✨ feat/llm: 完成deepseek、豆包、kimi、混元（腾讯元宝）的对接

.gitignore

@@ -64,3 +64,6 @@ STEP.md
 
 # config
 ec_config.h
+
+# codex
+.codex

README.md

@@ -203,23 +203,50 @@ Build-time configuration is layered:
 - [`components/embed_claw/ec_config_internal.h`](components/embed_claw/ec_config_internal.h) provides repo defaults and empty secret placeholders.
 - Create local [`main/ec_config.h`](main/ec_config.h) for project-specific overrides. Define only the macros you want to override. The build injects this header into `embed_claw`, so sensitive values do not need to live in the shared component tree.
 
-Create `main/ec_config.h` if needed, then set at least:
+Create `main/ec_config.h` if needed. The default preset is `Qwen`:
 
 ```c
+#define EC_USE_QWEN                1
+#define EC_USE_DEEPSEEK            0
+#define EC_USE_DOUBAO              0
+#define EC_USE_KIMI                0
+#define EC_USE_HUNYUAN             0
 #define EC_SECRET_SEARCH_KEY        "YOUR_TAVILY_API_KEY"
 #define EC_LLM_API_KEY              "YOUR_DASHSCOPE_API_KEY"
-#define EC_LLM_MODEL                "qwen-plus"
 #define EC_SECRET_FEISHU_APP_ID     "YOUR_FEISHU_APP_ID"
 #define EC_SECRET_FEISHU_APP_SECRET "YOUR_FEISHU_APP_SECRET"
 ```
 
-Default LLM URL (DashScope OpenAI-compatible):
+Under the default preset, the effective LLM settings are:
 
 ```c
-#define EC_LLM_API_URL "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions"
+#define EC_LLM_PROVIDER_NAME       "openai"
+#define EC_LLM_API_URL             "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions"
+#define EC_LLM_MODEL               "qwen-plus"
 ```
 
-If you skip Tavily or Feishu for now, you only need the Qwen-related keys.
+Built-in OpenAI-compatible LLM presets now include:
+
+- `Qwen`: DashScope + `qwen-plus`
+- `DeepSeek`: `https://api.deepseek.com/v1/chat/completions` + `deepseek-chat`
+- `Doubao`: `https://operator.las.cn-beijing.volces.com/api/v1/chat/completions` + `doubao-seed-1-6-251015`
+- `KiMi`: `https://api.moonshot.cn/v1/chat/completions` + `kimi-k2.5`
+- `Hunyuan`: `https://api.hunyuan.cloud.tencent.com/v1/chat/completions` + `hunyuan-turbos-latest`
+
+If you want to switch to DeepSeek, use:
+
+```c
+#define EC_USE_QWEN                0
+#define EC_USE_DEEPSEEK            1
+#define EC_LLM_API_KEY              "YOUR_DEEPSEEK_API_KEY"
+```
+
+Notes:
+
+- Explicit `EC_LLM_PROVIDER_NAME / EC_LLM_API_URL / EC_LLM_MODEL` overrides take precedence over preset-derived defaults.
+- Only one `EC_USE_*` LLM preset can be `1` at a time.
+- The Doubao preset defaults to the Volcengine Beijing OpenAI-compatible endpoint; override `EC_LLM_API_URL` or `EC_DOUBAO_LLM_API_URL` if your region or endpoint differs.
+- If you skip Tavily or Feishu for now, you only need the keys for the active LLM preset.
 
 Optional channel toggles:
 
@@ -597,9 +624,11 @@ esp_err_t ec_channel_demo(void)
 
 Currently used:
 
+- `EC_USE_QWEN=1`
+- Other built-in OpenAI-compatible presets: `EC_USE_DEEPSEEK / EC_USE_DOUBAO / EC_USE_KIMI / EC_USE_HUNYUAN`
 - `EC_LLM_PROVIDER_NAME` (default: `openai`)
-- DashScope OpenAI-compatible API
-- **qwen-plus**
+- Qwen preset defaults to DashScope OpenAI-compatible + **qwen-plus**
+- The other presets reuse the same OpenAI-compatible provider
 
 To add another provider:
 
@@ -610,7 +639,7 @@ To add another provider:
 5. Add one branch in `ec_llm_init_default()` (`components/embed_claw/llm/ec_llm.c`) to map provider name to getter
 6. Set `EC_LLM_PROVIDER_NAME` in `main/ec_config.h` (and set matching URL/model)
 
-The current runtime path is OpenAI-compatible providers (OpenAI, DeepSeek, Moonshot, Qwen, etc.). Other provider families are not wired yet.
+The current runtime path is OpenAI-compatible providers (OpenAI, DeepSeek, Moonshot/KiMi, Qwen, Doubao, Hunyuan, etc.). Other provider families are not wired yet.
 
 ## Possible next steps
 
@@ -624,7 +653,7 @@ For open-source distribution, repo defaults keep secret fields empty. Put real k
 
 Before running, set:
 
-- DashScope / Qwen API key  
+- API key for the active LLM preset  
 - `EC_LLM_PROVIDER_NAME` (default is `openai`)
 - Tavily API key  
 - Feishu App ID and App Secret  

README_ZH.md

@@ -202,23 +202,50 @@ EmbedClaw 使用 SPIFFS 保存人格、用户信息、会话与记忆：
 - [`components/embed_claw/ec_config_internal.h`](components/embed_claw/ec_config_internal.h) 提供仓库内默认值和空白密钥占位。
 - 如需放项目自己的参数或敏感信息，请新建本地 [`main/ec_config.h`](main/ec_config.h)，只定义你要覆盖的宏。构建系统会把这个头自动注入 `embed_claw`，因此不必把密钥写回共享组件目录。
 
-如不存在可先创建 `main/ec_config.h`，你至少需要填写：
+如不存在可先创建 `main/ec_config.h`。默认预设是 `Qwen`：
 
 ```c
+#define EC_USE_QWEN                1
+#define EC_USE_DEEPSEEK            0
+#define EC_USE_DOUBAO              0
+#define EC_USE_KIMI                0
+#define EC_USE_HUNYUAN             0
 #define EC_SECRET_SEARCH_KEY        "YOUR_TAVILY_API_KEY"
 #define EC_LLM_API_KEY              "YOUR_DASHSCOPE_API_KEY"
-#define EC_LLM_MODEL                "qwen-plus"
 #define EC_SECRET_FEISHU_APP_ID     "YOUR_FEISHU_APP_ID"
 #define EC_SECRET_FEISHU_APP_SECRET "YOUR_FEISHU_APP_SECRET"
 ```
 
-默认的 LLM 地址已经指向 DashScope OpenAI-Compatible 接口：
+在默认预设下，LLM 会自动落到：
 
 ```c
-#define EC_LLM_API_URL "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions"
+#define EC_LLM_PROVIDER_NAME       "openai"
+#define EC_LLM_API_URL             "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions"
+#define EC_LLM_MODEL               "qwen-plus"
 ```
 
-如果你暂时不用 Tavily 或飞书，可以只配置 Qwen 所需参数。
+当前内置的 OpenAI-Compatible LLM 预设有：
+
+- `Qwen`：DashScope + `qwen-plus`
+- `DeepSeek`：`https://api.deepseek.com/v1/chat/completions` + `deepseek-chat`
+- `Doubao`：`https://operator.las.cn-beijing.volces.com/api/v1/chat/completions` + `doubao-seed-1-6-251015`
+- `KiMi`：`https://api.moonshot.cn/v1/chat/completions` + `kimi-k2.5`
+- `Hunyuan`：`https://api.hunyuan.cloud.tencent.com/v1/chat/completions` + `hunyuan-turbos-latest`
+
+如果你想切到 DeepSeek，可在本地配置里改成：
+
+```c
+#define EC_USE_QWEN                0
+#define EC_USE_DEEPSEEK            1
+#define EC_LLM_API_KEY              "YOUR_DEEPSEEK_API_KEY"
+```
+
+说明：
+
+- 显式定义的 `EC_LLM_PROVIDER_NAME / EC_LLM_API_URL / EC_LLM_MODEL` 优先级高于预设宏。
+- 同一时刻只能有一个 `EC_USE_*` LLM 预设为 `1`。
+- Doubao 预设默认使用火山引擎北京区 OpenAI-Compatible endpoint；如果你的地域或接入点不同，请显式覆盖 `EC_LLM_API_URL` 或 `EC_DOUBAO_LLM_API_URL`。
+- 如果你暂时不用 Tavily 或飞书，可以只配置当前 LLM 预设所需参数。
 
 可选的 channel 开关示例：
 
@@ -628,9 +655,11 @@ esp_err_t ec_channel_demo(void)
 
 当前默认实际启用的是：
 
+- `EC_USE_QWEN=1`
+- 其它内置 OpenAI-Compatible 预设：`EC_USE_DEEPSEEK / EC_USE_DOUBAO / EC_USE_KIMI / EC_USE_HUNYUAN`
 - `EC_LLM_PROVIDER_NAME`（默认：`openai`）
-- DashScope OpenAI-Compatible 接口
-- `qwen-plus`
+- Qwen 预设默认落到 DashScope OpenAI-Compatible + `qwen-plus`
+- 其它预设同样复用这套 OpenAI-Compatible provider
 
 但抽象层已经拆出来了，扩展新 Provider 的路径很清晰：
 
@@ -644,7 +673,7 @@ esp_err_t ec_channel_demo(void)
 补充说明：
 
 - 但现阶段真正跑通的是 OpenAI-Compatible 路径
-- 所以如果你要接 OpenAI、DeepSeek、Moonshot、通义千问兼容接口，成本会最低
+- 所以如果你要接 OpenAI、DeepSeek、Moonshot/KiMi、通义千问、豆包或混元的兼容接口，成本会最低
 
 ## 适合继续演进的方向
 
@@ -658,7 +687,7 @@ esp_err_t ec_channel_demo(void)
 
 在正式运行前，请先填写：
 
-- DashScope / Qwen API Key
+- 当前预设对应的 LLM API Key
 - `EC_LLM_PROVIDER_NAME`（默认是 `openai`）
 - Tavily API Key
 - Feishu App ID / Secret

TODO.md

@@ -3,11 +3,10 @@
 ## LLM integration
 
 - [x] Qwen
-- [ ] XiaoZhi
-- [ ] DeepSeek
-- [ ] Doubao (ByteDance)
-- [ ] KiMi (Moonshot)
-- [ ] Hunyuan (Tencent)
+- [x] DeepSeek (OpenAI-compatible preset)
+- [x] Doubao (ByteDance, OpenAI-compatible preset)
+- [x] KiMi (Moonshot, OpenAI-compatible preset)
+- [x] Hunyuan (Tencent, OpenAI-compatible preset)
 
 ## Tools integration
 
@@ -35,4 +34,3 @@
 - [ ] Support more filesystems (beyond SPIFFS)
 - [ ] SD card support
 - [ ] Support ESP32C3
-

TODO_ZH.md

@@ -3,11 +3,10 @@
 ## LLM对接
 
 - [x] 对接qwen
-- [ ] 对接小智
-- [ ] 对接deepseek
-- [ ] 对接豆包
-- [ ] 对接KiMi
-- [ ] 对接混元
+- [x] 对接deepseek（OpenAI-Compatible 预设）
+- [x] 对接豆包（OpenAI-Compatible 预设）
+- [x] 对接KiMi（OpenAI-Compatible 预设）
+- [x] 对接混元（OpenAI-Compatible 预设）
 
 ## Tools对接
 

components/embed_claw/ec_config_internal.h

@@ -72,20 +72,113 @@ extern "C" {
 /*
  * [embed_claw/llm]
  * 如何配置:
- * - 覆盖 provider 名称、API 地址、密钥、模型和响应缓冲参数。
+ * - 可通过预设宏快速切换 Qwen / DeepSeek / Doubao / KiMi / Hunyuan，
+ *   也可直接覆盖 provider 名称、
+ *   API 地址、密钥、模型和响应缓冲参数。
  * 作用:
  * - 影响 LLM provider 选择、接入目标、推理容量、内存占用和工具调用解析上限。
+ * 优先级:
+ * - 显式定义的 EC_LLM_PROVIDER_NAME / EC_LLM_API_URL / EC_LLM_MODEL
+ *   优先于预设宏推导值。
  */
+#ifndef EC_USE_DEEPSEEK
+#define EC_USE_DEEPSEEK            0
+#endif
+
+#ifndef EC_USE_DOUBAO
+#define EC_USE_DOUBAO              0
+#endif
+
+#ifndef EC_USE_KIMI
+#define EC_USE_KIMI                0
+#endif
+
+#ifndef EC_USE_HUNYUAN
+#define EC_USE_HUNYUAN             0
+#endif
+
+#ifndef EC_USE_QWEN
+#if EC_USE_DEEPSEEK || EC_USE_DOUBAO || EC_USE_KIMI || EC_USE_HUNYUAN
+#define EC_USE_QWEN                0
+#else
+#define EC_USE_QWEN                1
+#endif
+#endif
+
+#if (EC_USE_QWEN + EC_USE_DEEPSEEK + EC_USE_DOUBAO + EC_USE_KIMI + EC_USE_HUNYUAN) > 1
+#error "Only one EC_USE_* LLM preset can be 1 at a time"
+#endif
+
+#ifndef EC_QWEN_LLM_API_URL
+#define EC_QWEN_LLM_API_URL        "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions"
+#endif
+
+#ifndef EC_QWEN_LLM_MODEL
+#define EC_QWEN_LLM_MODEL          "qwen-plus"
+#endif
+
+#ifndef EC_DEEPSEEK_LLM_API_URL
+#define EC_DEEPSEEK_LLM_API_URL    "https://api.deepseek.com/v1/chat/completions"
+#endif
+
+#ifndef EC_DEEPSEEK_LLM_MODEL
+#define EC_DEEPSEEK_LLM_MODEL      "deepseek-chat"
+#endif
+
+#ifndef EC_DOUBAO_LLM_API_URL
+#define EC_DOUBAO_LLM_API_URL      "https://ark.cn-beijing.volces.com/api/v3/chat/completions"
+#endif
+
+#ifndef EC_DOUBAO_LLM_MODEL
+#define EC_DOUBAO_LLM_MODEL        "doubao-seed-2-0-pro-260215"
+#endif
+
+#ifndef EC_KIMI_LLM_API_URL
+#define EC_KIMI_LLM_API_URL        "https://api.moonshot.cn/v1/chat/completions"
+#endif
+
+#ifndef EC_KIMI_LLM_MODEL
+#define EC_KIMI_LLM_MODEL          "kimi-k2.5"
+#endif
+
+#ifndef EC_HUNYUAN_LLM_API_URL
+#define EC_HUNYUAN_LLM_API_URL     "https://api.hunyuan.cloud.tencent.com/v1/chat/completions"
+#endif
+
+#ifndef EC_HUNYUAN_LLM_MODEL
+#define EC_HUNYUAN_LLM_MODEL       "hunyuan-turbos-latest"
+#endif
+
 #ifndef EC_LLM_API_URL
-#define EC_LLM_API_URL              "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions"
+#if EC_USE_DEEPSEEK
+#define EC_LLM_API_URL             EC_DEEPSEEK_LLM_API_URL
+#elif EC_USE_DOUBAO
+#define EC_LLM_API_URL             EC_DOUBAO_LLM_API_URL
+#elif EC_USE_KIMI
+#define EC_LLM_API_URL             EC_KIMI_LLM_API_URL
+#elif EC_USE_HUNYUAN
+#define EC_LLM_API_URL             EC_HUNYUAN_LLM_API_URL
+#else
+#define EC_LLM_API_URL             EC_QWEN_LLM_API_URL
+#endif
 #endif
 
 #ifndef EC_LLM_API_KEY
 #define EC_LLM_API_KEY              ""
 #endif
 
 #ifndef EC_LLM_MODEL
-#define EC_LLM_MODEL                "qwen-plus"
+#if EC_USE_DEEPSEEK
+#define EC_LLM_MODEL               EC_DEEPSEEK_LLM_MODEL
+#elif EC_USE_DOUBAO
+#define EC_LLM_MODEL               EC_DOUBAO_LLM_MODEL
+#elif EC_USE_KIMI
+#define EC_LLM_MODEL               EC_KIMI_LLM_MODEL
+#elif EC_USE_HUNYUAN
+#define EC_LLM_MODEL               EC_HUNYUAN_LLM_MODEL
+#else
+#define EC_LLM_MODEL               EC_QWEN_LLM_MODEL
+#endif
 #endif
 
 #ifndef EC_LLM_PROVIDER_NAME