0.6.0 - 2026-05-18

Release Notes

Added

• Reserved three new crate names in the workspace as placeholders for
  upcoming verticals: cobre-flow (power flow algorithms), cobre-uc
  (MILP-based unit commitment for hydrothermal dispatch), and
  cobre-emt (electromagnetic transient analysis). Each ships as an
  empty library following the existing cobre-tui reservation pattern
  and is not yet implemented.

• Five new columns in simulation/hydros.parquet, populated for every
  (stage, block, hydro) row:

  • equivalent_productivity_mw_per_m3s (MW/(m³/s)) — equivalent
    productivity ρ_eq. For ConstantProductivity and
    LinearizedHead hydros it is the stored input scalar; for FPHA
    hydros it is derived from VHA geometry, the specific productivity
    ρ_esp, and the reference operating point (V_ref, Q_ref).
  • accumulated_productivity_mw_per_m3s (MW/(m³/s)) — accumulated
    productivity ρ_acum: the sum of ρ_eq over each hydro and every
    plant downstream of it along the cascade.
  • incremental_inflow_energy_mw (MW) — incremental natural inflow
    expressed as energy, computed as
    ρ_acum · incremental_inflow_m3s.
  • stored_energy_initial_mwh (MWh) — stored reservoir energy at the
    start of the block,
    (storage_initial_hm3 − min_storage_hm3) · ρ_acum · 10⁶ / 3600.
  • stored_energy_final_mwh (MWh) — stored reservoir energy at the
    end of the block, using the same formula with the final storage.

Removed

• cobre_stochastic::par::fit_par_annual_with_reduction and its
  ReducedOrderFit return type. The function had been exposed alongside
  the wired PAR(p)-A reduction loop in cobre-sddp but was never called
  from production code, and its φ + ψ̂/12 "effective coefficient"
  contribution check disagreed with the φ-only check used by the actual
  estimation pipeline. The wired path
  (apply_annual_prepass_reductions → reduce_entity_orders_annual)
  remains the single source of truth.

• Column productivity_mw_per_m3s from simulation/hydros.parquet. The
  column was null for FPHA hydros and duplicated input for non-FPHA
  hydros. It is replaced by the five always-populated energy columns
  listed above.

• Inline productivity_mw_per_m3s field on
  system/hydros.json generation blocks. The per-stage
  productivity coefficient is now authored on every
  stage_ranges[] or seasons[] entry in
  system/hydro_production_models.json under the same key.
  Cases that previously omitted
  system/hydro_production_models.json and relied on the
  entity-level scalar must now ship the file with one entry per
  hydro.

• Input files system/scalar_parameter_definitions.parquet and
  system/scalar_parameter_values.parquet. Scalar parameters
  for generic constraints (@name references) are now authored
  in a single system/scalar_parameters.json carrying one
  object per parameter with a kind discriminator and
  kind-specific payload (value, values, or
  computed_spec).

• Several config.json fields whose values were no longer wired to any
  downstream consumer: modeling.inflow_non_negativity.penalty_cost (the
  per-hydro slack cost is now sourced exclusively from
  penalties.json::hydro.inflow_nonnegativity_cost),
  training.cut_formulation, training.forward_pass,
  simulation.policy_type, simulation.output_mode, simulation.output_path,
  and the legacy export flags
  (training, cuts, vertices, simulation, forward_detail,
  backward_detail, compression) under exports. Any case still carrying
  these keys is rejected at parse with an unknown field error.

• The "fixed" value of estimation.order_selection. Only "pacf" and
  "pacf_annual" remain valid; the alias was previously accepted and silently
  mapped to "pacf".

• The version field on penalties.json. The field was parsed but never
  consulted; schema-version gating is enforced via the $schema URL instead.

Changed

• system/hydro_production_models.json —
  productivity_mw_per_m3s is now optional for
  constant_productivity and linearized_head models. Omit the
  field (or set it to null) when the value is supplied per stage
  by system/hydro_energy_productivity.parquet.

• Productivity validation relaxed from strictly positive (> 0.0)
  to non-negative (>= 0.0) for productivity_mw_per_m3s in
  system/hydro_production_models.json,
  equivalent_productivity_mw_per_m3s in
  system/hydro_energy_productivity.parquet, and
  specific_productivity_mw_per_m3s_per_m in the same parquet.
  A value of 0.0 is accepted as a planned-outage marker for the
  affected (hydro, stage); the LP treats these coefficients as
  multipliers, so zero produces zero generation without any
  division-by-zero hazard. Negative values are still rejected.

• config.json::modeling.inflow_non_negativity.method is now a typed
  enum with four valid values ("none", "truncation", "penalty",
  "truncation_with_penalty"). Unrecognised strings were previously
  silently coerced to "none"; they are now rejected at parse.

• config.json::training.cut_selection.threshold is consulted only
  when method = "level1" (it is the activity-count cutoff for cut
  deactivation). The "lml1" and "domination" methods no longer
  fall back to threshold; they now require memory_window and
  domination_epsilon respectively, and reject configurations that
  omit them.

Breaking Changes

Output schema — simulation/hydros.parquet grows from 31 to 35
columns. The column productivity_mw_per_m3s is removed; five new
non-nullable Float64 columns are added after generation_mwh:
equivalent_productivity_mw_per_m3s,
accumulated_productivity_mw_per_m3s, incremental_inflow_energy_mw,
stored_energy_initial_mwh, stored_energy_final_mwh. The
accompanying variables.csv dictionary file is updated to reflect the
new columns and drops the entry for the removed one.

FPHA validation — Studies that declare a hydro with
generation_model: "fpha" must now supply either VHA geometry plus
specific_productivity_mw_per_m3s_per_m, or (when the optional
system/hydro_energy_productivity.parquet override is in place) a
per-(hydro, stage) equivalent_productivity entry. Cases that
previously loaded with neither source now fail fast at setup time with
an error that names the offending plant and lists the three accepted
remediations.

Hydro productivity authoring — Studies must remove
productivity_mw_per_m3s from every hydros.json generation
block and supply the same value via a
system/hydro_production_models.json entry. For non-FPHA models
(constant_productivity, linearized_head) the field is
positive when present on every stage_ranges[] / seasons[]
entry, and may be omitted (or null) to defer to the parquet
override (see the next entry). The previous opt-in
productivity_override key on those entries is renamed to
productivity_mw_per_m3s; the override semantics (replace the
entity-level base value) no longer apply because there is no
longer an entity-level base value.

Productivity resolution across files — The
equivalent_productivity_mw_per_m3s column in
system/hydro_energy_productivity.parquet now applies to all
hydro generation models, not only FPHA. A row supplying this value
for a non-FPHA hydro is honoured as the equivalent productivity
ρ_eq for that (hydro, stage) instead of being silently ignored.
Studies that supplied a value for the same (hydro, stage) pair
in both system/hydro_production_models.json
(productivity_mw_per_m3s) and system/hydro_energy_productivity.parquet
(equivalent_productivity_mw_per_m3s) are now rejected at load time
with a schema error naming both files. Studies that supplied a value
in neither file for a non-FPHA (hydro, stage) pair are likewise
rejected with a clear coverage-gap error, rather than failing deeper
in the SDDP setup layer.

Scalar parameters file format — The pair of input parquet
files system/scalar_parameter_definitions.parquet and
system/scalar_parameter_values.parquet is removed. Studies
that use scalar parameters in generic constraints must now ship
system/scalar_parameters.json instead. The new file uses one
array entry per parameter with a kind discriminator
(constant, per_stage, seasonal, or computed) and a
kind-specific payload. Authoring order is preserved at parse
time; LP coefficients are unaffected by authoring order
(parameters are looked up by id at LP-build time).

Strict JSON parsing — every input JSON file (config.json,
penalties.json, stages.json, initial_conditions.json, and every
file under system/, constraints/, and scenarios/) now rejects
unknown top-level and per-entry fields with a hard parse error.
Misspellings ("max_outflow" vs "max_outflow_m3s") and stale
configuration keys from earlier releases that previously survived
silently must be fixed before a case will load. The error message
names the unrecognised field.

Migration

• Replace reads of productivity_mw_per_m3s with
  equivalent_productivity_mw_per_m3s. For ConstantProductivity and
  LinearizedHead hydros the numeric value is identical to the
  previous column; for FPHA hydros it is the derived ρ_eq
  (previously null).
• Notebooks and dashboards that filtered out the old null values for
  FPHA hydros can now treat all hydros uniformly.
• Downstream consumers that need accumulated cascade productivity,
  energy-units inflows, or reservoir energy state should read from the
  four new columns instead of computing them externally.
• Remove productivity_mw_per_m3s from every generation block
  in system/hydros.json; transfer the same numeric value into
  the matching entry in system/hydro_production_models.json
  under the same key.
• Rename any productivity_override keys in
  system/hydro_production_models.json to
  productivity_mw_per_m3s. The values do not change; the
  positive-value validation rule still applies, but the
  override-versus-default distinction is gone.
• Replace system/scalar_parameter_definitions.parquet and
  system/scalar_parameter_values.parquet with a single
  system/scalar_parameters.json (see the schema at
  book/src/schemas/scalar_parameters.schema.json and the
  authoring guide at book/src/guide/scalar-parameters.md).

cobre-cli 0.6.0

Install cobre-cli 0.6.0

Install prebuilt binaries via shell script

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/cobre-rs/cobre/releases/download/v0.6.0/cobre-cli-installer.sh | sh

Install prebuilt binaries via powershell script

powershell -ExecutionPolicy Bypass -c "irm https://github.com/cobre-rs/cobre/releases/download/v0.6.0/cobre-cli-installer.ps1 | iex"

Download cobre-cli 0.6.0

File	Platform	Checksum
cobre-cli-aarch64-apple-darwin.tar.xz	Apple Silicon macOS	checksum
cobre-cli-x86_64-apple-darwin.tar.xz	Intel macOS	checksum
cobre-cli-x86_64-pc-windows-msvc.zip	x64 Windows	checksum
cobre-cli-aarch64-unknown-linux-gnu.tar.xz	ARM64 Linux	checksum
cobre-cli-x86_64-unknown-linux-gnu.tar.xz	x64 Linux	checksum

cobre-mcp 0.6.0

Install cobre-mcp 0.6.0

Install prebuilt binaries via shell script

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/cobre-rs/cobre/releases/download/v0.6.0/cobre-mcp-installer.sh | sh

Install prebuilt binaries via powershell script

powershell -ExecutionPolicy Bypass -c "irm https://github.com/cobre-rs/cobre/releases/download/v0.6.0/cobre-mcp-installer.ps1 | iex"

Download cobre-mcp 0.6.0

File	Platform	Checksum
cobre-mcp-aarch64-apple-darwin.tar.xz	Apple Silicon macOS	checksum
cobre-mcp-x86_64-apple-darwin.tar.xz	Intel macOS	checksum
cobre-mcp-x86_64-pc-windows-msvc.zip	x64 Windows	checksum
cobre-mcp-aarch64-unknown-linux-gnu.tar.xz	ARM64 Linux	checksum
cobre-mcp-x86_64-unknown-linux-gnu.tar.xz	x64 Linux	checksum