Architecture: How to make agents human-readable with persistent identity a...
Documentation for Sessions
🏗️ Architecture: How to make agents human-readable with persistent identity a...
Date: 2026-02-05
Time: 18:10:20 UTC
Session: editor2-92.5
Agent: backend
Confidence: 90%
Status: ✅ Documented
📋 Decision Context
Question:
How to make agents human-readable with persistent identity across sessions?
Task: Phase 1 - Agent Identity System
🎯 Decision Made
Chosen:
Added displayName field (auto-generated from platform+OS, manually overridable), agentSessionHistory array (last 20 sessions with decision counts), and WebSocket events agent.setName + agent.sessionHistory
Options Considered:
- No alternatives documented
💭 Reasoning
Machine-generated agentIds like 'vscode-darwin-ml9lt1wo' are meaningless to humans. displayName auto-generates as 'VSCode (macOS)' on first registration and can be overridden. Session history tracks which sessions each agent worked on and how many decisions it logged, enabling Vitto to understand agent expertise and activity patterns.
📁 Files Involved
apis/backend-ai-api/src/app/schemas/agent-registry.schema.tsapis/backend-ai-api/src/app/infrastructure/services/agent.service.tsapis/backend-ai-api/src/app/mcp-gateway/native-ws.gateway.tsapis/backend-ai-api/src/app/mcp-gateway/mcp-gateway.controller.tsapps/vscode-gograb-ai/src/panels/webview/types.tsapps/vscode-gograb-ai/src/panels/webview/components/agents/agents-view.ts
🔗 Related
- Decision ID:
dev_1770315018568_6xpvs41 - Session: 92
- Workspace: editor2
Auto-generated by MCP Smart Consolidation