Crate Overview

Cobre is organized as a Rust workspace of focused crates, each with a single responsibility and well-defined boundaries.

cobre/crates/
├── cobre/              Umbrella crate re-exporting workspace API
├── cobre-core/         Entity model (buses, hydros, thermals, lines)
├── cobre-io/           JSON/Parquet input, FlatBuffers/Parquet output
├── cobre-stochastic/   PAR(p) models, scenario generation
├── cobre-solver/       LP solver abstraction (HiGHS backend)
├── cobre-comm/         Communication abstraction (MPI, NUMA, shared-memory placeholder, local)
├── cobre-sddp/         SDDP training loop, simulation, cut management
├── cobre-cli/          Binary: run/validate/report/init/schema/summary/version
├── cobre-mcp/          Binary: MCP server for AI agent integration (reserved)
├── cobre-python/       cdylib: PyO3 Python bindings
├── cobre-tui/          Library: ratatui terminal UI (reserved)
├── cobre-flow/         Library: power flow algorithms (reserved)
├── cobre-uc/           Library: MILP unit commitment for hydrothermal dispatch (reserved)
└── cobre-emt/          Library: electromagnetic transient analysis (reserved)

Dependency Graph

The diagram below shows the primary dependency relationships between workspace crates. Arrows point from dependency to dependent (i.e., an arrow from cobre-core to cobre-io means cobre-io depends on cobre-core).

graph TD
    core[cobre-core]
    io[cobre-io]
    solver[cobre-solver]
    comm[cobre-comm]
    stochastic[cobre-stochastic]
    sddp[cobre-sddp]
    cli[cobre-cli]
    ferrompi[ferrompi]

    core --> io
    core --> stochastic
    stochastic --> io
    ferrompi --> comm
    io --> sddp
    solver --> sddp
    comm --> sddp
    stochastic --> sddp
    sddp --> cli

For the full dependency graph and crate responsibilities, see the methodology reference.

Feature Summary

The workspace provides an SDDP training and simulation pipeline:

• Entity model and topology validation (cobre-core)
• JSON/Parquet case loading with layered validation (cobre-io)
• LP solver abstraction with HiGHS backend, warm-start basis management, and bounded retry escalation (cobre-solver)
• Pluggable communication with MPI and local backends, execution topology reporting, and SLURM integration (cobre-comm)
• PAR(p) inflow models with deterministic correlated scenario generation, per-class sampling (InSample, OutOfSample, Historical, External), and inflow non-negativity enforcement (cobre-stochastic)
• SDDP training loop with forward/backward passes, Benders cut generation, cut synchronization, and composite stopping rules (cobre-sddp)
• Two-stage cut management pipeline with strategy-based selection (Level1/LML1/Dominated) and budget enforcement (cobre-sddp)
• Performance accelerators: LP scaling, model persistence, incremental cut injection, backward-pass work-stealing, parallel lower bound evaluation, basis-aware padding, and pre-allocated hot-path workspaces (cobre-sddp, cobre-solver)
• Simulation pipeline with Hive-partitioned Parquet output and FlatBuffers policy checkpointing (cobre-sddp)
• Policy warm-start and resume from checkpoint with per-stage cut counts (cobre-sddp)
• CLI subcommands (run, validate, report, init, schema, summary, version), rayon-based intra-rank thread parallelism, progress bars, and post-run summary (cobre-cli)
• Python bindings via PyO3 with Arrow zero-copy result loading (cobre-python)
• JSON Schema files for all input types, hosted for $schema editor integration

The workspace is covered by an automated test suite (cargo nextest run --workspace), including the deterministic example regression cases under examples/deterministic/ — one per modeled feature; see the Deterministic Regression Suite.