App Generation Handoff
ValorIDE completes the loop after App Generation produces an application. The web app can generate schemas, APIs, components, workflows, and deployments; ValorIDE then opens the generated code as a workspace for agentic refinement.
Handoff Inputs
| Input | Source | ValorIDE use |
|---|---|---|
| OpenAPI spec | OpenAPI Design Center / ThorAPI | Understand domain models, endpoints, generated clients. |
| Generated TypeScript client | ThorAPI | Call server APIs from React components and tests. |
| Generated components | App Generation component library | Customize UI without losing schema alignment. |
| Workflow definitions | ValkyrAI | Add automation, MCP endpoints, app-specific operations. |
| IntegrationAccounts | ValkyrAI / Valor | Configure external systems without leaking secrets. |
| Project record | ValkyrAI / GrayMatter | Track strategy, goals, tasks, generated apps, deployment follow-up, and MCP publication. |
| GrayMatter memory | GrayMatter | Keep durable project context, decisions, and preferences. |
| SWARM presence | SwarmOps | Coordinate with other active agents. |
Publish-Back Loop
Practical Workflow
- Generate or select an application in the web app.
- Download/open the generated project in ValorIDE.
- Let ValorIDE load the Project context from GrayMatter/ValkyrAI when available.
- Ask ValorIDE to inspect the generated clients and component surfaces.
- Refine behavior, UI, tests, and local automation.
- Use MCP/workflow tools when app behavior depends on server-side workflows.
- Publish back to the hosted instance or deployment target.
ValorIDE can create or update Project records through the GrayMatter/ValkyrAI client surface with sourceSurface: "valoride". That keeps agentic coding work connected to the same Project visible in the LCARS dashboard and to the same GrayMatter project memory used by other agents.
Model Providers
ValorIDE should expose model choice as a coding-agent concern: OpenAI, Claude, Ollama/open source models, or enterprise provider routing. The selected model changes how the agent reasons and edits code; it should not change RBAC or secret visibility.