Revise User Guide to reflect new tool calling update

Author: AvantMaker

README.md

@@ -147,8 +147,8 @@ void loop() {
 }
 ```
 
-## Tool Calls Support
-Tool calls (function calling) enable the AI model to request specific actions from your ESP32 application. This feature allows:
+## Tool Calls (Tool Calling) Support
+Tool calls (a.k.a. talling calling) enable the AI model to request specific actions from your ESP32 application. This feature allows:
 
 - Two-way interaction : AI can request data or actions from your device
 - Structured data exchange : Receive properly formatted JSON for easy parsing

doc/User Guides/3 Tool Calls Implementation Basics.md

@@ -1,8 +1,8 @@
 # ESP32_AI_Connect Library User Guide - 3 Tool Calls Implementation Basics
-> **Version 0.0.2** • Revised: May 09, 2025 • Author: AvantMaker • [https://www.AvantMaker.com](https://www.AvantMaker.com)
+> **Version 0.0.5** • Revised: May 12, 2025 • Author: AvantMaker • [https://www.AvantMaker.com](https://www.AvantMaker.com)
 ## Overview
 
-This guide will walk you through the process of setting up and using tool calls (function calling) with Large Language Models (LLMs) using the ESP32_AI_Connect library. We'll use the `tool_calls_demo.ino` sketch stored in the examples folder as our reference implementation, explaining each component in detail so you can understand how to integrate AI function calling capabilities into your ESP32 projects.
+This guide will walk you through the process of setting up and using tool calls (tool calling) with Large Language Models (LLMs) using the ESP32_AI_Connect library. We'll use the `tool_calling_demo.ino` sketch stored in the examples folder as our reference implementation, explaining each component in detail so you can understand how to integrate AI tool calling capabilities into your ESP32 projects with ESP32_AI_Connect.
 
 ## Prerequisites
 
@@ -12,7 +12,7 @@ Before you begin, make sure you have:
 - Arduino IDE installed with ESP32 board support
 - ESP32_AI_Connect library installed
 - WiFi connectivity
-- An API key for your chosen AI platform (OpenAI is recommended for tool calls)
+- An API key for your chosen AI platform (Ensure the AI Model you choose supports Tool Calls functionality)
 - Basic understanding of JSON and Arduino programming
 
 ## Step 1: Enable Tool Calls in Configuration
@@ -21,7 +21,7 @@ First, ensure that tool calls support is enabled in the library configuration fi
 
 ```cpp
 // --- Tool Calls Support ---
-// Uncomment the following line to enable tool calls (function calling) support
+// Uncomment the following line to enable tool calls (tool calling) support
 // This will add tcToolSetup and tcChat methods to the library
 // If you don't need tool calls, keep this commented out to save memory
 #define ENABLE_TOOL_CALLS
@@ -60,7 +60,7 @@ ESP32_AI_Connect aiClient(platform, apiKey, model);
 ```
 
 This line initializes the AI client with parameters:
-- The platform identifier (typically `"openai"` for tool calls)
+- The platform identifier ("openai", "gemini", "claude" or "deepseek", etc.)
 - Your API key
 - The model name (e.g., `"gpt-3.5-turbo"` or `"gpt-4"`)
 - Optional custom endpoint (if you're using a custom API endpoint)
@@ -93,7 +93,7 @@ void setup() {
 
 ## Step 5: Define Your Tools
 
-Tool calls require defining the functions that the AI can call. Each tool is defined as a JSON object that specifies the function name, description, and parameters:
+Tool calls require defining the tool(s) that the AI can call. Each tool is defined as a JSON object that specifies the function name, description, and parameters:
 
 ```cpp
 // --- Define Tools for Tool Calling ---
@@ -158,7 +158,7 @@ These optional methods allow you to:
 
 Each setter also has a corresponding getter method to retrieve the current value.
 
-Different AI platforms support different `toolChoice` parameters. The table below shows the allowed values for each platform:
+Different AI platforms support different `tool_choice` parameters. The table below shows the allowed values for each platform:
 
 | Tool Choice Mode | OpenAI API | Gemini API | Anthropic Claude API |
 |------------------|------------|------------|----------------------|
@@ -171,13 +171,13 @@ Different AI platforms support different `toolChoice` parameters. The table belo
 Notes on tool choice options:
 - `"auto"`: Let the AI decide whether to use tools (default behavior)
 - `"none"`: Don't use tools
-- `"required"`: (OpenAI only) Forces the model to use a tool
-- `"any"`: (Gemini and Claude only) Model should use a tool if relevant
+- `"required"`: (OpenAI and OpenAI Compatible) Forces the model to use a tool
+- `"any"`: (Gemini and Claude) Model should use a tool if relevant
 - Tool Choice JSON: Specifies a particular tool to use (format varies by platform)
 
 ## Step 8: Send a Message to Trigger Tool Calls
 
-Now we can send a message that will likely trigger a tool call:
+Now we can send a message that may trigger a tool call based on user prompt content:
 
 ```cpp
 // --- Perform Tool Calling Chat ---

doc/User Guides/4 Tool Calls Follow-Up Techniques.md

@@ -1,5 +1,5 @@
 # ESP32_AI_Connect Library User Guide - 4 Tool Calls Follow-Up Techniques
-> **Version 0.0.2** • Revised: May 09, 2025 • Author: AvantMaker • [https://www.AvantMaker.com](https://www.AvantMaker.com)
+> **Version 0.0.3** • Revised: May 12, 2025 • Author: AvantMaker • [https://www.AvantMaker.com](https://www.AvantMaker.com)
 ## Introduction
 
 This article is a follow-up to the previous guide "Tool Calls Implementation Basics". If you haven't read that article yet, please do so before continuing, as this guide builds upon the concepts introduced there.
@@ -94,12 +94,6 @@ String getWeatherData(const String& city, const String& units) {
 }
 ```
 
-The tool results must be formatted as a JSON array of objects, where each object contains:
-- `tool_call_id`: The ID of the tool call (provided by the AI in the original tool call)
-- `function`: An object containing:
-  - `name`: The name of the function that was called
-  - `output`: The result of the function execution (as a string)
-
 ## Step 2: Parsing Tool Calls and Executing Functions
 
 When you receive a response with tool calls, you need to parse the JSON and execute the appropriate functions:
@@ -167,15 +161,22 @@ String toolResultsJson;
 serializeJson(toolResults, toolResultsJson);
 ```
 
+
+The tool results must be formatted as a JSON array of objects, where each object contains:
+- `tool_call_id`: The ID of the tool call (provided by the AI in the original tool call)
+- `function`: An object containing:
+  - `name`: The name of the function that was called
+  - `output`: The result of the function execution (as a string)
+
 ## Step 3: Sending the Tool Results Back to the AI
 
 Before sending the results back to the AI, you can configure the follow-up request with separate parameters:
 
 ```cpp
-// Additional follow-up configuration
+// Additional follow-up optional configuration
 // These only affect the follow-up request, not the initial tool call
-aiClient.setTCReplyMaxTokens(350);   // Maximum tokens for the follow-up response
-aiClient.setTCReplyToolChoice("auto"); // Tool choice for the follow-up (can be different from initial)
+aiClient.setTCReplyMaxTokens(350);   // (Optional) Maximum tokens for the follow-up response
+aiClient.setTCReplyToolChoice("auto"); // (Optional) Tool choice for the follow-up (can be different from initial)
 
 // You can also use a JSON string to specify a particular tool
 // aiClient.setTCReplyToolChoice(R"({"type": "function","function": {"name": "control_device"}})");
@@ -226,11 +227,22 @@ else {
 }
 ```
 
-Note that different AI platforms use different terminology for finish reasons:
+Note that different AI platforms use different terminology for folow-up tool calling finish reasons:
 - OpenAI and Gemini use "tool_calls"
-- Anthropic Claude uses "tool_use"
-- OpenAI and Gemini use "stop" for completed responses
-- Anthropic Claude uses "end_turn" for completed responses
+  
+  This indicates the AI is requesting for more tool calling to complete user's requests.
+
+- OpenAI and Gemini use "stop" 
+
+  This indicates the AI has completed  user's follow-up tool calling requests.
+
+- Anthropic Claude uses "tool_use" 
+
+  This indicates the AI is requesting for more tool calling to complete user's requests.
+
+- Anthropic Claude uses "end_turn" 
+
+  This indicates the AI has completed  user's follow-up tool calling requests.
 
 ## Resetting Tool Call Configuration
 
@@ -266,7 +278,7 @@ Let's walk through the complete flow of a tool call interaction using the exampl
      while(1) { delay(1000); } // Halt on failure
    }
 
-   // Configure tool calling parameters
+   // Configure optional tool calling parameters
    aiClient.setTCChatSystemRole("You are a smart home assistant.");
    aiClient.setTCChatMaxTokens(300);
    aiClient.setTCChatToolChoice("auto");
@@ -293,7 +305,7 @@ Let's walk through the complete flow of a tool call interaction using the exampl
    // ... (code from Step 2 above) ...
    ```
 
-5. **Configure Follow-Up Request**: Set parameters for the follow-up request
+5. **Configure Follow-Up Request(Optional)**: Set parameters for the follow-up request
    ```cpp
    aiClient.setTCReplyMaxTokens(350);
    aiClient.setTCReplyToolChoice("auto");
@@ -310,11 +322,6 @@ Let's walk through the complete flow of a tool call interaction using the exampl
    // ... (code from Step 4 above) ...
    ```
 
-8. **Reset Configuration**: Clear all tool call settings
-   ```cpp
-   aiClient.tcChatReset();
-   ```
-
 ## Key Considerations for Tool Call Follow-Up
 
 ### 1. Memory Management