Update SPEAR architecture in /doc/design folder (#3)

docs/design/spear-architecture.md

@@ -0,0 +1,30 @@
+# SPEAR high-level design
+
+## Design principles
+SPEAR's first design principle is to help developers and end-users an easier approach to develop and deploy an AI Agent in Cloud-Edge integrated manner.
+
+### Scalable
+SPEAR aim to provide scalable agent service with the help of auto-scaling.
+
+### Performant
+SPEAR aim to serve Agent requests performant, with the help of prudently designed APIs and online scheduling system.
+
+## Workflow
+
+<div align="center">
+<img width="500px" src="./img/spear-arch.png"></img>
+</div>
+
+### Offline Phase
+- Input: Code of existing agent framework (e.g., LangChain)
+- Output: Deployment instance and configuration
+- Key technology
+  - Multi-instance selection
+  - Automatic code generation
+
+### Online Phase
+- Input: Generated instance of offline phase
+- Output: Scalable and performant user request serving
+- Key technology
+  - Cloud-edge-end integrated offloading
+  - Agent-oriented auto-scaling