How akilanpl/rlb-relaxlemmebuild- works
RLB is a multi-agent software-engineering workspace: a user submits a task, repository analysis informs planning, and approval gates govern isolated coding, diff review, testing, and final review before promotion. The persistent workflow engine and durable queue coordinate the stages rather than leaving control to the models. The README documents the broader product loop; sampled source confirms key API, analysis, reviewer, queue, and sandbox behavior, but not every service-to-service connection, so uncertain links are omitted.
Main entry points: engine.py Workflow engine · tasks.py Task API · service.py Analysis API · gateway.py AI gateway
Components 21
Workspace interface
- Workspace UI
frontend/src/app/dashboard/workspaces/[workspaceId]/page.tsx - Review panels
frontend/src/components/workspace/CodeProposalReview.tsx
Workflow control
- Task API
backend/app/api/v1/tasks.py - Workflow engine
backend/app/workflow/engine.py - Durable queue
backend/app/services/job_queue.py - Task worker
backend/app/services/durable_worker.py
Engineering stages
- Planner
backend/app/services/planner_service.py - Coder
backend/app/services/coder_service.py - Test architect
backend/app/services/testing_service.py - Reviewer
backend/app/services/reviewer_service.py
Execution infrastructure
- Staging and promotion
backend/app/services/staging_service.py - Sandbox
backend/app/sandbox/daytona.py - Model providers
backend/app/providers/resolver.py - AI gateway
backend/app/ai/gateway.py - Workspace service
backend/app/services/workspace_service.py - Artifact storage
backend/app/storage/base.py - Persistent repositories
backend/app/repositories/task.py
Repository intelligence
- Analysis API
backend/app/analysis/service.py - Codebase analysis
backend/app/analysis/indexer.py
Other components
- Engineer
- Model service
Connections 26
- Engineer → Workspace UI: submits request
- Workspace UI → Task API: calls task API
- Review panels → Task API: submits decisions
- Task API → Workflow engine: creates and transitions
- Task API → Durable queue: enqueues task
- Durable queue → Task worker: leases jobs
- Task worker → Workflow engine: runs workflow
- Workflow engine → Analysis API: requests analysis
- Analysis API → Codebase analysis: indexes and analyzes
- Analysis API → Workspace service: loads workspace
- Analysis API → Artifact storage: reads files
- Workflow engine → Planner: dispatches planning
- Workflow engine → Coder: dispatches coding
- Workflow engine → Test architect: dispatches testing
- Task API → Reviewer: runs review
- Reviewer → AI gateway: requests review
- Planner → AI gateway: requests plan
- Coder → AI gateway: requests changes
- Test architect → Sandbox: executes tests
- Coder → Staging and promotion: creates proposal
- Staging and promotion → Sandbox: isolates execution
- Staging and promotion → Artifact storage: stores artifacts
- AI gateway → Model providers: resolves model
- Model providers → Model service: calls model
- Workflow engine → Persistent repositories: persists task state
- Reviewer → Persistent repositories: persists report