From ca841f8b4ca6a5bcb36d1f80c5247357d84c3c5c Mon Sep 17 00:00:00 2001
From: LittleMouse <n15978654429@gmail.com>
Date: Tue, 31 Dec 2024 18:12:49 +0800
Subject: [PATCH 1/4] api_yolo add "exit" function.

---
 src/api/api_yolo.cpp | 21 +++++++++++++++++++++
 src/api/api_yolo.h   |  8 ++++++++
 2 files changed, 29 insertions(+)

diff --git a/src/api/api_yolo.cpp b/src/api/api_yolo.cpp
index 2774cc0..e7bdbe9 100644
--- a/src/api/api_yolo.cpp
+++ b/src/api/api_yolo.cpp
@@ -42,6 +42,27 @@ String ApiYolo::setup(ApiYoloSetupConfig_t config, String request_id)
     return work_id;
 }
 
+String ApiYolo::exit(String work_id, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["request_id"] = request_id;
+        doc["work_id"]    = work_id;
+        doc["action"]     = "exit";
+        serializeJson(doc, cmd);
+    }
+
+    _module_msg->sendCmdAndWaitToTakeMsg(
+        cmd.c_str(), request_id,
+        [&work_id](ResponseMsg_t& msg) {
+            // Copy work id
+            work_id = msg.work_id;
+        },
+        100);
+    return work_id;
+}
+
 int ApiYolo::inference(String& work_id, uint8_t* input, size_t& raw_len, String request_id)
 {
     String cmd;
diff --git a/src/api/api_yolo.h b/src/api/api_yolo.h
index fbd6368..e065182 100644
--- a/src/api/api_yolo.h
+++ b/src/api/api_yolo.h
@@ -28,6 +28,14 @@ public:
      */
     String setup(ApiYoloSetupConfig_t config = ApiYoloSetupConfig_t(), String request_id = "yolo_setup");
 
+    /**
+     * @brief Exit module YOLO, return YOLO work_id
+     *
+     * @param work_id
+     * @param request_id
+     * @return String
+     */
+    String exit(String work_id, String request_id = "yolo_exit");
     /**
      * @brief Inference input data by module LLM
      *

From ee5a466a82c2155cbe317fe60039a13938bc19ec Mon Sep 17 00:00:00 2001
From: LittleMouse <n15978654429@gmail.com>
Date: Mon, 6 Jan 2025 15:30:21 +0800
Subject: [PATCH 2/4] add vlm, depth anything module. add doc. update api.

---
 docs/cn.md                     | 640 +++++++++++++++++++++++++++++++++
 docs/en.md                     |  41 +++
 src/M5ModuleLLM.cpp            |   2 +
 src/M5ModuleLLM.h              |  14 +
 src/api/api_depth_anything.cpp | 106 ++++++
 src/api/api_depth_anything.h   |  60 ++++
 src/api/api_melotts.cpp        |  21 ++
 src/api/api_melotts.h          |   9 +
 src/api/api_vlm.cpp            | 137 +++++++
 src/api/api_vlm.h              |  72 ++++
 src/api/api_yolo.h             |   1 +
 11 files changed, 1103 insertions(+)
 create mode 100644 docs/cn.md
 create mode 100644 docs/en.md
 create mode 100644 src/api/api_depth_anything.cpp
 create mode 100644 src/api/api_depth_anything.h
 create mode 100644 src/api/api_vlm.cpp
 create mode 100644 src/api/api_vlm.h

diff --git a/docs/cn.md b/docs/cn.md
new file mode 100644
index 0000000..137bd59
--- /dev/null
+++ b/docs/cn.md
@@ -0,0 +1,640 @@
+# M5Module-LLM Arduino API
+
+[M5Module-LLM](https://github.com/m5stack/M5Module-LLM) Arduino驱动库API文档。
+
+## M5ModuleLLM Class
+
+`M5ModuleLLM`用于初始化LLM Module, 并且提供内部成员用于快速初始化LLM的各个单元, 方便根据自己的需求构建应用。
+
+```cpp
+class M5ModuleLLM {
+public:
+    bool begin(Stream* targetPort);
+    bool checkConnection();
+    void update();
+
+    m5_module_llm::ApiSys sys;
+    m5_module_llm::ApiLlm llm;
+    m5_module_llm::ApiAudio audio;
+    m5_module_llm::ApiTts tts;
+    m5_module_llm::ApiTts melotts;
+    m5_module_llm::ApiKws kws;
+    m5_module_llm::ApiAsr asr;
+    m5_module_llm::ApiAsr yolo;
+    m5_module_llm::ModuleMsg msg;
+    m5_module_llm::ModuleComm comm;
+private:
+};
+```
+
+### begin
+
+**函数原型:**
+
+```cpp
+bool begin(Stream* targetPort);
+```
+
+**功能说明:**
+
+- 初始化LLM Module UART接口配置
+
+**传入参数:**
+
+- Stream* targetPort:
+    - 传入Serial指针
+
+**返回值:**
+
+- bool:
+    - true: 初始化成功
+    - false: 初始化失败
+
+### checkConnection
+
+**函数原型:**
+
+```cpp
+bool checkConnection();
+```
+
+**功能说明:**
+
+- 发送`sys.ping`指令, 检查LLM Module连接状态
+
+**传入参数:**
+
+- null
+
+**返回值:**
+
+- bool:
+    - true: 模组响应
+    - false: 模组无响应
+
+### update
+
+**函数原型:**
+
+```cpp
+void update();
+```
+
+**功能说明:**
+
+- 拉取LLM Module UART响应数据, 该API需包含在Loop中循环执行。
+
+**传入参数:**
+
+- null
+
+**返回值:**
+
+- null
+
+## ApiSys Class
+
+`M5ModuleLLM`的内部成员`ApiSys sys`用于控制SYS单元实现系统复位等操作。
+
+### ping
+
+**函数原型:**
+
+```cpp
+int ping();
+```
+
+**功能说明:**
+
+- 发送`sys.ping`指令, 检查LLM Module连接状态
+
+**传入参数:**
+
+- null
+
+**返回值:**
+
+- int:
+    - MODULE_LLM_OK / Error Code
+
+### reset
+
+**函数原型:**
+
+```cpp
+int reset(bool waitResetFinish = true);
+```
+
+**功能说明:**
+
+- 发送`sys.reset`指令, 复位软件服务。
+
+**传入参数:**
+
+- bool waitResetFinish:
+    - true:阻塞等待复位
+    - false:非阻塞执行复位
+
+**返回值:**
+
+- int:
+    - MODULE_LLM_OK / Error Code
+
+### reboot
+
+**函数原型:**
+
+```cpp
+int reboot();
+```
+
+**功能说明:**
+
+- 发送`sys.reboot`指令, 复位系统。
+
+**传入参数:**
+
+- null
+
+**返回值:**
+
+- int:
+    - MODULE_LLM_OK / Error Code
+
+## ApiAudio Class
+
+`M5ModuleLLM`的内部成员`ApiAudio audio`用于控制AUDIO单元的初始化和配置。
+
+### setup
+
+**函数原型:**
+
+```cpp
+String setup(ApiAudioSetupConfig_t config = ApiAudioSetupConfig_t(), String request_id = "audio_setup");
+```
+
+**功能说明:**
+
+- 初始化Audio单元, 开启系统声卡。(使用KWS和TTS前需开启该功能)
+
+**传入参数:**
+
+ApiAudioSetupConfig_t config:
+
+- LLM单元初始化配置:
+- String request_id:
+    - 会话id, 使用默认即可。
+
+```cpp
+struct ApiAudioSetupConfig_t {
+    int capcard      = 0;
+    int capdevice    = 0;
+    float capVolume  = 0.5;
+    int playcard     = 0;
+    int playdevice   = 1;
+    float playVolume = 0.15;
+};
+```
+
+| 参数         | 描述       | 输入值                             |
+|------------|----------|---------------------------------|
+| capcard    | 麦克风声卡的索引 | 系统默认声卡:0                        |
+| capdevice  | 麦克风设备索引  | 板载硅麦:0                          |
+| capVolume  | 输入的音量    | 0.0～10.0 (1<volume将增益, 默认值为0.5) |
+| playcard   | 扬声器声卡的索引 | 系统默认声卡:0                        |
+| playdevice | 扬声器设备索引  | 板载扬声器:1                         |
+| playVolume | 输出的音量    | 0.0～10.0 (1<volume将增益, 默认值为0.5) |
+
+**返回值:**
+
+- String:
+    - audio_work_id: audio单元work_id
+
+## ApiKws Class
+
+`M5ModuleLLM`的内部成员`ApiKws kws`用于控制KWS单元的初始化和配置。
+
+### setup
+
+**函数原型:**
+
+```cpp
+String setup(ApiKwsSetupConfig_t config = ApiKwsSetupConfig_t(), String request_id = "kws_setup");
+```
+
+**功能说明:**
+
+- 初始化KWS单元, 并配置唤醒关键字。
+
+**传入参数:**
+
+ApiKwsSetupConfig_t config:
+
+- KWS单元初始化配置:
+- String request_id:
+    - 会话id, 使用默认即可。
+
+```cpp
+struct ApiKwsSetupConfig_t {
+    String kws             = "HELLO";
+    String model           = "sherpa-onnx-kws-zipformer-gigaspeech-3.3M-2024-01-01";
+    String response_format = "kws.bool";
+    String input           = "sys.pcm";
+    bool enoutput          = true;
+};
+```
+
+| 参数       | 描述         | 输入值                                                                                                                           |
+|----------|------------|-------------------------------------------------------------------------------------------------------------------------------|
+| model    | 转换模型       | 英文模型: "sherpa-onnx-kws-zipformer-gigaspeech-3.3M-2024-01-01"<br>中文模型: "sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01" |
+| kws      | KWS唤醒词文本设置 | 不允许中文/英文混合, 英文要求全大写                                                                                                           |
+| enoutput | 启用UART输出   | 启用: true<br>禁用: false                                                                                                         |
+
+**返回值:**
+
+- String:
+    - kws_work_id: kws单元work_id
+
+## ApiAsr Class
+
+`M5ModuleLLM`的内部成员`ApiAsr asr`用于控制ASR单元的初始化和配置。
+
+### setup
+
+**函数原型:**
+
+```cpp
+String setup(ApiAsrSetupConfig_t config = ApiAsrSetupConfig_t(), String request_id = "asr_setup");
+```
+
+**功能说明:**
+
+- 初始化ASR单元, 开启语音转文本功能。
+
+**传入参数:**
+
+ApiAsrSetupConfig_t config:
+
+- ASR单元初始化配置:
+- String request_id:
+    - 会话id, 使用默认即可。
+
+```cpp
+struct ApiAsrSetupConfig_t {
+    String model           = "sherpa-ncnn-streaming-zipformer-20M-2023-02-17";
+    String response_format = "asr.utf-8.stream";
+    String input           = ["sys.pcm", "kws.1000"];
+    bool enoutput          = true;
+    float rule1            = 2.4;
+    float rule2            = 1.2;
+    float rule3            = 30.0;
+};
+```
+
+| 参数              | 描述            | 输入值                                                                                                                 |
+|-----------------|---------------|---------------------------------------------------------------------------------------------------------------------|
+| model           | 转换模型          | 英文模型: "sherpa-ncnn-streaming-zipformer-20M-2023-02-17"<br>中文模型: "sherpa-ncnn-streaming-zipformer-zh-14M-2023-02-23" |
+| response_format | 输出格式          | 普通输出: "asr.utf-8"<br>流式输出: "asr.utf-8.stream"                                                                       |
+| input           | 输入            | KWS唤醒输入: "kws.xxx"(输入kws单元的work_id)<br>板载麦克风输入: "sys.pcm"<br>UART流式输入: "asr.wav.stream.base64"                      |
+| rule1           | 唤醒到未识别到内容超时时间 | 单位:秒                                                                                                                |
+| rule2           | 识别最大间隔时间      | 单位:秒                                                                                                                |
+| rule3           | 识别最长超时时间      | 单位:秒                                                                                                                |
+| enoutput        | 启用UART输出      | 启用: true<br>禁用: false                                                                                               |
+
+**返回值:**
+
+- String:
+    - asr_work_id: asr单元work_id
+
+## ApiLlm Class
+
+`M5ModuleLLM`的内部成员`ApiLlm llm`用于控制LLM单元的初始化和配置。
+
+### setup
+
+**函数原型:**
+
+```cpp
+String setup(ApiLlmSetupConfig_t config = ApiLlmSetupConfig_t(), String request_id = "llm_setup");
+```
+
+**功能说明:**
+
+- 初始化LLM单元, 支持配置LLM单元输入输出数据方式。
+
+**传入参数:**
+
+- ApiLlmSetupConfig_t config:
+    - LLM单元初始化配置:
+- String request_id:
+    - 会话id, 使用默认即可。
+
+```cpp
+struct ApiLlmSetupConfig_t {
+    String prompt;
+    String model           = "qwen2.5-0.5B-prefill-20e";
+    String response_format = "llm.utf-8.stream";
+    String input           = "llm.utf-8";
+    bool enoutput          = true;
+    bool enkws             = true;
+    int max_token_len      = 127;
+};
+```
+
+| 参数              | 描述                      | 输入值                                                                                             |
+|-----------------|-------------------------|-------------------------------------------------------------------------------------------------|
+| model           | 转换模型                    | 预置模型 "qwen2.5-0.5B-prefill-20e"                                                                 |
+| response_format | 输出格式                    | 普通输出: "llm.utf-8"<br>流式输出: "llm.utf-8.stream"                                                   |
+| input           | 输入                      | ASR输入: "asr.xxx"(输入asr单元的work_id)<br>UART输入: "llm.utf-8"<br>KWS唤醒打断: "kws.xxx"(输入kws单元的work_id) |
+| enkws           | KWS唤醒是否终止过程             | KWS打断过程: true<br>KWS不打断过程: false                                                                |
+| max_length      | 配置最大输出token(最大返回推理文本长度) | 最大值: 1024, 推荐使用127                                                                              |
+| prompt          | 模型初始化提示词                | String                                                                                          |
+| enoutput        | 启用UART输出                | 启用: true<br>禁用: false                                                                           |
+
+**返回值:**
+
+- String:
+    - llm_work_id: llm单元work_id
+
+### inference
+
+**函数原型:**
+
+```cpp
+int inference(String work_id, String input, String request_id = "llm_inference");
+```
+
+**功能说明:**
+
+- 输入数据, 开始推理。返回结果内容将进入`M5ModuleLLM.msg`中的`responseMsgList`列表容器中。
+
+**传入参数:**
+
+- String work_id:
+    - 调用的LLM单元work_id
+- String input:
+    - 输入文本
+- String request_id:
+    - 会话ID, 当同时存在多个会话的时候用于区分。
+
+**返回值:**
+
+- int:
+    - MODULE_LLM_OK / Error Code
+
+### inferenceAndWaitResult
+
+**函数原型:**
+
+```cpp
+int inferenceAndWaitResult(String work_id, String input, std::function<void(String&)> onResult, uint32_t timeout = 5000, String request_id = "llm_inference");
+```
+
+**功能说明:**
+
+- 输入数据, 开始推理。并阻塞等待返回结果, 然后调用callback函数。
+
+**传入参数:**
+
+- String work_id:
+    - 调用的LLM单元work_id
+- String input:
+    - 输入文本
+- void onResult(String&)
+    - 推理结果callback函数
+- uint32_t timeout:
+    - 等待推理超时时间
+- String request_id:
+    - 会话ID, 当同时存在多个会话的时候用于区分。
+
+**返回值:**
+
+- int:
+    - MODULE_LLM_OK / Error Code
+
+## ApiTts Class
+
+`M5ModuleLLM`的内部成员`ApiTts tts`用于控制TTS单元的初始化和配置。
+
+### setup
+
+**函数原型:**
+
+```cpp
+String setup(ApiTtsSetupConfig_t config = ApiTtsSetupConfig_t(), String request_id = "tts_setup");
+```
+
+**功能说明:**
+
+- 初始化TTS单元, 开启文本转语音功能。
+
+**传入参数:**
+
+ApiTtsSetupConfig_t config:
+
+- LLM单元初始化配置:
+- String request_id:
+    - 会话id, 使用默认即可。
+
+```cpp
+struct ApiTtsSetupConfig_t {
+    String model           = "single_speaker_english_fast";
+    String response_format = "tts.base64.wav";
+    String input           = "tts.utf-8.stream";
+    bool enoutput          = true;
+    bool enkws             = true;
+};
+```
+
+| 参数       | 描述          | 输入值                                                                                      |
+|----------|-------------|------------------------------------------------------------------------------------------|
+| model    | 转换模型        | 英文模型: "single_speaker_english_fast"<br>中文模型: "single_speaker_fast"                       |
+| input    | 输入          | LLM输入: "llm.xxx"(输入llm单元的work_id)<br>UART输入: "tts.utf-8"<br>UART流式输入: "tts.utf-8.stream" |
+| enkws    | KWS唤醒是否终止过程 | KWS打断过程: true<br>KWS不打断过程: false                                                         |
+| enoutput | 启用UART输出    | 启用: true<br>禁用: false                                                                    |
+
+**返回值:**
+
+- String:
+    - tts_work_id: tts单元work_id
+
+### inference
+
+**函数原型:**
+
+```cpp
+int inference(String work_id, String input, uint32_t timeout = 0, String request_id = "tts_inference");
+```
+
+**功能说明:**
+
+- 输入数据, 开始推理转换, 完成后将自动播放至扬声器。
+
+**传入参数:**
+
+- String work_id:
+    - 调用的TTS单元work_id
+- String input:
+    - 输入文本
+- uint32_t timeout:
+    - 等待推理超时时间
+- String request_id:
+    - 会话ID, 当同时存在多个会话的时候用于区分。
+
+**返回值:**
+
+- int:
+    - MODULE_LLM_OK / Error Code
+
+## ModuleMsg Class
+
+`M5ModuleLLM`的内部成员`ModuleMsg msg`提供了`responseMsgList`容器用于用于缓存接收LLM Module返回的各种信息。参考以下案例，在主循环中遍历获取返回结果。
+
+```cpp
+void loop()
+{
+    module_llm.update();
+
+    // Handle response msg
+    for (auto& msg : module_llm.msg.responseMsgList) {
+        // KWS msg
+        if (msg.work_id == kws_work_id) {
+            Serial.printf(">> Keyword detected\n");
+        }
+
+        // ASR msg
+        if (msg.work_id == asr_work_id) {
+            if (msg.object == "asr.utf-8.stream") {
+                // Parse and get asr result
+                JsonDocument doc;
+                deserializeJson(doc, msg.raw_msg);
+                String asr_result = doc["data"]["delta"].as<String>();
+                Serial.printf(">> %s\n", asr_result.c_str());
+            }
+        }
+    }
+    module_llm.msg.responseMsgList.clear();
+}
+
+```
+
+## VoiceAssistant Class
+
+`M5ModuleLLM_VoiceAssistant`用于快速创建LLM语音助手实例, 快速实现KWS(语音唤醒)->ASR(语音转文本)->LLM(大模型推理)->TTS(
+文本转语音)。
+
+- 初始化时候只需要将`M5ModuleLLM`实例传入构造函数, 并注册对应事件的回调函数即可完成语音助手创建。
+
+```cpp
+/*
+ * SPDX-FileCopyrightText: 2024 M5Stack Technology CO LTD
+ *
+ * SPDX-License-Identifier: MIT
+ */
+#include <Arduino.h>
+#include <M5Unified.h>
+#include <M5ModuleLLM.h>
+
+M5ModuleLLM module_llm;
+M5ModuleLLM_VoiceAssistant voice_assistant(&module_llm);
+
+/* On ASR data callback */
+void on_asr_data_input(String data, bool isFinish, int index)
+{
+    M5.Display.setTextColor(TFT_GREEN, TFT_BLACK);
+    M5.Display.printf(">> %s\n", data.c_str());
+
+    /* If ASR data is finish */
+    if (isFinish) {
+        M5.Display.setTextColor(TFT_YELLOW, TFT_BLACK);
+        M5.Display.print(">> ");
+    }
+};
+
+/* On LLM data callback */
+void on_llm_data_input(String data, bool isFinish, int index)
+{
+    M5.Display.print(data);
+
+    /* If LLM data is finish */
+    if (isFinish) {
+        M5.Display.print("\n");
+    }
+};
+
+void setup()
+{
+    M5.begin();
+    M5.Display.setTextSize(2);
+    M5.Display.setTextScroll(true);
+
+    /* Init module serial port */
+    Serial2.begin(115200, SERIAL_8N1, 16, 17);  // Basic
+    // Serial2.begin(115200, SERIAL_8N1, 13, 14);  // Core2
+    // Serial2.begin(115200, SERIAL_8N1, 18, 17);  // CoreS3
+
+    /* Init module */
+    module_llm.begin(&Serial2);
+
+    /* Make sure module is connected */
+    M5.Display.printf(">> Check ModuleLLM connection..\n");
+    while (1) {
+        if (module_llm.checkConnection()) {
+            break;
+        }
+    }
+
+    /* Begin voice assistant preset */
+    M5.Display.printf(">> Begin voice assistant..\n");
+    int ret = voice_assistant.begin("HELLO");
+    if (ret != MODULE_LLM_OK) {
+        while (1) {
+            M5.Display.setTextColor(TFT_RED);
+            M5.Display.printf(">> Begin voice assistant failed\n");
+        }
+    }
+
+    /* Register on ASR data callback function */
+    voice_assistant.onAsrDataInput(on_asr_data_input);
+
+    /* Register on LLM data callback function */
+    voice_assistant.onLlmDataInput(on_llm_data_input);
+
+    M5.Display.printf(">> Voice assistant ready\n");
+}
+
+void loop()
+{
+    /* Keep voice assistant preset update */
+    voice_assistant.update();
+}
+```
+
+## Error Code
+
+```cpp
+enum ModuleLLMErrorCode_t {
+    MODULE_LLM_OK                              = 0,
+    MODULE_LLM_RESET_WARN                      = -1,
+    MODULE_LLM_JSON_FORMAT_ERROR               = -2,
+    MODULE_LLM_ACTION_MATCH_FAILED             = -3,
+    MODULE_LLM_INFERENCE_DATA_PUSH_FAILED      = -4,
+    MODULE_LLM_MODEL_LOADING_FAILED            = -5,
+    MODULE_LLM_UNIT_NOT_EXIST                  = -6,
+    MODULE_LLM_UNKNOWN_OPERATION               = -7,
+    MODULE_LLM_UNIT_RESOURCE_ALLOCATION_FAILED = -8,
+    MODULE_LLM_UNIT_CALL_FAILED                = -9,
+    MODULE_LLM_MODEL_INIT_FAILED               = -10,
+    MODULE_LLM_MODEL_RUN_FAILED                = -11,
+    MODULE_LLM_MODULE_NOT_INITIALISED          = -12,
+    MODULE_LLM_MODULE_ALREADY_WORKING          = -13,
+    MODULE_LLM_MODULE_NOT_WORKING              = -14,
+    MODULE_LLM_NO_UPDATEABLE_MODULES           = -15,
+    MODULE_LLM_NO_MODULES_AVAILABLE_FOR_UPDATE = -16,
+    MODULE_LLM_FILE_OPEN_FAILED                = -17,
+    MODULE_LLM_WAIT_RESPONSE_TIMEOUT           = -97,
+    MODULE_LLM_RESPONSE_PARSE_FAILED           = -98,
+    MODULE_LLM_ERROR_NONE                      = -99,
+};
+```
+
diff --git a/docs/en.md b/docs/en.md
new file mode 100644
index 0000000..1714e53
--- /dev/null
+++ b/docs/en.md
@@ -0,0 +1,41 @@
+# LLM Module Arduino Quick Start
+
+## Overview
+
+The `LLM Module` can be used with various M5 controllers. This tutorial demonstrates how to control the LLM Module using the `M5Core` series in the `Arduino IDE` with the LLM Module driver library.
+
+<img src="https://m5stack.oss-cn-shenzhen.aliyuncs.com/resource/docs/static/assets/img/guide/llm/llm/llm_module_device_01.jpg" width="70%" />
+
+## Environment Setup
+
+- 1.Arduino IDE Installation: Refer to the [Arduino IDE Installation Guide](/en/arduino/arduino_ide) to complete the IDE installation.
+
+- 2.Board Manager Installation: Refer to the [Basic Environment Setup Guide](/en/arduino/arduino_board) to complete the M5Stack board manager installation and select the `M5Core` development board.
+
+<img src="https://m5stack.oss-cn-shenzhen.aliyuncs.com/resource/docs/static/assets/img/arduino/m5core/quickstart_arduino_core_selectboard.png" width="70%" />
+
+- 3.Library Installation: Refer to the [Library Management Guide](/en/arduino/arduino_library) to install the `LLM Module` driver library. (Follow prompts to install the dependency library `M5Unified`)
+
+<img src="https://m5stack.oss-cn-shenzhen.aliyuncs.com/resource/docs/static/assets/img/guide/llm/llm/llm_arduino_lib_01.jpg" width="70%" />
+<img src="https://m5stack.oss-cn-shenzhen.aliyuncs.com/resource/docs/static/assets/img/guide/llm/llm/llm_arduino_lib_02.jpg" width="70%" />
+
+## Program Compilation & Upload
+
+Open the example program "kws_asr" in the driver library, click the upload button, and the program will automatically compile and upload.The wake-up word used in the example program is "HELLO". After waiting for the device to be initialized, it will be woken up using the keyword.
+
+<img src="https://m5stack.oss-cn-shenzhen.aliyuncs.com/resource/docs/static/assets/img/guide/llm/llm/llm_arduino_example_01.jpg" width="70%" />
+<img src="https://m5stack.oss-cn-shenzhen.aliyuncs.com/resource/docs/static/assets/img/guide/llm/llm/llm_arduino_example_02.jpg" width="70%" />
+<img src="https://m5stack.oss-cn-shenzhen.aliyuncs.com/resource/docs/static/assets/img/guide/llm/llm/llm_arduino_example_03.jpg" width="70%" />
+<img src="https://m5stack.oss-cn-shenzhen.aliyuncs.com/resource/docs/static/assets/img/guide/llm/llm/llm_arduino_example_04.jpg" width="70%" />
+
+- Examples:
+  - `kws_asr`: Uses KWS to wake up and triggers ASR for speech-to-text conversion. (KWS+ASR)
+  - `text_assistant`: Inputs text into the LLM model, performs inference, and outputs the result in text form. (LLM)
+  - `tts`: Uses the TTS unit to convert text to speech for playback. (TTS)
+  - `voice_assistant`: Uses KWS to wake up, triggers ASR for speech-to-text conversion, inputs the converted text into the LLM for inference, and outputs the inference result through TTS as speech. (KWS+ASR+LLM+TTS)
+
+## Related Links
+
+- [LLM Module Arduino Lib](https://github.com/m5stack/M5Module-LLM)
+- [LLM Module Arduino Lib API](/en/guide/llm/llm/arduino_api)
+
diff --git a/src/M5ModuleLLM.cpp b/src/M5ModuleLLM.cpp
index 0bac44b..2ac9119 100644
--- a/src/M5ModuleLLM.cpp
+++ b/src/M5ModuleLLM.cpp
@@ -12,6 +12,7 @@ bool M5ModuleLLM::begin(Stream* serialPort)
     msg.init(&comm);
     sys.init(&msg);
     llm.init(&msg);
+    vlm.init(&msg);
     audio.init(&msg);
     tts.init(&msg);
     melotts.init(&msg);
@@ -19,6 +20,7 @@ bool M5ModuleLLM::begin(Stream* serialPort)
     asr.init(&msg);
     yolo.init(&msg);
     camera.init(&msg);
+    depthanything.init(&msg);
     return true;
 }
 
diff --git a/src/M5ModuleLLM.h b/src/M5ModuleLLM.h
index 42bae02..c652089 100644
--- a/src/M5ModuleLLM.h
+++ b/src/M5ModuleLLM.h
@@ -9,12 +9,14 @@
 #include "utils/msg.h"
 #include "api/api_sys.h"
 #include "api/api_llm.h"
+#include "api/api_vlm.h"
 #include "api/api_audio.h"
 #include "api/api_tts.h"
 #include "api/api_melotts.h"
 #include "api/api_kws.h"
 #include "api/api_asr.h"
 #include "api/api_yolo.h"
+#include "api/api_depth_anything.h"
 #include "api/api_camera.h"
 #include "api/api_version.h"
 
@@ -55,6 +57,12 @@ public:
      */
     m5_module_llm::ApiLlm llm;
 
+    /**
+     * @brief VLM module api set
+     *
+     */
+    m5_module_llm::ApiVlm vlm;
+
     /**
      * @brief Audio module api set
      *
@@ -97,6 +105,12 @@ public:
      */
     m5_module_llm::ApiYolo yolo;
 
+    /**
+     * @brief DepthAnything module api set
+     *
+     */
+    m5_module_llm::ApiDepthAnything depthanything;
+
     /**
      * @brief MSG module to handle module response message
      *
diff --git a/src/api/api_depth_anything.cpp b/src/api/api_depth_anything.cpp
new file mode 100644
index 0000000..b8ceb2d
--- /dev/null
+++ b/src/api/api_depth_anything.cpp
@@ -0,0 +1,106 @@
+/*
+ * SPDX-FileCopyrightText: 2024 M5Stack Technology CO LTD
+ *
+ * SPDX-License-Identifier: MIT
+ */
+#include "api_depth_anything.h"
+
+using namespace m5_module_llm;
+
+void ApiDepthAnything::init(ModuleMsg* moduleMsg)
+{
+    _module_msg = moduleMsg;
+}
+
+String ApiDepthAnything::setup(ApiDepthAnythingSetupConfig_t config, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["request_id"]              = request_id;
+        doc["work_id"]                 = "depth_anything";
+        doc["action"]                  = "setup";
+        doc["object"]                  = "depth_anything.setup";
+        doc["data"]["model"]           = config.model;
+        doc["data"]["response_format"] = config.response_format;
+        JsonArray inputArray           = doc["data"]["input"].to<JsonArray>();
+        for (const String& str : config.input) {
+            inputArray.add(str);
+        }
+        doc["data"]["enoutput"] = config.enoutput;
+        serializeJson(doc, cmd);
+    }
+
+    String work_id;
+    _module_msg->sendCmdAndWaitToTakeMsg(
+        cmd.c_str(), request_id,
+        [&work_id](ResponseMsg_t& msg) {
+            // Copy work id
+            work_id = msg.work_id;
+        },
+        5000);
+    return work_id;
+}
+
+int ApiDepthAnything::inference(String& work_id, uint8_t* input, size_t& raw_len, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["RAW"]        = raw_len;
+        doc["request_id"] = request_id;
+        doc["work_id"]    = work_id;
+        doc["action"]     = "inference";
+        doc["object"]     = "cv.jpeg.base64";
+        serializeJson(doc, cmd);
+    }
+
+    _module_msg->sendCmd(cmd.c_str());
+    _module_msg->sendRaw(input, raw_len);
+    return MODULE_LLM_OK;
+}
+
+int ApiDepthAnything::inferenceAndWaitResult(String& work_id, uint8_t* input, size_t& raw_len,
+                                             std::function<void(String&)> onResult, uint32_t timeout, String request_id)
+{
+    inference(work_id, input, raw_len, request_id);
+
+    uint32_t time_out_count = millis();
+    bool is_time_out        = false;
+    bool is_msg_finish      = false;
+    while (1) {
+        _module_msg->update();
+        _module_msg->takeMsg(request_id, [&time_out_count, &is_msg_finish, &onResult](ResponseMsg_t& msg) {
+            String response_msg;
+            {
+                JsonDocument doc;
+                deserializeJson(doc, msg.raw_msg);
+                response_msg = doc["data"]["delta"].as<String>();
+                if (!doc["data"]["finish"].isNull()) {
+                    is_msg_finish = doc["data"]["finish"];
+                    if (is_msg_finish) {
+                        response_msg += '\n';
+                    }
+                }
+            }
+            if (onResult) {
+                onResult(response_msg);
+            }
+            time_out_count = millis();
+        });
+
+        if (is_msg_finish) {
+            break;
+        }
+
+        if (millis() - time_out_count > timeout) {
+            is_time_out = true;
+            break;
+        }
+    }
+
+    if (is_time_out) {
+        return MODULE_LLM_WAIT_RESPONSE_TIMEOUT;
+    }
+    return MODULE_LLM_OK;
+}
diff --git a/src/api/api_depth_anything.h b/src/api/api_depth_anything.h
new file mode 100644
index 0000000..c1138e7
--- /dev/null
+++ b/src/api/api_depth_anything.h
@@ -0,0 +1,60 @@
+/*
+ * SPDX-FileCopyrightText: 2024 M5Stack Technology CO LTD
+ *
+ * SPDX-License-Identifier: MIT
+ */
+#pragma once
+#include "../utils/msg.h"
+#include <Arduino.h>
+
+namespace m5_module_llm {
+struct ApiDepthAnythingSetupConfig_t {
+    String model              = "depth_anything";
+    String response_format    = "jpeg.base64.stream";
+    std::vector<String> input = {"depth_anything.jpeg.raw"};
+    bool enoutput             = true;
+};
+
+class ApiDepthAnything {
+public:
+    void init(ModuleMsg* moduleMsg);
+
+    /**
+     * @brief Setup module YOLO, return YOLO work_id
+     *
+     * @param config
+     * @param request_id
+     * @return String
+     */
+    String setup(ApiDepthAnythingSetupConfig_t config = ApiDepthAnythingSetupConfig_t(),
+                 String request_id                    = "depth_anything_setup");
+
+    /**
+     * @brief Inference input data by module LLM
+     *
+     * @param raw_len
+     * @param work_id
+     * @param input
+     * @param request_id
+     * @return int
+     */
+    int inference(String& work_id, uint8_t* input, size_t& raw_len, String request_id = "depth_anything_inference");
+
+    /**
+     * @brief Inference input data by module LLM, and wait inference result
+     *
+     * @param raw_len
+     * @param work_id
+     * @param input
+     * @param onResult On inference result callback
+     * @param timeout
+     * @param request_id
+     * @return int
+     */
+    int inferenceAndWaitResult(String& work_id, uint8_t* input, size_t& raw_len, std::function<void(String&)> onResult,
+                               uint32_t timeout = 5000, String request_id = "depth_anything_inference");
+
+private:
+    ModuleMsg* _module_msg = nullptr;
+};
+}  // namespace m5_module_llm
diff --git a/src/api/api_melotts.cpp b/src/api/api_melotts.cpp
index b4e6f8b..adcadbf 100644
--- a/src/api/api_melotts.cpp
+++ b/src/api/api_melotts.cpp
@@ -44,6 +44,27 @@ String ApiMelotts::setup(ApiMelottsSetupConfig_t config, String request_id, Stri
     return work_id;
 }
 
+String ApiMelotts::exit(String work_id, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["request_id"] = request_id;
+        doc["work_id"]    = work_id;
+        doc["action"]     = "exit";
+        serializeJson(doc, cmd);
+    }
+
+    _module_msg->sendCmdAndWaitToTakeMsg(
+        cmd.c_str(), request_id,
+        [&work_id](ResponseMsg_t& msg) {
+            // Copy work id
+            work_id = msg.work_id;
+        },
+        100);
+    return work_id;
+}
+
 int ApiMelotts::inference(String work_id, String input, uint32_t timeout, String request_id)
 {
     String cmd;
diff --git a/src/api/api_melotts.h b/src/api/api_melotts.h
index 7a35219..22ab4ea 100644
--- a/src/api/api_melotts.h
+++ b/src/api/api_melotts.h
@@ -31,6 +31,15 @@ public:
     String setup(ApiMelottsSetupConfig_t config = ApiMelottsSetupConfig_t(), String request_id = "melotts_setup",
                  String language = "en_US");
 
+    /**
+     * @brief Exit module TTS, return TTS work_id
+     *
+     * @param work_id
+     * @param request_id
+     * @return String
+     */
+    String exit(String work_id, String request_id = "yolo_exit");
+
     /**
      * @brief Inference input data by TTS module
      *
diff --git a/src/api/api_vlm.cpp b/src/api/api_vlm.cpp
new file mode 100644
index 0000000..651b986
--- /dev/null
+++ b/src/api/api_vlm.cpp
@@ -0,0 +1,137 @@
+/*
+ * SPDX-FileCopyrightText: 2024 M5Stack Technology CO LTD
+ *
+ * SPDX-License-Identifier: MIT
+ */
+#include "api_vlm.h"
+#include "api_version.h"
+
+using namespace m5_module_llm;
+
+void ApiVlm::init(ModuleMsg* moduleMsg)
+{
+    _module_msg = moduleMsg;
+}
+
+String ApiVlm::setup(ApiVlmSetupConfig_t config, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["request_id"]              = request_id;
+        doc["work_id"]                 = "vlm";
+        doc["action"]                  = "setup";
+        doc["object"]                  = "vlm.setup";
+        doc["data"]["model"]           = config.model;
+        doc["data"]["response_format"] = config.response_format;
+        doc["data"]["enoutput"]        = config.enoutput;
+        doc["data"]["enkws"]           = config.enkws;
+        doc["data"]["max_token_len"]   = config.max_token_len;
+        doc["data"]["prompt"]          = config.prompt;
+        if (!llm_version) {
+            doc["data"]["model"] = "qwen2.5-0.5b";
+            doc["data"]["input"] = config.input[0];
+        } else {
+            JsonArray inputArray = doc["data"]["input"].to<JsonArray>();
+            for (const String& str : config.input) {
+                inputArray.add(str);
+            }
+        }
+        serializeJson(doc, cmd);
+    }
+
+    String llm_work_id;
+    _module_msg->sendCmdAndWaitToTakeMsg(
+        cmd.c_str(), request_id,
+        [&llm_work_id](ResponseMsg_t& msg) {
+            // Copy work id
+            llm_work_id = msg.work_id;
+        },
+        20000);
+    return llm_work_id;
+}
+
+String ApiVlm::exit(String work_id, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["request_id"] = request_id;
+        doc["work_id"]    = work_id;
+        doc["action"]     = "exit";
+        serializeJson(doc, cmd);
+    }
+
+    _module_msg->sendCmdAndWaitToTakeMsg(
+        cmd.c_str(), request_id,
+        [&work_id](ResponseMsg_t& msg) {
+            // Copy work id
+            work_id = msg.work_id;
+        },
+        100);
+    return work_id;
+}
+
+int ApiVlm::inference(String work_id, String input, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["request_id"]     = request_id;
+        doc["work_id"]        = work_id;
+        doc["action"]         = "inference";
+        doc["object"]         = "vlm.utf-8.stream";
+        doc["data"]["delta"]  = input;
+        doc["data"]["index"]  = 0;
+        doc["data"]["finish"] = true;
+        serializeJson(doc, cmd);
+    }
+
+    _module_msg->sendCmd(cmd.c_str());
+    return MODULE_LLM_OK;
+}
+
+int ApiVlm::inferenceAndWaitResult(String work_id, String input, std::function<void(String&)> onResult,
+                                   uint32_t timeout, String request_id)
+{
+    inference(work_id, input, request_id);
+
+    uint32_t time_out_count = millis();
+    bool is_time_out        = false;
+    bool is_msg_finish      = false;
+    while (1) {
+        _module_msg->update();
+        _module_msg->takeMsg(request_id, [&time_out_count, &is_msg_finish, &onResult](ResponseMsg_t& msg) {
+            String response_msg;
+            {
+                JsonDocument doc;
+                deserializeJson(doc, msg.raw_msg);
+                response_msg = doc["data"]["delta"].as<String>();
+                if (!doc["data"]["finish"].isNull()) {
+                    is_msg_finish = doc["data"]["finish"];
+                    if (is_msg_finish) {
+                        response_msg += '\n';
+                    }
+                }
+            }
+            if (onResult) {
+                onResult(response_msg);
+            }
+            time_out_count = millis();
+        });
+
+        if (is_msg_finish) {
+            break;
+        }
+
+        if (millis() - time_out_count > timeout) {
+            is_time_out = true;
+            break;
+        }
+    }
+
+    if (is_time_out) {
+        return MODULE_LLM_WAIT_RESPONSE_TIMEOUT;
+    }
+    return MODULE_LLM_OK;
+}
diff --git a/src/api/api_vlm.h b/src/api/api_vlm.h
new file mode 100644
index 0000000..01878d7
--- /dev/null
+++ b/src/api/api_vlm.h
@@ -0,0 +1,72 @@
+/*
+ * SPDX-FileCopyrightText: 2024 M5Stack Technology CO LTD
+ *
+ * SPDX-License-Identifier: MIT
+ */
+#pragma once
+#include "../utils/msg.h"
+#include <Arduino.h>
+
+namespace m5_module_llm {
+
+struct ApiVlmSetupConfig_t {
+    String prompt;
+    String model              = "internvl2.5-1B-ax630c";
+    String response_format    = "vlm.utf-8.stream";
+    std::vector<String> input = {"vlm.utf-8.stream"};
+    bool enoutput             = true;
+    bool enkws                = true;
+    // int max_token_len         = 127;
+    int max_token_len = 255;
+};
+
+class ApiVlm {
+public:
+    void init(ModuleMsg* moduleMsg);
+
+    /**
+     * @brief Setup module LLM, return LLM work_id
+     *
+     * @param config
+     * @param request_id
+     * @return String
+     */
+    String setup(ApiVlmSetupConfig_t config = ApiVlmSetupConfig_t(), String request_id = "vlm_setup");
+
+    /**
+     * @brief Exit module YOLO, return YOLO work_id
+     *
+     * @param work_id
+     * @param request_id
+     * @return String
+     */
+    String exit(String work_id, String request_id = "vlm_exit");
+
+    /**
+     * @brief Inference input data by module LLM
+     *
+     * @param work_id
+     * @param input
+     * @param request_id
+     * @return int
+     */
+    int inference(String work_id, String input, String request_id = "vlm_inference");
+
+    /**
+     * @brief Inference input data by module LLM, and wait inference result
+     *
+     * @param work_id
+     * @param input
+     * @param onResult On inference result callback
+     * @param timeout
+     * @param request_id
+     * @return int
+     */
+    int inferenceAndWaitResult(String work_id, String input, std::function<void(String&)> onResult,
+                               uint32_t timeout = 5000, String request_id = "vlm_inference");
+
+private:
+    ModuleMsg* _module_msg = nullptr;
+};
+
+}  // namespace m5_module_llm
diff --git a/src/api/api_yolo.h b/src/api/api_yolo.h
index e065182..ebccc1a 100644
--- a/src/api/api_yolo.h
+++ b/src/api/api_yolo.h
@@ -36,6 +36,7 @@ public:
      * @return String
      */
     String exit(String work_id, String request_id = "yolo_exit");
+
     /**
      * @brief Inference input data by module LLM
      *

From af55e48bc04f4da3edd97e2b220418f885fa4fd3 Mon Sep 17 00:00:00 2001
From: LittleMouse <n15978654429@gmail.com>
Date: Mon, 6 Jan 2025 17:22:47 +0800
Subject: [PATCH 3/4] update module api.

---
 src/api/api_asr.cpp            | 21 +++++++++++++++++++++
 src/api/api_asr.h              |  9 +++++++++
 src/api/api_audio.cpp          | 21 +++++++++++++++++++++
 src/api/api_audio.h            | 11 ++++++++++-
 src/api/api_camera.cpp         | 21 +++++++++++++++++++++
 src/api/api_camera.h           | 11 ++++++++++-
 src/api/api_depth_anything.cpp | 21 +++++++++++++++++++++
 src/api/api_depth_anything.h   | 11 ++++++++++-
 src/api/api_kws.cpp            | 21 +++++++++++++++++++++
 src/api/api_kws.h              |  9 +++++++++
 src/api/api_llm.cpp            | 21 +++++++++++++++++++++
 src/api/api_llm.h              |  9 +++++++++
 src/api/api_melotts.h          |  2 +-
 src/api/api_tts.cpp            | 21 +++++++++++++++++++++
 src/api/api_tts.h              |  9 +++++++++
 src/api/api_vlm.h              |  8 ++++----
 16 files changed, 218 insertions(+), 8 deletions(-)

diff --git a/src/api/api_asr.cpp b/src/api/api_asr.cpp
index 900d531..2766fa6 100644
--- a/src/api/api_asr.cpp
+++ b/src/api/api_asr.cpp
@@ -51,3 +51,24 @@ String ApiAsr::setup(ApiAsrSetupConfig_t config, String request_id, String langu
         10000);
     return work_id;
 }
+
+String ApiAsr::exit(String work_id, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["request_id"] = request_id;
+        doc["work_id"]    = work_id;
+        doc["action"]     = "exit";
+        serializeJson(doc, cmd);
+    }
+
+    _module_msg->sendCmdAndWaitToTakeMsg(
+        cmd.c_str(), request_id,
+        [&work_id](ResponseMsg_t& msg) {
+            // Copy work id
+            work_id = msg.work_id;
+        },
+        100);
+    return work_id;
+}
diff --git a/src/api/api_asr.h b/src/api/api_asr.h
index 34d2364..0cd34a0 100644
--- a/src/api/api_asr.h
+++ b/src/api/api_asr.h
@@ -34,6 +34,15 @@ public:
     String setup(ApiAsrSetupConfig_t config = ApiAsrSetupConfig_t(), String request_id = "asr_setup",
                  String language = "en_US");
 
+    /**
+     * @brief Exit module ASR, return ASR work_id
+     *
+     * @param work_id
+     * @param request_id
+     * @return String
+     */
+    String exit(String work_id, String request_id = "asr_exit");
+
 private:
     ModuleMsg* _module_msg = nullptr;
 };
diff --git a/src/api/api_audio.cpp b/src/api/api_audio.cpp
index 343a586..caae9ae 100644
--- a/src/api/api_audio.cpp
+++ b/src/api/api_audio.cpp
@@ -40,3 +40,24 @@ String ApiAudio::setup(ApiAudioSetupConfig_t config, String request_id)
         5000);
     return work_id;
 }
+
+String ApiAudio::exit(String work_id, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["request_id"] = request_id;
+        doc["work_id"]    = work_id;
+        doc["action"]     = "exit";
+        serializeJson(doc, cmd);
+    }
+
+    _module_msg->sendCmdAndWaitToTakeMsg(
+        cmd.c_str(), request_id,
+        [&work_id](ResponseMsg_t& msg) {
+            // Copy work id
+            work_id = msg.work_id;
+        },
+        100);
+    return work_id;
+}
diff --git a/src/api/api_audio.h b/src/api/api_audio.h
index a1ab88f..2bf9eff 100644
--- a/src/api/api_audio.h
+++ b/src/api/api_audio.h
@@ -23,7 +23,7 @@ public:
     void init(ModuleMsg* moduleMsg);
 
     /**
-     * @brief Setup module audio, return work_id
+     * @brief Setup module audio, return audio work_id
      *
      * @param config
      * @param request_id
@@ -31,6 +31,15 @@ public:
      */
     String setup(ApiAudioSetupConfig_t config = ApiAudioSetupConfig_t(), String request_id = "audio_setup");
 
+    /**
+     * @brief Exit module audio, return audio work_id
+     *
+     * @param work_id
+     * @param request_id
+     * @return String
+     */
+    String exit(String work_id, String request_id = "audio_exit");
+
 private:
     ModuleMsg* _module_msg = nullptr;
 };
diff --git a/src/api/api_camera.cpp b/src/api/api_camera.cpp
index 8b2c7da..395dc8b 100644
--- a/src/api/api_camera.cpp
+++ b/src/api/api_camera.cpp
@@ -39,3 +39,24 @@ String ApiCamera::setup(ApiCameraSetupConfig_t config, String request_id)
         5000);
     return work_id;
 }
+
+String ApiCamera::exit(String work_id, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["request_id"] = request_id;
+        doc["work_id"]    = work_id;
+        doc["action"]     = "exit";
+        serializeJson(doc, cmd);
+    }
+
+    _module_msg->sendCmdAndWaitToTakeMsg(
+        cmd.c_str(), request_id,
+        [&work_id](ResponseMsg_t& msg) {
+            // Copy work id
+            work_id = msg.work_id;
+        },
+        100);
+    return work_id;
+}
diff --git a/src/api/api_camera.h b/src/api/api_camera.h
index f0d0ea3..973de50 100644
--- a/src/api/api_camera.h
+++ b/src/api/api_camera.h
@@ -22,7 +22,7 @@ public:
     void init(ModuleMsg* moduleMsg);
 
     /**
-     * @brief Setup module camera, return work_id
+     * @brief Setup module camera, return camera work_id
      *
      * @param config
      * @param request_id
@@ -30,6 +30,15 @@ public:
      */
     String setup(ApiCameraSetupConfig_t config = ApiCameraSetupConfig_t(), String request_id = "camera_setup");
 
+    /**
+     * @brief Exit module camera, return camera work_id
+     *
+     * @param work_id
+     * @param request_id
+     * @return String
+     */
+    String exit(String work_id, String request_id = "camera_exit");
+
 private:
     ModuleMsg* _module_msg = nullptr;
 };
diff --git a/src/api/api_depth_anything.cpp b/src/api/api_depth_anything.cpp
index b8ceb2d..85d0a97 100644
--- a/src/api/api_depth_anything.cpp
+++ b/src/api/api_depth_anything.cpp
@@ -42,6 +42,27 @@ String ApiDepthAnything::setup(ApiDepthAnythingSetupConfig_t config, String requ
     return work_id;
 }
 
+String ApiDepthAnything::exit(String work_id, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["request_id"] = request_id;
+        doc["work_id"]    = work_id;
+        doc["action"]     = "exit";
+        serializeJson(doc, cmd);
+    }
+
+    _module_msg->sendCmdAndWaitToTakeMsg(
+        cmd.c_str(), request_id,
+        [&work_id](ResponseMsg_t& msg) {
+            // Copy work id
+            work_id = msg.work_id;
+        },
+        100);
+    return work_id;
+}
+
 int ApiDepthAnything::inference(String& work_id, uint8_t* input, size_t& raw_len, String request_id)
 {
     String cmd;
diff --git a/src/api/api_depth_anything.h b/src/api/api_depth_anything.h
index c1138e7..36cb450 100644
--- a/src/api/api_depth_anything.h
+++ b/src/api/api_depth_anything.h
@@ -20,7 +20,7 @@ public:
     void init(ModuleMsg* moduleMsg);
 
     /**
-     * @brief Setup module YOLO, return YOLO work_id
+     * @brief Setup module DepthAnything, return DepthAnything work_id
      *
      * @param config
      * @param request_id
@@ -29,6 +29,15 @@ public:
     String setup(ApiDepthAnythingSetupConfig_t config = ApiDepthAnythingSetupConfig_t(),
                  String request_id                    = "depth_anything_setup");
 
+    /**
+     * @brief Exit module DepthAnything, return DepthAnything work_id
+     *
+     * @param work_id
+     * @param request_id
+     * @return String
+     */
+    String exit(String work_id, String request_id = "depth_anything_exit");
+
     /**
      * @brief Inference input data by module LLM
      *
diff --git a/src/api/api_kws.cpp b/src/api/api_kws.cpp
index 07f417e..a4c3103 100644
--- a/src/api/api_kws.cpp
+++ b/src/api/api_kws.cpp
@@ -48,3 +48,24 @@ String ApiKws::setup(ApiKwsSetupConfig_t config, String request_id, String langu
         30000);
     return work_id;
 }
+
+String ApiKws::exit(String work_id, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["request_id"] = request_id;
+        doc["work_id"]    = work_id;
+        doc["action"]     = "exit";
+        serializeJson(doc, cmd);
+    }
+
+    _module_msg->sendCmdAndWaitToTakeMsg(
+        cmd.c_str(), request_id,
+        [&work_id](ResponseMsg_t& msg) {
+            // Copy work id
+            work_id = msg.work_id;
+        },
+        100);
+    return work_id;
+}
diff --git a/src/api/api_kws.h b/src/api/api_kws.h
index 6400eec..71dc51c 100644
--- a/src/api/api_kws.h
+++ b/src/api/api_kws.h
@@ -31,6 +31,15 @@ public:
     String setup(ApiKwsSetupConfig_t config = ApiKwsSetupConfig_t(), String request_id = "kws_setup",
                  String language = "en_US");
 
+    /**
+     * @brief Exit module KWS, return KWS work_id
+     *
+     * @param work_id
+     * @param request_id
+     * @return String
+     */
+    String exit(String work_id, String request_id = "kws_exit");
+
 private:
     ModuleMsg* _module_msg = nullptr;
 };
diff --git a/src/api/api_llm.cpp b/src/api/api_llm.cpp
index 57b4ee9..93e4fa9 100644
--- a/src/api/api_llm.cpp
+++ b/src/api/api_llm.cpp
@@ -51,6 +51,27 @@ String ApiLlm::setup(ApiLlmSetupConfig_t config, String request_id)
     return llm_work_id;
 }
 
+String ApiLlm::exit(String work_id, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["request_id"] = request_id;
+        doc["work_id"]    = work_id;
+        doc["action"]     = "exit";
+        serializeJson(doc, cmd);
+    }
+
+    _module_msg->sendCmdAndWaitToTakeMsg(
+        cmd.c_str(), request_id,
+        [&work_id](ResponseMsg_t& msg) {
+            // Copy work id
+            work_id = msg.work_id;
+        },
+        100);
+    return work_id;
+}
+
 int ApiLlm::inference(String work_id, String input, String request_id)
 {
     String cmd;
diff --git a/src/api/api_llm.h b/src/api/api_llm.h
index 27536e9..45a3455 100644
--- a/src/api/api_llm.h
+++ b/src/api/api_llm.h
@@ -33,6 +33,15 @@ public:
      */
     String setup(ApiLlmSetupConfig_t config = ApiLlmSetupConfig_t(), String request_id = "llm_setup");
 
+    /**
+     * @brief Exit module LLM, return LLM work_id
+     *
+     * @param work_id
+     * @param request_id
+     * @return String
+     */
+    String exit(String work_id, String request_id = "llm_exit");
+
     /**
      * @brief Inference input data by module LLM
      *
diff --git a/src/api/api_melotts.h b/src/api/api_melotts.h
index 22ab4ea..2708c82 100644
--- a/src/api/api_melotts.h
+++ b/src/api/api_melotts.h
@@ -38,7 +38,7 @@ public:
      * @param request_id
      * @return String
      */
-    String exit(String work_id, String request_id = "yolo_exit");
+    String exit(String work_id, String request_id = "tts_exit");
 
     /**
      * @brief Inference input data by TTS module
diff --git a/src/api/api_tts.cpp b/src/api/api_tts.cpp
index 6cfe661..7aee3c8 100644
--- a/src/api/api_tts.cpp
+++ b/src/api/api_tts.cpp
@@ -52,6 +52,27 @@ String ApiTts::setup(ApiTtsSetupConfig_t config, String request_id, String langu
     return work_id;
 }
 
+String ApiTts::exit(String work_id, String request_id)
+{
+    String cmd;
+    {
+        JsonDocument doc;
+        doc["request_id"] = request_id;
+        doc["work_id"]    = work_id;
+        doc["action"]     = "exit";
+        serializeJson(doc, cmd);
+    }
+
+    _module_msg->sendCmdAndWaitToTakeMsg(
+        cmd.c_str(), request_id,
+        [&work_id](ResponseMsg_t& msg) {
+            // Copy work id
+            work_id = msg.work_id;
+        },
+        100);
+    return work_id;
+}
+
 int ApiTts::inference(String work_id, String input, uint32_t timeout, String request_id)
 {
     String cmd;
diff --git a/src/api/api_tts.h b/src/api/api_tts.h
index e5b06b9..80aea4e 100644
--- a/src/api/api_tts.h
+++ b/src/api/api_tts.h
@@ -32,6 +32,15 @@ public:
     String setup(ApiTtsSetupConfig_t config = ApiTtsSetupConfig_t(), String request_id = "tts_setup",
                  String language = "en_US");
 
+    /**
+     * @brief Exit module TTS, return TTS work_id
+     *
+     * @param work_id
+     * @param request_id
+     * @return String
+     */
+    String exit(String work_id, String request_id = "tts_exit");
+
     /**
      * @brief Inference input data by TTS module
      *
diff --git a/src/api/api_vlm.h b/src/api/api_vlm.h
index 01878d7..cad9e30 100644
--- a/src/api/api_vlm.h
+++ b/src/api/api_vlm.h
@@ -25,7 +25,7 @@ public:
     void init(ModuleMsg* moduleMsg);
 
     /**
-     * @brief Setup module LLM, return LLM work_id
+     * @brief Setup module VLLM, return VLLM work_id
      *
      * @param config
      * @param request_id
@@ -34,7 +34,7 @@ public:
     String setup(ApiVlmSetupConfig_t config = ApiVlmSetupConfig_t(), String request_id = "vlm_setup");
 
     /**
-     * @brief Exit module YOLO, return YOLO work_id
+     * @brief Exit module VLLM, return VLLM work_id
      *
      * @param work_id
      * @param request_id
@@ -43,7 +43,7 @@ public:
     String exit(String work_id, String request_id = "vlm_exit");
 
     /**
-     * @brief Inference input data by module LLM
+     * @brief Inference input data by module VLLM
      *
      * @param work_id
      * @param input
@@ -53,7 +53,7 @@ public:
     int inference(String work_id, String input, String request_id = "vlm_inference");
 
     /**
-     * @brief Inference input data by module LLM, and wait inference result
+     * @brief Inference input data by module VLLM, and wait inference result
      *
      * @param work_id
      * @param input

From dfd4c1e5d8c94e52c60c603b5e33566f14961c0a Mon Sep 17 00:00:00 2001
From: LittleMouse <n15978654429@gmail.com>
Date: Mon, 6 Jan 2025 17:33:41 +0800
Subject: [PATCH 4/4] Increase llm setup timeout.

---
 src/api/api_llm.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/api/api_llm.cpp b/src/api/api_llm.cpp
index 93e4fa9..617c8bd 100644
--- a/src/api/api_llm.cpp
+++ b/src/api/api_llm.cpp
@@ -47,7 +47,7 @@ String ApiLlm::setup(ApiLlmSetupConfig_t config, String request_id)
             // Copy work id
             llm_work_id = msg.work_id;
         },
-        10000);
+        20000);
     return llm_work_id;
 }