migrate-bluejay

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Bluejay → Coval Migration

Bluejay → Coval 迁移指南

Migrate all configuration from Bluejay to Coval for

$ARGUMENTS

为

$ARGUMENTS

将所有配置从Bluejay迁移至Coval。

API References

API 参考文档

references/bluejay-api.md — Bluejay endpoints, auth, pagination
references/coval-api.md — Coval endpoints, auth, request schemas
references/entity-mapping.md — Field-by-field mapping tables
references/metric-type-mapping.md — Bluejay response_type → Coval metric_type

references/bluejay-api.md — Bluejay接口、鉴权、分页机制
references/coval-api.md — Coval接口、鉴权、请求 schema
references/entity-mapping.md — 字段级映射表
references/metric-type-mapping.md — Bluejay response_type → Coval metric_type 映射

Step 0: Pre-flight

步骤0：预检查

Collect API Keys

收集API密钥

Parse

$ARGUMENTS

for two positional args:

[bluejay-api-key] [coval-api-key]

If not provided, ask the user for each key individually.

从

$ARGUMENTS

中解析两个位置参数：

[bluejay-api-key] [coval-api-key]

。

如果未提供，分别向用户索要每个密钥。

Validate Bluejay Key

验证Bluejay密钥

bash

curl -s -o /dev/null -w "%{http_code}" \
  -H "X-API-Key: <BLUEJAY_KEY>" \
  "https://api.getbluejay.ai/v1/all-agents"

Expect

. On

, tell user key is invalid.

bash

curl -s -o /dev/null -w "%{http_code}" \
  -H "X-API-Key: <BLUEJAY_KEY>" \
  "https://api.getbluejay.ai/v1/all-agents"

预期返回

。若返回

，告知用户密钥无效。

Validate Coval Key

验证Coval密钥

bash

curl -s -o /dev/null -w "%{http_code}" \
  -H "X-API-Key: <COVAL_KEY>" \
  "https://api.coval.dev/v1/agents?page_size=1"

Expect

. On

, tell user key is invalid.

bash

curl -s -o /dev/null -w "%{http_code}" \
  -H "X-API-Key: <COVAL_KEY>" \
  "https://api.coval.dev/v1/agents?page_size=1"

预期返回

。若返回

，告知用户密钥无效。

Fetch Coval Voice Catalog

获取Coval语音库

bash

curl -s -H "X-API-Key: <COVAL_KEY>" \
  "https://api.coval.dev/v1/personas/voices"

Store the voice list for persona mapping in Step 4. Response shape:

json

{ "voices": [{ "voice_name": "aria", "supported_languages": ["en-US", ...] }] }

bash

curl -s -H "X-API-Key: <COVAL_KEY>" \
  "https://api.coval.dev/v1/personas/voices"

存储语音列表，用于步骤4中的角色映射。响应格式：

json

{ "voices": [{ "voice_name": "aria", "supported_languages": ["en-US", ...] }] }

Step 1: Inventory

步骤1：资源盘点

Fetch all Bluejay entities. Use the endpoints in references/bluejay-api.md.

Fetch these in parallel where possible:

Entity	Endpoint
Agents	`GET /v1/all-agents`
Simulations	`GET /v1/get-all-simulations`
Custom Metrics	`GET /v1/custom-metrics?page=1&page_size=100` (paginate)
Communities	`GET /v1/communities`
Schedules	Derive from simulations (no list-all endpoint — check each simulation)
Customer Personas	Fetch per agent: `GET /v1/customer-personas-by-agent/{agent_id}`
Folders	`GET /v1/all-folders`

For each agent, also fetch:

Latest prompt:

GET /v1/agents/{agent_id}/prompts/versions?page_size=1

(latest version)

Latest knowledge base:

GET /v1/agents/{agent_id}/knowledge-bases/versions?page_size=1

Digital humans: Fetch per simulation:

GET /v1/digital-humans-by-simulation/{simulation_id}

Present summary to user:

Found:
  - X agents
  - Y digital humans (across Z simulations)
  - N custom metrics
  - M communities
  - P customer personas
  - Q schedules
  - R folders

Proceed with migration?

Wait for user confirmation before continuing.

获取所有Bluejay实体，使用references/bluejay-api.md中的接口。

尽可能并行获取以下资源：

实体	接口
Agents	`GET /v1/all-agents`
Simulations	`GET /v1/get-all-simulations`
Custom Metrics	`GET /v1/custom-metrics?page=1&page_size=100` （分页获取）
Communities	`GET /v1/communities`
Schedules	从Simulations中提取（无批量获取接口——需逐个检查模拟场景）
Customer Personas	按Agent获取： `GET /v1/customer-personas-by-agent/{agent_id}`
Folders	`GET /v1/all-folders`

对于每个Agent，还需获取：

Step 2: Check Existing Coval State

步骤2：检查Coval现有状态

Fetch existing Coval entities to enable idempotency:

bash

undefined

获取Coval现有实体以实现幂等性：

bash

undefined

Agents

curl -s -H "X-API-Key: <COVAL_KEY>" "https://api.coval.dev/v1/agents?page_size=100"

Personas

curl -s -H "X-API-Key: <COVAL_KEY>" "https://api.coval.dev/v1/personas?page_size=100"

Test Sets

curl -s -H "X-API-Key: <COVAL_KEY>" "https://api.coval.dev/v1/test-sets?page_size=100"

Metrics

curl -s -H "X-API-Key: <COVAL_KEY>" "https://api.coval.dev/v1/metrics?page_size=100"

Run Templates

curl -s -H "X-API-Key: <COVAL_KEY>" "https://api.coval.dev/v1/run-templates?page_size=100"

Scheduled Runs

curl -s -H "X-API-Key: <COVAL_KEY>" "https://api.coval.dev/v1/scheduled-runs?page_size=100"


Paginate if `next_page_token` is present.

Build a lookup map of `display_name → ID` for each entity type. When checking for duplicates:
- If entity exists with `bluejay_migration: true` in metadata → **skip** (already migrated)
- If name collision with non-migrated entity → **ask user** whether to skip or rename

curl -s -H "X-API-Key: <COVAL_KEY>" "https://api.coval.dev/v1/scheduled-runs?page_size=100"


若存在`next_page_token`则分页获取。

为每种实体类型构建`display_name → ID`的查找映射。检查重复项时：
- 若实体元数据中存在`bluejay_migration: true` → **跳过**（已完成迁移）
- 若与非迁移实体重名 → **询问用户**是否跳过或重命名

Phase 1: Independent Entities (Steps 3–5)

第一阶段：独立实体（步骤3–5）

These have no cross-references. Execute in order but each step is independent.

这些实体无交叉引用，可按顺序执行，且每个步骤相互独立。

Step 3: Migrate Metrics

步骤3：迁移指标

Map Bluejay Custom Metrics → Coval Metrics.

For each Bluejay metric, use the mapping in references/metric-type-mapping.md.

For each metric:

Determine
```
metric_type
```
from Bluejay
```
response_type
```
Build
```
prompt
```
from Bluejay
```
scoring_guidance
```
+
```
description
```
For
```
qualitative
```
→ ask user for category list
For
```
quantitative
```
→ use Bluejay
```
min_value
```
/
```
max_value
```
if present, else ask
For
```
enum
```
→ use
```
enum_options
```
as
```
categories
```
If
```
eval_route == "AUDIO"
```
→ use
```
METRIC_AUDIO_LLM_*
```
variants
For
```
json
```
→ use
```
METRIC_LLM_BINARY
```
, warn user this is an approximation

bash

curl -s -X POST "https://api.coval.dev/v1/metrics" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "metric_name": "<bluejay_metric_name>",
    "description": "<bluejay_description>",
    "metric_type": "<mapped_type>",
    "prompt": "<from_scoring_guidance>",
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<bluejay_metric_id>"
    }
  }'

Add type-specific fields (

categories

min_value

max_value

) per the mapping table.

Record:

bluejay_metric_id → coval_metric_id

将Bluejay自定义指标映射至Coval指标。

对于每个Bluejay指标，使用references/metric-type-mapping.md中的映射规则。

每个指标的处理步骤：

根据Bluejay的
```
response_type
```
确定
```
metric_type
```
从Bluejay的
```
scoring_guidance
```
+
```
description
```
构建
```
prompt
```
若为
```
qualitative
```
类型 → 向用户询问分类列表
若为
```
quantitative
```
类型 → 若存在Bluejay的
```
min_value
```
/
```
max_value
```
则使用，否则询问用户
若为
```
enum
```
类型 → 将
```
enum_options
```
作为
```
categories
```
若
```
eval_route == "AUDIO"
```
→ 使用
```
METRIC_AUDIO_LLM_*
```
变体
若为
```
json
```
类型 → 使用
```
METRIC_LLM_BINARY
```
，并告知用户这是近似处理

bash

curl -s -X POST "https://api.coval.dev/v1/metrics" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "metric_name": "<bluejay_metric_name>",
    "description": "<bluejay_description>",
    "metric_type": "<mapped_type>",
    "prompt": "<from_scoring_guidance>",
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<bluejay_metric_id>"
    }
  }'

根据映射表添加类型特定字段（

categories

、

min_value

、

max_value

）。

记录映射关系：

bluejay_metric_id → coval_metric_id

Step 4: Migrate Personas

步骤4：迁移角色

Map Bluejay Digital Humans (voice/behavior) AND Customer Personas → Coval Personas.

将Bluejay数字人（语音/行为）和客户角色映射至Coval角色。

Digital Human → Persona

数字人 → 角色

For each unique digital human:

Voice mapping — Present the Bluejay voice config to the user:

Bluejay Digital Human: "<name>"
  Gender: <gender>
  Accent: <accent>
  Language: <language>

Available Coval voices: aria, asteria, athena, luna, ...

Which Coval voice should map to this configuration?

Cache voice mappings — if same gender+accent+language combo appears again, reuse the mapping without asking.

Map
```
language
```
→
```
language_code
```
(e.g.,
```
en
```
→
```
en-US
```
,
```
es
```
→
```
es-ES
```
)
Map
```
background_noise
```
→
```
background_sound
```
(see entity mapping reference)
Build
```
persona_prompt
```
from
```
role_description
```
+
```
traits
```
+ behavioral fields

bash

curl -s -X POST "https://api.coval.dev/v1/personas" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "<digital_human_name>",
    "voice_name": "<mapped_coval_voice>",
    "language_code": "<mapped_language_code>",
    "persona_prompt": "<built_from_role_description_and_traits>",
    "background_sound": "<mapped_background_noise>",
    "conversation_initiation": "<from_speaks_first_config>",
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<digital_human_id>"
    }
  }'

Record:

bluejay_digital_human_id → coval_persona_id

对于每个唯一的数字人：

语音映射 — 向用户展示Bluejay语音配置：

Bluejay数字人："<name>"
  性别: <gender>
  口音: <accent>
  语言: <language>

可用Coval语音：aria, asteria, athena, luna, ...

该配置应映射到哪个Coval语音？

缓存语音映射 — 若相同性别+口音+语言组合再次出现，直接复用映射无需询问。

将
```
language
```
映射为
```
language_code
```
（例如：
```
en
```
→
```
en-US
```
，
```
es
```
→
```
es-ES
```
）
将
```
background_noise
```
映射为
```
background_sound
```
（参考实体映射文档）
从
```
role_description
```
+
```
traits
```
+ 行为字段构建
```
persona_prompt
```

bash

curl -s -X POST "https://api.coval.dev/v1/personas" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "<digital_human_name>",
    "voice_name": "<mapped_coval_voice>",
    "language_code": "<mapped_language_code>",
    "persona_prompt": "<built_from_role_description_and_traits>",
    "background_sound": "<mapped_background_noise>",
    "conversation_initiation": "<from_speaks_first_config>",
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<digital_human_id>"
    }
  }'

记录映射关系：

bluejay_digital_human_id → coval_persona_id

Customer Persona → Persona

客户角色 → 角色

For each Bluejay Customer Persona:

bash

curl -s -X POST "https://api.coval.dev/v1/personas" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "<customer_persona_name>",
    "voice_name": "aria",
    "language_code": "en-US",
    "persona_prompt": "Name: <name>\nDescription: <description>\nGoal: <goal>",
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<customer_persona_id>",
      "bluejay_entity_type": "customer_persona"
    }
  }'

Ask user which voice to use for customer personas (default to

aria

对于每个Bluejay客户角色：

bash

curl -s -X POST "https://api.coval.dev/v1/personas" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "<customer_persona_name>",
    "voice_name": "aria",
    "language_code": "en-US",
    "persona_prompt": "Name: <name>\nDescription: <description>\nGoal: <goal>",
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<customer_persona_id>",
      "bluejay_entity_type": "customer_persona"
    }
  }'

询问用户客户角色使用的语音（默认使用

aria

）。

Step 5: Migrate Agents

步骤5：迁移Agent

Map Bluejay Agents → Coval Agents.

For each Bluejay agent:

Determine

model_type

: Bluejay

mode=VOICE

→

MODEL_TYPE_VOICE

mode=TEXT

→

MODEL_TYPE_TEXT

Build
```
prompt
```
from: latest Prompt version
```
prompt_text
```
+ agent
```
goals
```
(append as numbered list)
If agent has a
```
system_prompt
```
, prepend it to the prompt
Store
```
knowledge_base
```
content in agent metadata

If agent has a

folder

, prefix

display_name

with folder name:

"[FolderName] AgentName"

bash

curl -s -X POST "https://api.coval.dev/v1/agents" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "display_name": "<folder_prefix><agent_name>",
    "model_type": "<MODEL_TYPE_VOICE|MODEL_TYPE_TEXT>",
    "prompt": "<system_prompt + latest_prompt + goals>",
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<bluejay_agent_id>",
      "knowledge_base": "<kb_text_content>",
      "bluejay_connection_type": "<connection_type>",
      "bluejay_phone_number": "<phone_number>"
    }
  }'

Record:

bluejay_agent_id → coval_agent_id

将Bluejay Agent映射至Coval Agent。

对于每个Bluejay Agent：

确定

model_type

：Bluejay

mode=VOICE

→

MODEL_TYPE_VOICE

，

mode=TEXT

→

MODEL_TYPE_TEXT

构建
```
prompt
```
：最新提示词版本的
```
prompt_text
```
+ Agent的
```
goals
```
（以编号列表追加）
若Agent存在
```
system_prompt
```
，将其前置到prompt中
将知识库内容存储在Agent元数据中
若Agent属于某个文件夹，在
```
display_name
```
前添加文件夹名称：
```
"[FolderName] AgentName"
```

bash

curl -s -X POST "https://api.coval.dev/v1/agents" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "display_name": "<folder_prefix><agent_name>",
    "model_type": "<MODEL_TYPE_VOICE|MODEL_TYPE_TEXT>",
    "prompt": "<system_prompt + latest_prompt + goals>",
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<bluejay_agent_id>",
      "knowledge_base": "<kb_text_content>",
      "bluejay_connection_type": "<connection_type>",
      "bluejay_phone_number": "<phone_number>"
    }
  }'

记录映射关系：

bluejay_agent_id → coval_agent_id

Phase 2: Dependent on Phase 1 (Steps 6–7)

第二阶段：依赖第一阶段（步骤6–7）

Step 6: Migrate Test Sets

步骤6：迁移测试集

Map Bluejay Simulations + Communities → Coval Test Sets.

将Bluejay Simulations + Communities映射至Coval Test Sets。

Simulation → Test Set

模拟场景 → 测试集

For each Bluejay simulation:

bash

curl -s -X POST "https://api.coval.dev/v1/test-sets" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "display_name": "<simulation_name>",
    "description": "Migrated from Bluejay simulation <simulation_id>",
    "test_set_type": "SCENARIO",
    "test_set_metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<simulation_id>",
      "bluejay_entity_type": "simulation"
    }
  }'

Record:

bluejay_simulation_id → coval_test_set_id

对于每个Bluejay模拟场景：

bash

curl -s -X POST "https://api.coval.dev/v1/test-sets" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "display_name": "<simulation_name>",
    "description": "从Bluejay模拟场景<simulation_id>迁移而来",
    "test_set_type": "SCENARIO",
    "test_set_metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<simulation_id>",
      "bluejay_entity_type": "simulation"
    }
  }'

记录映射关系：

bluejay_simulation_id → coval_test_set_id

Community → Test Set

社区 → 测试集

For each Bluejay community:

bash

curl -s -X POST "https://api.coval.dev/v1/test-sets" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "display_name": "<community_title>",
    "description": "<community_description>",
    "test_set_type": "SCENARIO",
    "test_set_metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<community_id>",
      "bluejay_entity_type": "community"
    }
  }'

Record:

bluejay_community_id → coval_test_set_id

对于每个Bluejay社区：

bash

curl -s -X POST "https://api.coval.dev/v1/test-sets" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "display_name": "<community_title>",
    "description": "<community_description>",
    "test_set_type": "SCENARIO",
    "test_set_metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<community_id>",
      "bluejay_entity_type": "community"
    }
  }'

记录映射关系：

bluejay_community_id → coval_test_set_id

Step 7: Migrate Test Cases

步骤7：迁移测试用例

Map Bluejay Digital Human intents → Coval Test Cases.

For each digital human associated with a simulation:

Use
```
intent
```
→
```
input_str
```
Use
```
success_criteria
```
→
```
expected_behaviors
```
(split into array if multi-line)
Look up
```
coval_test_set_id
```
from the simulation mapping (Step 6)

scripted_responses

present → set

input_type: "SCRIPT"

with

simulation_metadata_input.script_turns

bash

curl -s -X POST "https://api.coval.dev/v1/test-cases" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "test_set_id": "<coval_test_set_id>",
    "input_str": "<digital_human_intent>",
    "expected_behaviors": ["<success_criteria_line_1>", "<success_criteria_line_2>"],
    "description": "<digital_human_name>: <digital_human_tag>",
    "input_type": "SCENARIO",
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<digital_human_id>"
    }
  }'

For scripted test cases:

bash

curl -s -X POST "https://api.coval.dev/v1/test-cases" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "test_set_id": "<coval_test_set_id>",
    "input_str": "<digital_human_intent>",
    "expected_behaviors": ["<success_criteria>"],
    "input_type": "SCRIPT",
    "simulation_metadata_input": {
      "script_turns": ["<turn_1>", "<turn_2>", "..."]
    },
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<digital_human_id>"
    }
  }'

Record:

bluejay_digital_human_id → coval_test_case_id

将Bluejay数字人意图映射至Coval测试用例。

对于每个与模拟场景关联的数字人：

使用
```
intent
```
作为
```
input_str
```
使用
```
success_criteria
```
作为
```
expected_behaviors
```
（若为多行则拆分为数组）
从步骤6的模拟场景映射中查找
```
coval_test_set_id
```

若存在

scripted_responses

→ 设置

input_type: "SCRIPT"

并传入

simulation_metadata_input.script_turns

bash

curl -s -X POST "https://api.coval.dev/v1/test-cases" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "test_set_id": "<coval_test_set_id>",
    "input_str": "<digital_human_intent>",
    "expected_behaviors": ["<success_criteria_line_1>", "<success_criteria_line_2>"],
    "description": "<digital_human_name>: <digital_human_tag>",
    "input_type": "SCENARIO",
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<digital_human_id>"
    }
  }'

对于脚本化测试用例：

bash

curl -s -X POST "https://api.coval.dev/v1/test-cases" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "test_set_id": "<coval_test_set_id>",
    "input_str": "<digital_human_intent>",
    "expected_behaviors": ["<success_criteria>"],
    "input_type": "SCRIPT",
    "simulation_metadata_input": {
      "script_turns": ["<turn_1>", "<turn_2>", "..."]
    },
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<digital_human_id>"
    }
  }'

记录映射关系：

bluejay_digital_human_id → coval_test_case_id

Phase 3: Depends on Phase 1 + 2 (Step 8)

第三阶段：依赖第一、二阶段（步骤8）

Step 8: Migrate Run Templates

步骤8：迁移运行模板

Map Bluejay Simulations → Coval Run Templates.

For each Bluejay simulation, create a Run Template that ties together the migrated resources:

Look up
```
coval_agent_id
```
from
```
bluejay_agent_id
```
(the simulation's agent)
Pick the first persona mapped from a digital human in this simulation, or ask the user
Look up
```
coval_test_set_id
```
from the simulation mapping
Collect
```
coval_metric_ids
```
from the agent's associated metrics

bash

curl -s -X POST "https://api.coval.dev/v1/run-templates" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "display_name": "<simulation_name> Template",
    "description": "Migrated from Bluejay simulation <simulation_id>",
    "agent_id": "<coval_agent_id>",
    "persona_id": "<coval_persona_id>",
    "test_set_id": "<coval_test_set_id>",
    "metric_ids": ["<coval_metric_id_1>", "<coval_metric_id_2>"],
    "iteration_count": 1,
    "concurrency": 5,
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<simulation_id>"
    }
  }'

Record:

bluejay_simulation_id → coval_run_template_id

将Bluejay Simulations映射至Coval Run Templates。

对于每个Bluejay模拟场景，创建一个关联已迁移资源的运行模板：

从
```
bluejay_agent_id
```
（模拟场景关联的Agent）查找
```
coval_agent_id
```
选择该模拟场景中数字人映射的第一个角色，或询问用户
从模拟场景映射中查找
```
coval_test_set_id
```
收集Agent关联指标对应的
```
coval_metric_ids
```

bash

curl -s -X POST "https://api.coval.dev/v1/run-templates" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "display_name": "<simulation_name> Template",
    "description": "从Bluejay模拟场景<simulation_id>迁移而来",
    "agent_id": "<coval_agent_id>",
    "persona_id": "<coval_persona_id>",
    "test_set_id": "<coval_test_set_id>",
    "metric_ids": ["<coval_metric_id_1>", "<coval_metric_id_2>"],
    "iteration_count": 1,
    "concurrency": 5,
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<simulation_id>"
    }
  }'

记录映射关系：

bluejay_simulation_id → coval_run_template_id

Phase 4: Depends on Phase 3 (Step 9)

第四阶段：依赖第三阶段（步骤9）

Step 9: Migrate Scheduled Runs

步骤9：迁移调度任务

Map Bluejay Schedules → Coval Scheduled Runs.

All scheduled runs are created DISABLED for safety. The customer must manually enable them after review.

For each Bluejay schedule:

Look up
```
coval_run_template_id
```
from the simulation mapping

Convert schedule format:

Bluejay
```
cron
```
expression → Coval
```
cron(...)
```
format
Bluejay
```
every_n_minutes
```
→
```
rate(<n> minutes)
```
Bluejay
```
every_n_hours
```
→
```
rate(<n> hours)
```
Bluejay
```
daily
```
at HH:MM →
```
cron(<MM> <HH> ? * * *)
```
Bluejay
```
weekly
```
→
```
cron(<MM> <HH> ? * <DAY> *)
```
Bluejay
```
monthly
```
→
```
cron(<MM> <HH> <DOM> * ? *)
```

bash

curl -s -X POST "https://api.coval.dev/v1/scheduled-runs" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "display_name": "<simulation_name> Schedule",
    "run_template_id": "<coval_run_template_id>",
    "schedule_expression": "<converted_schedule>",
    "enabled": false,
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<schedule_id>"
    }
  }'

Record:

bluejay_schedule_id → coval_scheduled_run_id

将Bluejay Schedules映射至Coval Scheduled Runs。

**出于安全考虑，所有创建的调度任务默认处于禁用状态。**客户需在审核后手动启用。

对于每个Bluejay调度任务：

从模拟场景映射中查找
```
coval_run_template_id
```

转换调度格式：

Bluejay
```
cron
```
表达式 → Coval
```
cron(...)
```
格式
Bluejay
```
every_n_minutes
```
→
```
rate(<n> minutes)
```
Bluejay
```
every_n_hours
```
→
```
rate(<n> hours)
```
Bluejay
```
daily
```
于HH:MM →
```
cron(<MM> <HH> ? * * *)
```
Bluejay
```
weekly
```
→
```
cron(<MM> <HH> ? * <DAY> *)
```
Bluejay
```
monthly
```
→
```
cron(<MM> <HH> <DOM> * ? *)
```

bash

curl -s -X POST "https://api.coval.dev/v1/scheduled-runs" \
  -H "X-API-Key: <COVAL_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "display_name": "<simulation_name> Schedule",
    "run_template_id": "<coval_run_template_id>",
    "schedule_expression": "<converted_schedule>",
    "enabled": false,
    "metadata": {
      "bluejay_migration": true,
      "bluejay_original_id": "<schedule_id>"
    }
  }'

记录映射关系：

bluejay_schedule_id → coval_scheduled_run_id

Step 10: Migration Report

步骤10：迁移报告

Print a final summary:

undefined

打印最终汇总：

undefined

Migration Complete

迁移完成

Created Entities

创建的实体

Type	Count	Details
Metrics	N	bluejay_id → coval_id for each
Personas	N	...
Agents	N	...
Test Sets	N	...
Test Cases	N	...
Run Templates	N	...
Scheduled Runs	N	(all disabled)

类型	数量	详情
Metrics	N	每个实体的bluejay_id → coval_id映射
Personas	N	...
Agents	N	...
Test Sets	N	...
Test Cases	N	...
Run Templates	N	...
Scheduled Runs	N	（全部禁用）

Skipped (Already Existed)

已跳过（已存在）

List any entities skipped due to idempotency

列出因幂等性而跳过的实体

Warnings

警告

List any ambiguous mappings or approximations made
Note any
```
json
```
metrics mapped to METRIC_LLM_BINARY
Note any voice mappings that may need review

列出所有模糊映射或近似处理的内容
记录所有映射为METRIC_LLM_BINARY的json类型指标
记录所有需审核的语音映射

Next Steps

后续步骤

Review migrated entities in the Coval dashboard: https://app.coval.dev
Enable scheduled runs after verifying configuration
Run a test evaluation from a migrated Run Template to verify
Review persona voice assignments and adjust if needed

undefined

在Coval控制台查看已迁移实体：https://app.coval.dev
验证配置后启用调度任务
运行一个迁移后的运行模板进行测试验证
审核角色语音分配并按需调整

undefined

Error Handling

错误处理

HTTP 429 (Rate Limited): Wait 5 seconds and retry. Max 3 retries per request.
HTTP 400 (Bad Request): Show the error details to the user. Ask whether to fix the input and retry, or skip this entity.
HTTP 401/403 (Auth Error): Stop and ask user to verify their API key.
HTTP 500 (Server Error): Show error, skip entity, continue with others.
Network Error: Retry once after 3 seconds. If still failing, stop and report what was created so far.

After any failure, always show what has been successfully created so far so the user can resume.

HTTP 429（请求限流）：等待5秒后重试，每个请求最多重试3次。
HTTP 400（请求错误）：向用户展示错误详情，询问是否修正输入后重试，或跳过该实体。
HTTP 401/403（鉴权错误）：停止操作并要求用户验证API密钥。
HTTP 500（服务器错误）：展示错误信息，跳过该实体，继续处理其他内容。
网络错误：等待3秒后重试一次。若仍失败，停止操作并报告已成功创建的内容。

无论发生何种失败，均需向用户展示已成功创建的内容，以便用户可以恢复迁移。

Idempotency Rules

幂等性规则

Before creating any entity, check the name lookup from Step 2
If entity exists with
```
bluejay_migration: true
```
in metadata → skip silently
If name collision with non-migrated entity → ask user: "A Coval <entity_type> named '<name>' already exists but wasn't from a Bluejay migration. Skip, rename, or overwrite?"

All created entities get metadata:

{ "bluejay_migration": true, "bluejay_original_id": "<id>" }

Re-running the migration should produce zero new entities if nothing changed

创建任何实体前，检查步骤2中的名称查找映射
若实体元数据中存在
```
bluejay_migration: true
```
→ 静默跳过
若与非迁移实体重名 → 询问用户：“名为'<name>'的Coval <entity_type>已存在，但并非来自Bluejay迁移。是否跳过、重命名或覆盖？”

所有创建的实体均添加元数据：

{ "bluejay_migration": true, "bluejay_original_id": "<id>" }

若未发生任何变更，重新运行迁移不应创建新实体