chore: refine prompt

server/mcp_server_vefaas_function/README.md

@@ -1,76 +1,10 @@
 # veFaaS MCP Server
 
-veFaaS MCP Server 提供创建、更新、发布函数和添加触发器的能力。
+veFaaS MCP Server 提供创建、更新、发布 veFaaS 函数以及管理应用的自动化能力，帮助在 MCP 体系内快速对接 veFaaS 服务。
 
-| | |
-|------|------|
-| 描述 | veFaaS MCP Server 助你轻松管理函数和触发器生命周期|
-| 分类 | 容器与中间件 |
-| 标签 | FaaS，函数服务，函数，生命周期 |
+## 安装与快速开始
 
-## 能力概览
-
-- 创建/更新/发布 veFaaS 函数，并管理依赖安装与触发器配置。
-- 支持上传本地代码或通过已有制品（TOS、镜像）更新函数。
-- 覆盖 API 网关触发器和函数列表/存在性查询等补充能力。
-
-最新的参数、调用约束和注意事项都写在 `src/mcp_server_vefaas_function/vefaas_server.py` 的工具描述中，优先以代码为准。
-
-## 推荐流程
-
-1. `create_function` 创建函数（或确认已有函数 ID）。
-2. `upload_code` 推送代码，必要时排除 `.venv`、`node_modules` 等本地依赖。
-3. 如果触发了依赖安装任务，使用 `get_dependency_install_task_status` 轮询直至 `Succeeded/Failed`。
-4. 重新发布：`release_function`，随后持续调用 `get_function_release_status` 直到成功或失败。
-5. 发布成功后，再创建 API 网关触发器或其它下游资源。
-
-## 支持的运行时
-
-- `native-python3.12/v1`
-- `native-node20/v1`
-- `native/v1`
-
-以上运行时均为原生环境，仅提供解释器/运行时，不包含任何 Web 框架或额外工具。请把依赖声明在 `requirements.txt` / `package.json` 中，由 veFaaS 自动安装。
-
-## 常见注意事项
-
-- 本地修改完成后务必重新执行 `upload_code`；直接调用 `release_function` 只会复用旧制品。
-- 依赖安装长时间 `InProgress` 时，遵循工具描述中的退避策略并及时上报。
-- API 网关相关操作在创建触发器前应复用现有关联服务，避免重复创建。
-
-## 可适配平台
-
-streamable-http: 方舟，Python
-stdio: Python, Cursor, Claude macOS App, Cline
-
-## 服务开通链接 (整体产品)
-
-<https://console.volcengine.com/vefaas>
-
-## 鉴权方式
-
-OAuth 2.0
-
-## 在不同平台的配置
-
-### 方舟
-
-#### 体验中心
-
-1. 查看 MCP Server 详情
-在大模型生态广场，选择合适的 veFaaS MCP Server，并查看详情
-2. 选择 MCP Server 即将运行的平台
-检查当前 MCP Server 已适配的平台，并选择合适的平台
-3. 查看并对比可用的 Tools
-仔细查看可用的 Tools 的功能描述与所需的输入参数，并尝试运行对应的功能。
-4. 获取专属的URL或代码示例
-检查账号登录状态与服务开通情况，生成唯一URL
-5. 去对应的Client的平台进行使用
-点击快捷跳转按钮，前往方舟平台的体验中心进行对应MCP Server的体验
-
-### UVX
-
-请预先获取环境变量 VOLCENGINE_ACCESS_KEY 和 VOLCENGINE_SECRET_KEY。
+推荐使用 `uvx` 在本地快速拉起服务，请将 VOLCENGINE_ACCESS_KEY 和 VOLCENGINE_SECRET_KEY 内容替换为您账号的 AK/SK。
 
 ```json
 {
@@ -91,6 +25,33 @@ OAuth 2.0
 }
 ```
 
+## 前置条件与注意事项
+
+- 凭证管理：AK/SK 为账号敏感信息，请妥善保管，避免泄露。
+- 服务开通：生成公网访问链接依赖 API 网关等前置资源，使用前请确认账号已在控制台开通相关服务。
+- 代码包与运行时：编译型语言需先在本地生成 Linux 可执行文件；解释型语言需提供启动脚本和依赖声明文件（`requirements.txt` / `package.json`）以便平台自动安装依赖。
+- 模型与 Agent 效果：MCP 的执行效果受所选模型和 Agent 策略影响，若结果不理想，可补充上下文、调整提示词或切换模型/Agent。
+
+## 能力概览
+
+- 函数生命周期：创建、更新、发布、删除 veFaaS 函数，支持本地代码包、TOS 制品或镜像等多种来源。
+- 代码与依赖：上传代码、触发依赖安装任务、轮询安装状态，并可下载线上代码。
+- 运行时与配置：管理命令、环境变量、VPC 配置等基础属性。
+- 应用与访问：查询可用的 API 网关资源，创建并发布 veFaaS 应用，检索模板信息。
+
+## 支持的运行时
+
+- `native-python3.12/v1`
+- `native-node20/v1`
+- `native/v1`
+
+以上运行时均为原生环境，仅提供解释器/运行时，不包含额外框架或工具。
+
+## 官方资源
+
+- veFaaS 控制台：<https://console.volcengine.com/vefaas>
+- veFaaS MCP 操作文档：<https://www.volcengine.com/docs/6662/1852853#VCm9Uhw2>
+
 ## License
 
 volcengine/mcp-server is licensed under the MIT License.

server/mcp_server_vefaas_function/src/mcp_server_vefaas_function/vefaas_server.py

@@ -77,8 +77,6 @@ def validate_and_set_region(region: str = None) -> str:
 
 Note:
  - veFaaS Application is the top-level collection that contains veFaaS function, api-gateway and other production.
- - After all steps done, provide veFaaS Application infos like application_id, name, region and platform link if these info can get from context directly.
-    - veFaaS platform link template: https://console.volcengine.com/vefaas/region:vefaas+`region`/application/detail/`application_id`?tab=detail
 
 Error Handle Tips:
  - If there is **any authentication** error about vefaas application(create/release/get), let user apply auth by link: https://console.volcengine.com/iam/service/attach_custom_role?ServiceName=vefaas&policy1_1=APIGFullAccess&policy1_2=VeFaaSFullAccess&role1=ServerlessApplicationRole, then retry.
@@ -87,10 +85,10 @@ def validate_and_set_region(region: str = None) -> str:
 - This tool **CAN ONLY** be called if `create_function` tool be called in previous step.
 - If user want to create a veFaaS Application, **MUST** follow the steps:
  1. Use `list_vefaas_application_templates` to get all available templates.
- 2. Find most suitable template from all the templates and use `get_application_template_detail` to get template code.
+ 2. Find most suitable template from all the templates and use `get_application_template_detail` to get template code. If no related template found, no need to use template.
  3. Create/Release vefaas function.
  4. Create/Release vefaas application.
-- MUST EDIT vefaas.yml: Add application_id to `vefaas.yml` immediately after application created successfully.
+- MUST EDIT vefaas.yml after application created successfully: Add application_id to `vefaas.yml` immediately after application created successfully.
 
 """)
 def create_vefaas_application(function_id: Required[str], function_name: Required[str], gateway_name: Required[str], region: Optional[str] = None):
@@ -234,6 +232,7 @@ def get_vefaas_application(application_id: Required[str], region: Optional[str]
 Note:
  - veFaaS Application status **NOT** related to veFaaS function release status.
  - Usually use this tool to check application deployment status after application release.
+ - **MUST** provide some important info for user after application deployment finished: application_id, region, deployment status, access_url and app_platform_url. (can get these info from tool return value)
 
 """)
 def poll_vefaas_application_status(application_id: Required[str], region: Optional[str] = None):
@@ -291,13 +290,23 @@ def poll_vefaas_application_status(application_id: Required[str], region: Option
                 errLogs.append("Failed to release application due to an authentication error. Please visit https://console.volcengine.com/iam/service/attach_custom_role?ServiceName=vefaas&policy1_1=APIGFullAccess&policy1_2=VeFaaSFullAccess&role1=ServerlessApplicationRole to grant the required permissions and then try again.")    
         except Exception as e:
             logger.error(f"Failed to get application log: {str(e)}")              
-
+
+    # get system_url
+    system_url = ""
+    try:
+        cloud_resource = json.loads(result["CloudResource"])
+        system_url = cloud_resource['framework']['url']['system_url']
+    except Exception as e:
+        logger.error(f"Failed to get system_url: {str(e)}")
+
     responseInfo = {
         "Id": result["Id"],
         "Name": result["Name"],
         "Status": result["Status"],
         "Config": result["Config"],
         "Region": result["Region"],
+        "AccessUrl": system_url,
+        "AppPlatformUrl": f"https://console.volcengine.com/vefaas/region:vefaas+{region}/application/detail/{application_id}?tab=detail",
         "NewRevisionNumber": result.get("NewRevisionNumber"),
     }
     if len(errLogs) > 0:
@@ -319,10 +328,6 @@ def poll_vefaas_application_status(application_id: Required[str], region: Option
  - command **MUST** be a runnable script, e.g., `./run.sh`.
  - region **MUST** be `cn-beijing`, `cn-shanghai`, `cn-guangzhou`, or `ap-southeast-1`.
  - Supplying `enable_vpc=true` requires `vpc_id`, `subnet_ids`, and `security_group_ids`.
- - Declare every framework/server dependency in `requirements.txt` / `package.json`; do not bundle virtualenvs.
- - Module CLIs are not on PATH. Invoke them with `python -m module_name ...` or start the server in code—running `gunicorn ...` or `uvicorn ...` directly will fail.
- - Keep startup scripts focused on launching the app; skip extra installs/build once `upload_code` has run.
- - Store templates/static assets as files and sanity-check imports before uploading.
  - After all steps done, provide veFaaS function infos like function_id, name, region, runtime and platfrom link if these infos can get from context directly.
     - veFaaS platform link template: https://console.volcengine.com/vefaas/region:vefaas+`region`/function/detail/`function_id`?tab=config
 
@@ -614,6 +619,8 @@ def get_function_release_status(function_id: str, region: str = None):
 
 **CRITICAL REQUIREMENT**:
  - Can **only** use this tool to check vefaas function release status, **NEVER** try to get release status by other ways.
+ - After function release finished, provide some important info for user: function_id, region, release status, and vefaas function platform url.
+  - `vefaas function platform url` is `https://console.volcengine.com/vefaas/function/detail?functionId={function_id}&region={region}`
 """)
 def poll_function_release_status(function_id: str, region: str = None):
     region = validate_and_set_region(region)
@@ -631,6 +638,7 @@ def poll_function_release_status(function_id: str, region: str = None):
             time.sleep(interval)
         else:
             return response
+
     return response
 
 def generate_random_name(prefix="mcp", length=8):
@@ -691,7 +699,7 @@ def list_existing_api_gateways(region: str = None):
             if gateway.get("Region") == region and gateway.get("Status") in ["Running", "Creating"]:
                 result.append({
                     "Name": gateway.get("Name", ""),
-                    "ID": gateway.get("ID", ""),
+                    "ID": gateway.get("Id", ""),
                     "Region": gateway.get("Region", ""),
                     "Type": gateway.get("Type", ""),
                     "Status": gateway.get("Status", ""),
@@ -746,7 +754,7 @@ def create_api_gateway(name: str = None, region: str = "cn-beijing"):
 Note:
  - Use this tool to select one running api gateway for create veFaaS Application.
  - If no running gateway exists, will create a new one and wait for it to be running.
- - If can this tool can not fetch a running gateway after timeout, please retry and at most try 3 times.
+ - If can this tool returned error, should retry and at most retry 3 times.
 """)
 
 def fetch_running_api_gateway(region: str = None):
@@ -784,9 +792,9 @@ def fetch_running_api_gateway(region: str = None):
                     raise Exception(f"Failed to create API Gateway after {create_api_gateway_failed_times} times")
                 time.sleep(interval)
     except Exception as e:
-        return f"Failed to fetch an running API Gateway: {str(e)}"
+        raise Exception(f"Failed to fetch an running API Gateway: {str(e)}")
 
-    return f"Failed to fetch an running API Gateway after {timeout} seconds"
+    raise Exception(f"Failed to fetch an running API Gateway after {timeout} seconds")
 
 def ensure_executable_permissions(folder_path: str):
     for root, _, files in os.walk(folder_path):
@@ -922,7 +930,11 @@ def _get_upload_code_description() -> str:
         "  - Rule 4: Python/Node deps: put them in 'requirements.txt'/'package.json'; veFaaS installs them as needed.\n"
         "  - Rule 5: HTTP server **MUST** listen on IP: 0.0.0.0, PORT: 8000.\n"
         "  - Rule 6: Store templates/static assets as files and sanity-check imports before uploading.\n"
-        "  - Rule 7: When uploading code, exclude local deps and noise (e.g., `.venv`, `site-packages`, `node_modules`, `.git`, build artifacts) via `local_folder_exclude`.\n\n"
+        "  - Rule 7: Declare every framework/server dependency in `requirements.txt` / `package.json`; do not bundle virtualenvs.\n"
+        "  - Rule 8: Module CLIs are not on PATH. Invoke them with `python -m module_name ...` or start the server in code—running `gunicorn ...` or `uvicorn ...` directly will fail.\n"
+        "  - Rule 9: Keep startup scripts focused on launching the app; skip extra installs/build once `upload_code` has run.\n"
+        "  - Rule 10: Store templates/static assets as files and sanity-check imports before uploading.\n"
+        "  - Rule 11: When uploading code, exclude local deps and noise (e.g., `.venv`, `site-packages`, `node_modules`, `.git`, build artifacts) via `local_folder_exclude`.\n\n"
     )
 
     # Detect run mode via FASTMCP_* environment variables.