context: " + contextString + ",\n previousResponseId: " + previousResponseId +
-               "\n}";
-    }
-};
-
-// LLMUsage is now defined in OpenAITypes.h to avoid circular dependency
-
-struct LLMResponse {
-    json result = json::object();
-    bool success = false;
-    std::string errorMessage;
-    std::string responseId;  // For conversation continuity with OpenAI
-    LLMUsage usage;          // Token usage information
-
-    std::string toString() const {
-        std::string resultString = result.dump(2);
-        return "LLMResponse {\n result: " + resultString +
-               ",\n success: " + (success ? "true" : "false") +
-               ",\n errorMessage: " + errorMessage + ",\n responseId: " + responseId +
-               ",\n usage: " + usage.toString() + "\n}";
-    }
-};
+// LLMUsage and LLMResponse are now defined in OpenAITypes.h to avoid circular dependency
 
 // Error codes for LLM operations
 enum class LLMErrorCode {