Constraint-based workload scheduling agent with what-if scenario analysis for operations planning.
Active development begins Week 25. This scaffold establishes the repository structure, documentation standards, and evaluation targets that will guide the Phase 4 build.
workload-scheduler is a Planner-Executor agent that generates and optimizes work schedules under real-world operational constraints — technician availability, equipment due dates, instrument capacity, and regulatory priorities. It supports what-if scenario analysis so planners can evaluate "what happens if I add three urgent jobs this week" without committing to the change.
The agent first plans — generating a candidate schedule that satisfies all hard constraints. It then executes iterative optimization passes, evaluating alternatives and selecting the highest-scoring schedule. What-if queries run the same pipeline against a modified constraint set without touching the live schedule.
| Metric | Target |
|---|---|
| Constraint Satisfaction | 100% (no violations in generated schedules) |
| Due Date Coverage | ≥90% of overdue and near-due jobs scheduled |
| Resource Load | No resource exceeds 110% capacity |
| What-If Query Time | <15s per scenario |
| Agent | Pattern | Repo |
|---|---|---|
| knowledge-navigator | RAG Agent | link |
| intake-agent | Structured Output + Tool Use | link |
| compliance-checker | Multi-Agent Pipeline | link |
| workload-scheduler | Planner-Executor | this repo |
| ops-orchestrator | Orchestrator / Router | link |
Active development begins Week 25. Full demo available at Phase 4 completion.
git clone https://github.com/Jlowpez/workload-scheduler
cd workload-scheduler
pip install uv && uv pip install -e ".[dev]"
pytest tests/ -vReplace the constraint data in data/synthetic/ with your organization's resource calendars, job queues, and capacity limits. The scheduling engine and optimization logic are unchanged — only the data source differs.
- ARCHITECTURE.md — System design and component breakdown
- TOOLS.md — Tool and library selection rationale
- ops-orchestrator — full platform view