> ## Documentation Index
> Fetch the complete documentation index at: https://docs.veadk.xyz/llms.txt
> Use this file to discover all available pages before exploring further.

# 评测

评测集保存用于检验智能体行为的用例，评测运行会重新执行这些用例并记录指标结果。本页索引 Agent API Server 的开发评测接口

这些 `/dev/...` 接口需要启用开发界面，并安装评测依赖，见 [Agent API Server 概述](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/overview.md)。Harness 部署使用[对应评测接口](https://docs.veadk.xyz/productions/api-reference/preview/zh/harness-runtime/create-eval-set.md)

> [!WARNING]
>
> 评测可能调用模型和工具，产生费用或执行外部操作。准备样本时应检查用户数据、工具权限和指标配置

## 评测工作流

先创建评测集，从已有会话添加用例，再选择指标运行评测。用例可以单独读取、替换或删除；运行后根据逐用例状态与指标结果判断是否通过

| 方法 | 路径 | 接口 |
| - | - | - |
| POST | `/dev/apps/{app_name}/eval-sets` | [创建评测集](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/create-eval-set.md) |
| GET | `/dev/apps/{app_name}/eval-sets` | [列出评测集](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/list-eval-sets.md) |
| POST | `/dev/apps/{app_name}/eval-sets/{eval_set_id}/add-session` | [从会话添加评测用例](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/add-session-to-eval-set.md) |
| GET | `/dev/apps/{app_name}/eval_sets/{eval_set_id}/evals` | [列出评测用例](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/list-eval-cases.md) |
| GET | `/dev/apps/{app_name}/eval-sets/{eval_set_id}/eval-cases/{eval_case_id}` | [获取评测用例](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/get-eval-case.md) |
| PUT | `/dev/apps/{app_name}/eval-sets/{eval_set_id}/eval-cases/{eval_case_id}` | [更新评测用例](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/update-eval-case.md) |
| DELETE | `/dev/apps/{app_name}/eval-sets/{eval_set_id}/eval-cases/{eval_case_id}` | [删除评测用例](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/delete-eval-case.md) |
| GET | `/dev/apps/{app_name}/metrics-info` | [列出评测指标](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/list-eval-metrics.md) |
| POST | `/dev/apps/{app_name}/eval-sets/{eval_set_id}/run` | [运行评测](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/run-evaluation.md) |
| GET | `/dev/apps/{app_name}/eval-results` | [列出评测结果](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/list-eval-results.md) |
| GET | `/dev/apps/{app_name}/eval-results/{eval_result_id}` | [获取评测结果](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/get-eval-result.md) |

## 兼容入口

以下接口供已有调用方继续使用。迁移时需检查各页的请求和响应差异，不能只替换 URL；列出评测用例目前仅提供上表中的下划线路径

| 方法 | 路径 | 接口 |
| - | - | - |
| PUT | `/dev/apps/{app_name}/eval_sets/{eval_set_id}/evals/{eval_case_id}` | [更新评测用例兼容接口](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/update-eval-case-legacy.md) |
| POST | `/dev/apps/{app_name}/eval_sets/{eval_set_id}/run_eval` | [运行评测兼容接口](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/run-evaluation-legacy.md) |
| GET | `/dev/apps/{app_name}/eval_results` | [列出评测结果兼容接口](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/list-eval-results-legacy.md) |
| POST | `/dev/apps/{app_name}/eval_sets/{eval_set_id}/add_session` | [添加评测用例兼容接口](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/add-session-to-eval-set-legacy.md) |
| DELETE | `/dev/apps/{app_name}/eval_sets/{eval_set_id}/evals/{eval_case_id}` | [删除评测用例兼容接口](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/delete-eval-case-legacy.md) |
| GET | `/dev/apps/{app_name}/eval_sets/{eval_set_id}/evals/{eval_case_id}` | [获取评测用例兼容接口](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/get-eval-case-legacy.md) |
| GET | `/dev/apps/{app_name}/eval_results/{eval_result_id}` | [获取评测结果兼容接口](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/get-eval-result-legacy.md) |
| POST | `/dev/apps/{app_name}/eval_sets/{eval_set_id}` | [创建评测集兼容接口](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/create-eval-set-legacy.md) |
| GET | `/dev/apps/{app_name}/eval_sets` | [列出评测集兼容接口](https://docs.veadk.xyz/productions/api-reference/preview/zh/agent-api-server/list-eval-sets-legacy.md) |


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.