Architecture

Raiken is built as an Nx monorepo with a clear separation between the CLI, dashboard, and core logic.

Modules

Module	Location	Responsibility
CLI	apps/cli	Fastify server, SSE output, tRPC API, discover/auth/init command surface
Dashboard	apps/dashboard	React + Vite UI for test generation and discovery runtime monitoring
Core	libs/core	LangGraph agent, site discovery runtime, browser session, DB, testing
Shared	libs/shared	tRPC router, discovery runtime orchestration, shared types
Playground (internal)	tools/playground	Route-rich test surface for discovery and generated E2E validation

Data flow

1. User submits a prompt or starts discovery from the dashboard.
2. CLI receives the request and streams runtime output via SSE/tRPC.
3. Shared router coordinates runtime phase, events, and active jobs.
4. Core executes LangGraph tools or discovery crawler operations.
5. Results and session state are persisted locally in .raiken/raiken.db.

Local persistence

Raiken keeps local state under .raiken/ so it can:

• store the code graph, keywords, embeddings, and memory
• cache analysis for faster follow-up prompts
• track test run history and outcomes
• persist discovery sessions and resume metadata

Ports

• CLI server + dashboard: http://localhost:7101

Discovery runtime

Discovery now operates as a stateful subsystem with explicit runtime phases: idle, running, paused, completed, and error.

Runtime state is exposed to the dashboard through shared tRPC endpoints for:

• start/continue operations
• runtime state and event timeline
• auth-assist and cleanup flows

Recent persistence changes include discovery session fields such as max_pages and max_depth to support predictable continuation behavior across pause/resume runs.

See Site Discovery for implementation details.