{
  "id": "page:docs-reference-repos-sero-labs-sero-sero-gap-analysis",
  "_kind": "Page",
  "_file": "wiki/docs/reference-repos/sero-labs/sero/sero-gap-analysis.md",
  "_cluster": "wiki",
  "attributes": {
    "nodeKind": "Page",
    "sourcePath": "docs/reference-repos/sero-labs/sero/sero-gap-analysis.md",
    "sourceKind": "repo-docs",
    "title": "Sero Agent OS — Gap Analysis vs Babysitter Stack",
    "displayName": "Sero Agent OS — Gap Analysis vs Babysitter Stack",
    "slug": "docs/reference-repos/sero-labs/sero/sero-gap-analysis",
    "articlePath": "wiki/docs/reference-repos/sero-labs/sero/sero-gap-analysis.md",
    "article": "\n# Sero Agent OS — Gap Analysis vs Babysitter Stack\n\n## What Sero Is\n\nSero is a **personal agent operating system** (Agent OS) — an Electron desktop app built on the Pi coding agent framework. It consolidates chat, terminals, file explorer, visual browser, and plugins into a unified local-first workspace where agents maintain persistent awareness across sessions.\n\n## Feature Mapping\n\n| Sero Feature | Our Status | Our Implementation | Gap |\n|---|---|---|---|\n| **Unified Desktop Workspace** | PRESENT | genty/desktop-app (Electron), genty/web-app (Vite), genty/tui (Ink), mobile/TV/watch apps | None — we have MORE platforms |\n| **Visual Browser / Screenshots** | PARTIAL | genty/core agenticTools/browser (Puppeteer-based screenshots) | Missing: screen recording, video clips, Playwright integration |\n| **Project Memory & Persistence** | PRESENT | sdk/profiles (user+project), sdk/session (state+transcript), platform/daemon (cron scheduler) | None — we have equivalent or richer persistence |\n| **Self-Extending Plugin System** | PRESENT | platform/plugins (sandbox, loader, permissions), platform/skills (router, discovery, chaining) | None — our plugin system is more advanced (permissions, sandboxing, marketplace) |\n| **Runtime Isolation** | PRESENT | Docker, SSH, K8s transports, secure-sandbox, governance layer | None — we have MORE isolation modes |\n| **Built-in Specialists / Subagents** | PRESENT | genty/core/subagent (3 invocation modes), .claude/agents/ (markdown specs), AgentPersona CRDs | None — our subagent system is richer |\n| **Multi-Provider LLM** | PRESENT | 25+ providers via adapters/core, transport-adapter, provider translation tables | None — we support far more providers |\n| **Agent Orchestration** | PRESENT | babysitter-sdk (orchestrate, iterate, effects, journal), genty/core/loop, genty/platform/harness | None — our orchestration is more sophisticated (process definitions, effects, breakpoints) |\n| **MCP Integration** | PRESENT | genty/platform/mcp (client, channels, transport, executor, tool registry) | None — full MCP support |\n| **Evaluation Framework** | PARTIAL | Vitest tests, E2E pipeline tests, contract tests, live-stack tests | Missing: Promptfoo-style evaluation DSL, LLM benchmark harness, continuous quality scoring |\n| **Module Federation** | MISSING | — | No dynamic module federation for live plugin loading in desktop/web app |\n| **Screen Recording** | MISSING | — | No video/clip recording capability |\n\n## What We're Missing (Gaps)\n\n### Gap 1: Screen Recording for Agents\n**Sero**: Agents can record short video clips of running applications.\n**Us**: Only Puppeteer screenshots. No video recording.\n**Fix**: Add screen recording to the browser agentic tool — use Puppeteer's `page.screencast()` API or integrate `ffmpeg` for recording.\n**Effort**: Small — Puppeteer already supports screencast.\n\n### Gap 2: Promptfoo-Style Evaluation Framework\n**Sero**: Includes promptfoo-based evaluation framework in `eval/` directory.\n**Us**: Solid test infrastructure (vitest, contract tests, live-stack E2E) but no evaluation-as-a-service framework for scoring agent responses, comparing prompts, or continuous quality benchmarking.\n**Fix**: Add an `eval/` package with promptfoo integration. Define evaluation suites for:\n- Agent response quality scoring\n- Prompt comparison (A/B testing)\n- Regression detection across model versions\n- Task completion benchmarks\n**Effort**: Medium — framework setup + evaluation suite authoring.\n\n### Gap 3: Module Federation for Live Plugin Loading\n**Sero**: Uses Webpack Module Federation to dynamically load plugins at runtime without app restart.\n**Us**: Plugin loader exists but uses traditional dynamic import, not Module Federation. Desktop and web apps would need to restart or rebuild to pick up new plugins.\n**Fix**: Add Module Federation to genty/desktop-app and genty/web-app Vite configs. Use `@module-federation/vite` to expose/consume plugin containers dynamically.\n**Effort**: Medium — Vite Module Federation plugin is available but requires plugin architecture adaptation.\n\n### Gap 4: Unified Desktop Experience Polish\n**Sero**: Polished single-pane Electron app with integrated chat + terminal + file explorer + browser in one window.\n**Us**: Desktop app exists but is a scaffold with page-based routing (Home, Sessions, Agents, Kanban). Not yet a unified workspace with simultaneous chat + terminal + file explorer panes.\n**Fix**: Redesign genty/desktop-app layout to support split-pane views:\n- Left: File explorer + project tree\n- Center: Chat / session conversation\n- Right: Terminal / tool output\n- Bottom: Activity feed / notifications\nUse a layout manager (like `react-mosaic` or `allotment`) for resizable panes.\n**Effort**: Large — significant UI architecture work.\n\n## What We Have That Sero Doesn't\n\n| Our Feature | Sero Equivalent |\n|---|---|\n| **Kubernetes-native forge (Kradle)** | Nothing — Sero has no K8s/CRD/infrastructure management |\n| **89+ CRD resource types** | Nothing — Sero operates purely locally |\n| **Process definitions + babysitter orchestration** | Nothing — Sero uses Pi's simpler agent loop |\n| **Multi-agent meeting (Jitsi integration)** | Nothing — no video conferencing or multi-agent meetings |\n| **Agent identity system (Persona, Soul, Appearance, Voice)** | Basic specialist agents — no rich identity model |\n| **Atlas knowledge graph** | Nothing — no structured knowledge graph |\n| **7 platform surfaces (desktop, web, TUI, mobile×2, TV×2, watch×2)** | Only desktop (Electron) |\n| **Breakpoints + approval gates** | Nothing — no human-in-the-loop orchestration gates |\n| **Agent marketplace / process library** | Plugin system but no marketplace |\n| **CI/CD integration (GitHub Actions, workflows)** | Nothing — local-only |\n| **Governance layer (authority chains, mandates)** | Basic plugin permissions only |\n| **Transport-adapter (multi-provider proxy routing)** | Basic provider support |\n| **Hooks-adapter (cross-harness hook normalization)** | Plugin hooks only |\n\n## Priority Fixes\n\n| Priority | Gap | Effort | Impact |\n|---|---|---|---|\n| P1 | Unified desktop workspace (split-pane layout) | Large | High — this is Sero's core UX differentiator |\n| P2 | Evaluation framework (promptfoo) | Medium | High — quality assurance for agent responses |\n| P3 | Screen recording | Small | Medium — agents can provide richer visual feedback |\n| P4 | Module Federation for live plugins | Medium | Medium — smoother plugin development experience |\n",
    "documents": []
  },
  "outgoingEdges": [],
  "incomingEdges": [
    {
      "from": "page:docs-reference-repos",
      "to": "page:docs-reference-repos-sero-labs-sero-sero-gap-analysis",
      "kind": "contains_page"
    }
  ]
}