Contract index
Every Brushline contract has a schema version, a model module, a validator, stable named errors, and a test suite. This page is the map.
Contract |
Schema |
Validator |
Checks |
|---|---|---|---|
Compiled dataset |
|
|
Manifest, schema, relative paths, tables, columns, dtypes, keys, row counts, checksums, digest |
Corridor graph |
|
|
Ids, CRS, references, span order, contiguous sequence, treatment units, access, structures, length |
State and treatment |
|
|
State keys, ranges, unknown states, years, references, required segments |
Plan |
|
|
Action and assumption ids, segments, currencies, parent, assumptions |
Run |
|
|
Artifact ids, availability, hashes, status |
Pareto archive |
|
|
Objectives, scalarizations, member keys, dominance |
Operations timeline |
|
|
Event ids, ordering, segment references, open events |
Demo manifest |
|
|
Slug, sources, licences, artifacts, caveats, publication status |
Data catalogue and connectors |
|
|
Sources, licences, classifications, coverage, connectors, auth references, scope, outputs, audit, approval, secret scan |
Fetch plan |
|
|
Items, estimates, licences, access, materialization pointers, bulk approvals, catalogue drift |
Materialization |
|
|
Checksums, sizes, paths, containment, licences, pointers, retrieval time, missing reasons |
Coverage and gaps |
|
|
Six coverage states, evidence, reasons, next action, and stop/continue |
Agent capabilities |
|
|
Six capabilities, prompt ids, tool allowlists, validators, retry budgets, private markers, authority fields, sanitized outcomes |
Model builder session |
|
|
AOI, plan preview, availability, approvals, stages, review, handoff, and dry-run artifacts |
FEMIC LiDAR bundle |
|
|
Producer pin, selection digest, product checksums, CRS, datum, resolution, licence, paths, and provenance |
Terrain support |
|
|
Raster CRS, resolution, datum, buffer, statistic, valid pixels, coverage, slope, aspect, terrain class, and access proxy |
Canopy support |
|
|
CHM CRS, datum, buffer, height statistics, cover thresholds, structure, optional tree, VRI comparison, and applicability |
Initial vegetation state |
|
|
Field/CHM rules, evidence coverage, class probabilities, audit decisions, model card, and state-contract validity |
LiDAR coverage and applicability |
|
|
Acquisition recency, usable coverage, valid pixels, domain, six coverage states, next action, and stop/continue |
LiDAR benchmark |
|
|
Bundle pin, shared contract, absolute/relative tolerance, missing values, case statuses, and bundle hash |
Low-vegetation state space |
|
|
Functional groups, variables, thresholds, coverage states, class probabilities, uncertainty, and state-contract conversion |
Growth |
|
|
Group curves, site/climate/competition modifiers, monotonicity, asymptotes, age, uncertainty, and domain |
Treatment transitions |
|
|
Method/season efficacy, top-kill, residual, resprouting, delay, cost, trajectory, and monotonicity |
Treatment calibration |
|
|
Evidence, pooling, shrinkage, intervals, sample sizes, source ids, and leave-one-out checks |
Hazard and service indicators |
|
|
Clearance, fall-in, fuel, access, reliability components and weights, unknown states, and caveats |
Vegetation validation |
|
|
Spatial holdout, group error, interval coverage, time-lag sensitivity, candidate benchmark, and exclusions |
Low-vegetation model card |
|
|
Predictors, training/validation evidence, metrics, candidates, uncertainty, domain, licence, citation, caveats, and decision boundary |
Optimization formulation |
|
|
Stable variables, options, periods, targets, hard/soft constraints, and four objectives |
Optimization evaluation |
|
|
Exact four-objective result, costs, feasibility, named violations, and period trail |
Optimization search |
|
|
Adapter, seed, budget, assignment, objective vector, evidence, checkpoint, and replay flag |
Promotion record |
|
|
Promoted plan, parent, reason, objective vector, time, and promoter |
Tiny optimum |
|
|
Enumerated assignment count, best assignment/vector, ties, and exact/batch agreement |
Tiny validation |
|
|
Optimum, optional search, gap, reach flag, replay, and expected-match evidence |
Search budget |
|
|
Per-budget and per-seed objectives, feasibility, evaluations, replay, and no-optimality caveat |
Budget report |
|
|
Quality, scaling, checkpoint, adapter and Fresh Solutions versions, and caveat |
Optimization event |
|
|
Progress event with run, seed, budget, payload, objective, plan, and evidence references |
Optimization telemetry |
|
|
Versioned run, model/data versions, archive path, and ordered events |
Strategic dynamic program |
|
|
States, actions, transitions, weights, horizon, discount, and terminal values |
Strategic DP result |
|
|
Policy, value table, expected cost, service, risk, and uncertainty |
Strategic portfolio |
|
|
Selections, period budgets/capacity, service/risk/uncertainty, continuity, and violations |
Strategic monitoring |
|
|
Span decisions, frequency, route, sensor, cost, value of information, and reductions |
Strategic BAU |
|
|
Fixed-cycle selections, period summaries, outcomes, baselines, and budget/capacity flags |
Strategic report |
|
|
Optimized/BAU yearly cost, budget, service, risk, uncertainty, monitoring, and comparison |
Strategic scenario |
|
|
Data/model/config/seed versions, artifact digests, paths, and replay |
Strategic scenario archive |
|
|
Versioned scenario set with creation metadata |
Strategic validation |
|
|
Coverage, budget, service, monitoring, BAU, totals, and replay checks |
Tactical program |
|
|
Units, actions, year budgets, contractor allocation, cost, and violations |
Tactical resources |
|
|
Contractor requests, cost lines, capacity, cost by contractor/year, and violations |
Tactical access |
|
|
Road status/source, windows, route length, allowed/blocked decisions, and caveats |
Operational plan |
|
|
Work orders, source action, crew/day, cost, access gates, and violations |
Tactical FHOPS |
|
|
Optional boundary, request mapping, dry-run machine hours/productivity/cost, and notes |
Operations timeline |
|
|
Start/finish events, source work orders, segments, positions, cost, and unknown-position count |
Tactical validation |
|
|
Capacity, access, seasonality, traceability, FHOPS dry-run, and timeline checks |
Observation |
|
|
Coverage, detection, observed class, measurement error, confidence, and latency |
Plan exchange |
|
|
Cost category totals, owner/contractor views, bid comparison, and approval digest |
Field validation |
|
|
Required columns, plot/unit/position references, GPS distance, duplicates, ranges, photo IDs, and normalized observations |
Field comparison |
|
|
Field/LiDAR residuals, unit bias/MAE/RMSE, sample gates, review multipliers, and treatment-evidence status |
Execution uncertainty |
|
|
Seeded cost/productivity/capacity sampling, percentile bands, budget/capacity/contractor shortfall probabilities, and one-at-a-time sensitivities |
Scenario optimization |
|
|
Hindsight scenario re-solves, fixed/adaptive cost/hour/risk bands, budget and capacity violation probabilities, normalized feasibility penalty, and regret |
Planning brief |
|
|
Self-contained Markdown and HTML decision record with selected frontier, tactical and shift summaries, workforce, bid, uncertainty, scenario comparison, caveats, and reproduction metadata |
Rolling policy |
|
|
Annual-factor episodes, observed current conditions, persistence forecast, short lookahead, committed-year state carry-over, fixed-versus-rolling bands, and regret |
Policy execution |
|
|
Median rolling episode, first-year segment actions, tactical/access/resource/work-order/shift handoff, feasibility, and fixed-year comparison |
Policy execution scan |
|
|
All-episode first-year handoff feasibility, access checks, cost/hours/risk bands, and failure reasons |
Static plan evaluation |
|
|
Frontier candidate stress against a common actual budget, annual-factor cost/hours bands, budget/capacity violations, and budget-risk recommendation |
Operational stress |
|
|
Assumed weather/outage shift calendar, completion rate, blocked days, unscheduled hours, and failure reporting |
Operational stress scan |
|
|
All-episode weather/outage calendar stress, baseline and scenario completion rates, failure sources, and window sensitivity |
Operational remedies |
|
|
Added crew/machine capacity, scope reduction, and window remedies with scenario completion, mean completion, unscheduled hours, and full-episode counts |
Risk-ranked scope |
|
|
Whole-action deferral by first-year risk reduction per hour, actual deferred fraction, risk loss, residual risk, calendar completion, cost, and budget feasibility |
Risk trajectory |
|
|
Fixed-future 2027-2031 risk paths for full, risk-ranked, and proportional first-year deferral |
Rolling recovery |
|
|
Future-year re-optimization after first-year deferral, annual risk paths, recovery gaps, and five-year cost |
Out-of-sample recovery |
|
|
Resampled future factors, final risk gaps, recovery bands, cost bands, and budget outcomes |
Deferral frontier |
|
|
0-25% risk-ranked first-year deferral sweep with calendar completion, fixed and recovered risk gaps, and five-year cost |
Frontier validation |
|
|
Out-of-sample risk gaps, recovery, budget, and cost for every deferral candidate |
Robust plan |
|
|
Scenario-robust static MILP with training-path budget/capacity constraints and independent out-of-sample evaluation |
Chance plan |
|
|
Sampled budget chance constraint with hard capacity, training-path budget rows, and independent out-of-sample evaluation |
Chance capacity plan |
|
|
Sampled budget and capacity chance constraints with independent out-of-sample evaluation |
Chance 100-path plan |
|
|
100-path sampled budget chance constraint with hard capacity and independent test paths |
Two-stage recourse |
|
|
Fixed first-stage plan, observed annual factors, risk-ranked action deferral, budget/capacity feasibility, completion, and risk loss |
Multi-year two-stage recourse |
|
|
Annual recourse across a fixed five-year plan, budget/capacity feasibility, completion, risk trajectories, and deferred workload |
Rolling recourse |
|
|
Static, fixed-future recourse, rolling-only, and rolling-recourse comparison on common annual factors |
Weather recourse |
|
|
Calendar-naive versus calendar-aware annual recourse under assumed weather/outage calendars |
Forecast-skill sensitivity |
|
|
Forecast-aware recourse across a coarse forecast-skill sweep |
Window remedy |
|
|
Observed-calendar awareness across 75-, 90-, and 105-day fall windows |
Capacity remedy |
|
|
Added crew and machine versus longer fall windows with explicit resource costs |
Stateful strategic plan |
|
|
Linear risk state, treatment transitions, budget/capacity constraints, objective, annual outcomes, solver provenance, and replay |
Stateful heuristic plan |
|
|
Fresh Solutions adapter, greedy seed, linear state replay, budget/capacity feasibility, seed, iterations, evaluations, and accepted moves |
Stateful decomposition |
|
|
Cluster grouping, allocated budgets, exact cluster solves, baseline fallback, aggregate replay, and gap |
Stateful coordination |
|
|
Lagrangian budget/crew prices, cluster priced subproblems, subgradient iterations, and dual bounds |
Shift plan |
|
|
Crew/machine skills, contractor matching, daily shift hours, travel, split work, utilization, and violations |
Studio |
|
|
Artifact-backed block scene with pre/post variants and unknown-position labels |
Demo portfolio |
|
|
Public synthetic instances, licences, caveats, artifacts, and pilot link |
Verification |
|
|
Criterion evidence, BAU/monitoring/exchange, performance, reproducibility, and recommendation |
Error prefixes
Prefix |
Contract |
|---|---|
|
Compiled dataset |
|
Corridor graph |
|
State and treatment |
|
Plan |
|
Run |
|
Pareto archive |
|
Operations timeline |
|
Demo manifest |
|
Data catalogue and connectors |
|
Fetch plan |
|
Materialization manifest |
|
Coverage and gap report |
|
Agent capabilities and sanitized outcome record |
|
FEMIC LiDAR bundle |
|
Terrain support |
|
Canopy support |
|
LiDAR coverage and applicability |
|
LiDAR benchmark |
|
Low-vegetation state space |
|
Growth model |
|
Treatment transitions |
|
Treatment calibration |
|
Hazard and service indicators |
|
Vegetation validation |
|
Low-vegetation model card |
|
Optimization formulation |
|
Optimization evaluation |
|
Optimization move validation |
|
Promotion record |
|
Telemetry stream |
|
Strategic dynamic program |
|
Strategic portfolio |
|
Strategic monitoring |
|
Business-as-usual baseline |
|
Strategic reports |
|
Strategic scenario |
|
Tactical program |
|
Tactical resources |
|
Tactical access |
|
Operational plan |
|
Tactical FHOPS boundary |
|
Operations timeline |
|
Field validation |
|
Field comparison |
|
Stateful strategic plan |
|
Shift scheduling |
|
Studio compatibility |
Test layers
Unit tests for each model and validator.
Contract tests for the public interfaces.
Integration tests for the CLI (
validate,field validate,build,simulate,prototype).Catalogue tests for licences, classifications, unknown references, connector scope, approval, and the secret scan.
Fetch tests for estimates, bulk approval, catalogue drift, materialization checksums and containment, the six coverage states, and the gap decision.
Capability tests for the six prompts, tool allowlists, validator failure, retry budget, fake provider, private markers, authoritative fields, and sanitized records.
Model-builder tests for the static shell, plan preview, approvals, blocked builds, dry-run artifacts, progress SSE, review, handoff, and custom AOI gaps.
LiDAR bundle tests for the producer pin, selection digest, checksums, CRS, datum, licence, path containment, product kinds, derived flags, and provenance.
Terrain tests for buffered aggregation, coverage states, every statistic, slope and access classes, RASTER CRS/datum/resolution checks, and provenance.
Canopy tests for height and cover statistics, optional structure/tree products, VRI comparison, partial/missing/out-of-domain states, and applicability.
State-initialization tests for field-first selection, every prototype class, unknown/partial states, species fallback, audit decisions, probabilities, and contract validity.
LiDAR coverage tests for observed, stale, partial, missing, out-of-domain, excluded, optional gaps, domain reasons, next actions, and stop/continue.
Benchmark tests for the committed fixture, absolute and relative tolerance, contract mismatch, required/optional/unexplained missing values, duplicate keys, and report consistency.
State-space tests for every class, coverage state, strata, probabilities, config validation, unknown and partial projection, and state-contract conversion.
Growth tests for monotonicity, asymptotes, all functional groups, site/climate/competition modifiers, domain blocking, uncertainty, conversion, and validator errors.
Treatment tests for method/season effects, immediate residual, resprouting, delay, trajectory, cost, conversion, and validator errors.
Calibration tests for exact/pooled/prior levels, shrinkage, intervals, holdout checks, model validation, and parameter application.
Hazard tests for clearance statuses, fall-in risk, fuel connectivity, access, reliability weighting, missing components, caveats, and validator errors.
Validation tests for spatial holdout, eligibility exclusions, group metrics, interval coverage, lag sensitivity, candidate disagreement, and report validation.
Model-card tests for predictor/domain ranges, evidence, metrics, candidates, uncertainty, domain checks, out-of-domain rejection, and card validation.
Formulation tests for stable indices, options, allowed methods, access constraints, objectives, validation, and build reproduction.
Evaluator tests for exact growth/treatment replay, costs, budgets, capacity, access, service, monitoring allocation, discounting, batch equality, scalarization, and apply/undo.
Operator tests for seeded determinism, every move kind, access and locks, move validation, apply/undo, budget/capacity/monitoring/access/service repair, and non-repairable blocked models.
Adapter tests for the Fresh Solutions contract, scalarizations, random/mutation/crossover, single and compound proposals, all search modes, checkpoint replay, and rejected non-replayable results.
Archive tests for dominance, duplicates, infeasible rejection, crowding, size eviction, round-trip, invalid archives, and promotion records.
Tiny-validation tests for complete enumeration, the committed optimum, exact/batch agreement, search gap, replay, infeasible instances, ties, and report round-trips.
Budget tests for every budget/seed point, feasible and infeasible runs, scaling timing and memory, checkpoint save/replay, composed reports, and input guards.
Telemetry tests for committed stream/archive compatibility, run events, archive insertion, sequence and order validation, missing/mismatched artifacts, SSE, and JSON round-trips.
Dynamic-program tests for the committed policy, transition/availability validation, tie-breaking, terminal values, expected outcomes, and invalid models.
Portfolio tests for service repair, budget/capacity limits, risk warnings, continuity, local swaps, empty options, and invalid references.
Monitoring tests for value-ranked selection, intervals/event years, budget/value floors, empty options, invalid references, and model guards.
BAU tests for annual and multi-year cycles, budget truncation and unenforced violations, capacity, action filters, empty/invalid inputs, and like-for-like comparison.
Report tests for optimized/BAU rows, monitoring allocation, overruns, mismatched scopes, row totals, and JSON round-trips.
Scenario tests for canonical digests, artifact replay, missing/tampered artifacts, archive validation, model guards, and the committed example.
Validation tests for the 10-year case, every named check, failed service/budget/monitoring/BAU/totals/ replay, writers, and model guards.
Tactical program tests for feasible allocation, budget/method/season/year violations, missing refs, empty actions, and model guards.
Resource tests for the five cost lines, cost by contractor/year, capacity/method/profile failures, duplicate profiles, and model guards.
Access tests for open/restricted/closed/unknown status, source preservation, route totals, missing links, duplicates, and model guards.
Operations tests for work-order traceability, crew/day assignment, blocked/missing access, and totals.
FHOPS tests for mapping, deterministic dry-run, boundary pin, optional-version reporting, and guards.
Timeline tests for start/finish events, known/unknown positions, segment/crew traceability, ordering, and model guards.
Tactical validation tests for the synthetic case, every named check, fixture loading, writers, and model guards.
Property tests for duplicate detection, dominance, corridor length conservation, and range checks.
The self-check that the planning surfaces pass the workflow validator.
The committed synthetic fixture is the shared contract example. A fresh build must match its artifact digests and content digest; the fixture tests enforce that.
Adding a contract
Add the schema version and the models.
Add the validator and named error prefixes.
Add a valid example and at least one failing case per invariant.
Add the JSON round-trip.
Add the reference page and the API page.
Add the contract to this index.
Run
python scripts/dashboard.py.