Published at: 2026-09-17
AI PaaS core terms
| Term | Description |
|---|---|
| Agent | An AI program that runs tasks autonomously in a secure cloud sandbox. It uses a “plan-execute-evaluate” Agent Loop architecture |
| Agent Studio | The Setup console for ShareCRM Agents. It lets you visually and with low code configure, manage, and publish enterprise agents |
| ShareAgent | The unified conversational workspace (main Agent). It discovers and invokes enterprise SubAgents |
| SubAgent | An internal enterprise Agent that ShareAgent discovers and invokes as a child agent |
| Skill | A capability unit that an Agent can mount. The enterprise skill library manages Skills. Agents discover and use Skills to handle business requests |
| Prompt template | A reusable prompt that contains placeholder variables for customer, product, and other details. It combines CRM data to generate personalized prompts |
| Semantic Retrieval Index | A “smart catalog system” that structures enterprise private-domain knowledge into a searchable index. It provides precise recall for RAG |
| RAG | Retrieval-Augmented Generation. It first queries the knowledge index to find relevant chunks, then generates accurate answers based on those chunks |
| Credit | The unified billing unit for AI PaaS. 1 credit = 0.10 RMB |
| Context | Environmental information that an Agent automatically senses while running a task. It is injected into the Agent System Prompt as key-value pairs |
| Event log | The core entry point for Agent troubleshooting, tuning, and auditing. It records the complete run of each user query |
| Prediction model | A binary classification model trained on business data. It predicts business outcomes such as opportunity win or loss |