AI Center#

Path: /interaction

The AI center manages AI interactions, agent runs and session history.

Operations Dashboard#

Path: /interaction

The dashboard shows: total interaction count, agent usage distribution, average response latency, token consumption, cost statistics.

Interaction statistics:

Metric	Description
total_interactions	Total interactions
by_status	By status
by_agent	By agent
by_intent	By intent
avg_latency_ms	Average latency
total_tokens	Total tokens
total_cost_usd	Total cost (USD)

AI Assistant#

Path: /interaction/copilot

The full-page version of the AI Copilot chat — same features as the sidebar Copilot panel with more space.

Session management:

• Clear conversation: clicking the top "Clear Conversation" button deletes the current backend session (not just the screen) — history and AI action to-dos tied to the session are cleaned up and no longer appear in the task inbox; if AI is mid-response, output is aborted first, then deleted; on failure it shows "Clear failed, please retry later" and local messages are kept — you can retry
• To-do navigation: clicking an AI action to-do in the task inbox jumps to this page and locates the original session (URL carries the conversationId parameter), letting you review the original context

Agents#

Path: /interaction/agents

Manage the 7 specialized AI agents:

Agent ID	Name	Color	Description
chat_agent	General Chat	Blue #1677ff	General Q&A and chat
assessment_agent	Assessment Analysis	Purple #722ed1	Assessment-related queries
decision_agent	Decision Support	Cyan #13c2c2	Decision analysis support
planning_agent	Planning	Pink #eb2f96	Planning and roadmaps
delivery_agent	Delivery Management	Orange #fa8c16	Project and task management
knowledge_agent	Knowledge Retrieval	Green #52c41a	Knowledge search and Q&A
operations_agent	AI Operations Agent	Rose #eb2f96	App monitoring/evaluation/cost/feedback/optimization analysis

Automatic Agent Routing#

The system switches agents automatically based on the current page:

Page prefix	Routed to
/assessment	assessment_agent
/decision	decision_agent
/planning	planning_agent
/delivery	delivery_agent
/knowledge	knowledge_agent
/aiops	operations_agent
Others	chat_agent

Agent execution states:

Intent Routing Methods#

Method	Description
rule	Rule matching
keyword	Keyword matching
regex	Regex matching
context	Context inference
llm	LLM intent recognition

Intent classification:

Intent	Routed to
assessment_query	assessment_agent
decision_query	decision_agent
planning_query	planning_agent
delivery_query	delivery_agent
knowledge_query	knowledge_agent
task_execute	chat_agent
general_chat	chat_agent
data_query	chat_agent
help	chat_agent
feedback	chat_agent

Keyword routing rules:

• Contains "评估/就绪度/成熟度/assessment/readiness" → assessment_agent
• Contains "决策/场景/方案选择/decision/scenario" → decision_agent
• Contains "规划/路线图/计划/planning/roadmap/plan" → planning_agent
• Contains "交付/项目/任务/sprint/delivery/task/project" → delivery_agent
• Contains "知识/文档/搜索/knowledge/document/search/asset" → knowledge_agent
• Contains "执行/操作/创建/删除/更新" → chat_agent

Session History#

Path: /interaction/sessions

Browse the history of all AI chat sessions.

Session attributes:

Attribute	Values
Channel	web / api / wecom / dingtalk / feishu / clawbot / telegram
Type	chat / assessment / decision / planning / delivery / knowledge / system
Context scope	global / project / scenario / plan / delivery_project
Status	active / archived / deleted

Interaction states:

Steps:

1. Browse the session list
2. Click a session to view message details
3. Filter by channel/type/status
4. Archive or delete old sessions

Message content types:

Type	Description
text	Plain text
markdown	Markdown
json	JSON data
html	HTML content
stream_chunk	Streaming chunk
action_result	Action result
error	Error message

For the sidebar AI chat panel, see AI Copilot.