Desktop app
Dette innholdet er ikke tilgjengelig på ditt språk.
The desktop app is Khadim’s local automation workspace. It is built with Tauri and React, backed by SQLite, and designed for tasks that need screen access, local files, credentials, Docker runners, and native process control.
Current state
Section titled “Current state”The desktop app currently includes these foundations:
- SQLite persistence for workspaces, conversations, messages, and settings.
- AI engine adapters for OpenCode, Claude Code, and the native Khadim orchestrator.
- Normalized streaming events such as
text_delta,step_start,step_update,step_complete,question,done, anderror. - Docker runner scaffolding with environment and secret injection.
- Git operations for repository validation, branch listing, status, and diff statistics.
- Process spawning, health checks, and cleanup.
- React UI surfaces for chat, approvals, questions, and folder picking.
RPA direction
Section titled “RPA direction”The desktop app is pivoting from a developer coding tool to a local-first RPA platform.
Planned domain work includes:
- Screenshot capture with
xcap. - OCR with Tesseract.
- Mouse and keyboard simulation with
enigo. - Browser automation with
chromiumoxide. - Email, spreadsheet, HTTP, and file connectors.
- Environment, credential, memory, schedule, connector, automation, and session tables.
Managed agents
Section titled “Managed agents”Managed agents are persistent automation personas. Each agent can carry:
- Instructions and approval mode.
- Tool and connector access.
- Trigger and schedule configuration.
- Runner settings for local, Docker, or cloud execution.
- Memory stores shared across sessions.
Run the desktop app
Section titled “Run the desktop app”From the repository root, run:
bun run desktop:devBuild the desktop app:
bun run desktop:buildRelated pages
Section titled “Related pages”Read Architecture for the shared engine and Tools and domains for the planned RPA domain model.