add
This commit is contained in:
V-LiuShuang
@@ -1,27 +0,0 @@
|
||||
---
|
||||
name: mis-database-ops
|
||||
description: 生成SQL的参考文档。
|
||||
---
|
||||
|
||||
# 能力索引
|
||||
|
||||
| 能力名称 | 触发条件 |
|
||||
| --- | --- |
|
||||
| [查询智能体对话日志](./references/query-agent-chat-log.md) | 明确要求查智能体的对话日志 |
|
||||
| [查询智能体信息](./references/query-agent-info.md) | 明确要求查智能体信息 |
|
||||
| [查询智能体使用数据](./references/query-agent-usage.md) | 明确要求查智能体使用数据 |
|
||||
| [查询项目统计数据](./references/query-project-stats.md) | 明确要求查项目统计数据 |
|
||||
| [新增devco账号](./references/add-devco-accout.md) | 当用户提到新增devco账号的时候 |
|
||||
| [查询paw-agent信息](./references/query-user-paw-agent.md) | 当用户提到查询paw-agent信息 |
|
||||
|
||||
## 工作流程
|
||||
|
||||
```mermaid
|
||||
flowchart LR
|
||||
A[开始] --> B[意图识别]
|
||||
B --> C{请求是否被能力索引覆盖？}
|
||||
C -->|是| D[参考文档执行相关操作]
|
||||
C -->|否| E[利用相关工具想办法解决]
|
||||
D --> F[结束]
|
||||
E --> F
|
||||
```
|
||||
@@ -1,18 +0,0 @@
|
||||
# 能力说明
|
||||
|
||||
在`mis-dev`资源下的`mis-cloud-paw`库中执行批量插入sql，用户的输入格式如下。
|
||||
|
||||
```text
|
||||
devco
|
||||
username1
|
||||
username2
|
||||
username3
|
||||
```
|
||||
在用户的输入中，`devco`是触发d当前技能的关键词。从第二行开始，去掉空白行以后，每一个都是用户名。如果用户提供的用户名存在重复，请忽略大小写保留1个即可。`client_id`和`state`字段是固定值，用户名对应的是`account`字段，生成一条下面这样的sql执行即可。
|
||||
|
||||
```sql
|
||||
INSERT INTO external_login_client_account (client_id, account, state) VALUES
|
||||
('226c18408b5705449f37b621a432a6d3', 'username1', 1),
|
||||
('226c18408b5705449f37b621a432a6d3', 'username2', 1),
|
||||
('226c18408b5705449f37b621a432a6d3', 'username3', 1);
|
||||
```
|
||||
@@ -1,19 +0,0 @@
|
||||
# 输入要求
|
||||
|
||||
必须要求用户提供**会话ID**。
|
||||
|
||||
## 工作流程
|
||||
|
||||
1. 查询所有切片号，假设查询出来的`part`是[1,2]
|
||||
|
||||
```sql
|
||||
select part from log_agent_chat where del_flag = '0';
|
||||
```
|
||||
|
||||
2. 根据切片号查询并合并各个切片表的数据
|
||||
|
||||
```sql
|
||||
select * from log_agent_chat_1 where chat_window_id = <会话ID>
|
||||
UNION ALL
|
||||
select * from log_agent_chat_2 where chat_window_id = <会话ID>
|
||||
```
|
||||
@@ -1,63 +0,0 @@
|
||||
# 输入要求
|
||||
|
||||
必须提供**智能体ID**，在不明确用户查询的是单智能体还是多智能体时，友好的询问要查询哪种智能体。
|
||||
|
||||
## 查询单智能体信息
|
||||
|
||||
```sql
|
||||
with agent_info AS (
|
||||
select
|
||||
agent_id,
|
||||
agent_name,
|
||||
system_prompt,
|
||||
`description`,
|
||||
max_messages,
|
||||
max_tokens,
|
||||
kb_uuids,
|
||||
kb_configs,
|
||||
kb_similarity_threshold,
|
||||
kb_top_k,
|
||||
kb_max_recall_toekn,
|
||||
llm_name,
|
||||
long_term_memory,
|
||||
case kb_mode when '1' then '直接检索' when '2' then 'tool检索' else '' end AS kb_mode,
|
||||
case kb_search_type when '0' then '向量检索' when '1' then '混合检索' when '2' then '智能检索' else '' end AS kb_search_type
|
||||
from mis-agent.agent
|
||||
where agent_id = <智能体ID>
|
||||
),
|
||||
agent_skill AS (
|
||||
select
|
||||
<智能体ID> AS agent_id,
|
||||
GROUP_CONCAT(skill_name) AS skill_names
|
||||
from mis-mcp.skill
|
||||
where skill_id in (
|
||||
select skill_id from mis-agent.agent_skill_mapping
|
||||
where agent_id = <智能体ID>
|
||||
)
|
||||
),
|
||||
agent_mcp_tool AS (
|
||||
select
|
||||
<智能体ID> AS agent_id,
|
||||
GROUP_CONCAT(tool_name) AS tool_names
|
||||
from mis-mcp.map_tools
|
||||
where mcp_server_uuid in (
|
||||
select mcp_server_uuid from mis-agent.agent_tool_mapping
|
||||
where relate_agent_id = <智能体ID>
|
||||
)
|
||||
)
|
||||
|
||||
select * from agent_info AS a
|
||||
left join agent_skill AS b on a.agent_id = b.agent_id
|
||||
left join agent_mcp_tool AS c on a.agent_id = c.agent_id
|
||||
```
|
||||
|
||||
## 查询多智能体信息
|
||||
|
||||
```sql
|
||||
select * from mis-agent.agent_release_info where release_id in (
|
||||
select
|
||||
release_id
|
||||
from mis-agent.multi_agent_sub
|
||||
where main_agent_id = <智能体ID>
|
||||
)
|
||||
```
|
||||
@@ -1,9 +0,0 @@
|
||||
# 输入要求
|
||||
|
||||
必须提供**智能体ID**。
|
||||
|
||||
## 执行查询
|
||||
|
||||
```sql
|
||||
select * from mis-agent.agent_usage_stats_daily where agent_id = <智能体ID> order by daily desc
|
||||
```
|
||||
@@ -1,17 +0,0 @@
|
||||
# 输入要求
|
||||
|
||||
必须提供**项目名称**。
|
||||
|
||||
## 工作流程
|
||||
|
||||
1. 查询项目注册表，判断是否包含该项目。用户提供的项目名称大概率是模糊的，比如"认证"、"ebiz"，要根据语义判断究竟是哪一个，如果猜不准就询问一下用户是哪个。重要：`project_stats_registry`里的`id`字段，称为项目ID，或者叫project_id，后面的步骤需要用到。
|
||||
|
||||
```sql
|
||||
select * from mis-agent.project_stats_registry
|
||||
```
|
||||
|
||||
2. 查询项目统计数据
|
||||
|
||||
```sql
|
||||
select * from mis-agent.project_data_stats_daily where project_id = <project_id>
|
||||
```
|
||||
@@ -1,29 +0,0 @@
|
||||
# 能力说明
|
||||
|
||||
查看用户上传的`skill`审批详情数据。
|
||||
|
||||
# 工作流程
|
||||
|
||||
如果用户没有提供**流程单号**则先询问用户，然后在`mis-dev`资源下的`mis-workflow-claw`库中执行如下sql，把`{process_id}`替换成用户提供的流程单号。
|
||||
|
||||
```sql
|
||||
SELECT
|
||||
ht.ID_ AS task_id,
|
||||
ht.NAME_ AS task_name,
|
||||
ht.TASK_DEF_KEY_ AS task_definition_key,
|
||||
ht.ASSIGNEE_ AS assignee,
|
||||
ht.START_TIME_ AS start_time,
|
||||
ht.END_TIME_ AS end_time,
|
||||
c.TYPE_ AS comment_type,
|
||||
c.MESSAGE_ AS comment_msg,
|
||||
h.PROC_INST_ID_ AS process_instance_id,
|
||||
h.BUSINESS_KEY_ AS business_key
|
||||
FROM ACT_HI_TASKINST ht
|
||||
INNER JOIN ACT_HI_PROCINST h ON h.PROC_INST_ID_ = ht.PROC_INST_ID_
|
||||
LEFT JOIN ACT_HI_COMMENT c
|
||||
ON c.TASK_ID_ = ht.ID_
|
||||
AND c.TYPE_ IN ('pass', 'back')
|
||||
WHERE h.BUSINESS_KEY_ = '{process_id}'
|
||||
AND (ht.PARENT_TASK_ID_ IS NULL OR ht.PARENT_TASK_ID_ = '')
|
||||
ORDER BY ht.START_TIME_ ASC, ht.ID_ ASC;
|
||||
```
|
||||
@@ -1,17 +0,0 @@
|
||||
# 能力说明
|
||||
|
||||
在`mis-dev`资源下，根据用户名（username）查询paw-agent信息，如果用户没有提供username，友好的提问要查哪个用户的，一个可用的username示例：`Agnes2.wang`。
|
||||
|
||||
## 工作流程
|
||||
|
||||
1. 根据用户提供的`username`查询`user_id`，只会查出来1行数据，示例：
|
||||
|
||||
```sql
|
||||
select user_id from `mis-cloud-claw`.`sys_user` where user_name = '{username}';
|
||||
```
|
||||
|
||||
2. 根据`user_id`查询agent信息，可能返回多行数据，使用markdown表格的形式展示每一行的数据，示例：
|
||||
|
||||
```sql
|
||||
select agent_id, agent_mode, publish_status, publish_scope, published_at, version_lane, key_scope_default from `mis-paw-claw`.`agent` where create_by = '1889506629046902786';
|
||||
```
|
||||
Reference in New Issue
Block a user