Supply-chain reports

SupplyChainAnalyzer solves a nominal scenario for exactly one positive target good. Its immutable result exposes the reports below. All schemas remain present when a report is empty. See shared conventions and the supply-chain guide.

Scope, prices, and graph views

• System means every configuration in the solved state, including any production outside the terminal good's upstream chain.
• Chain means processes in the realized terminal graph: active upstream producers, with their input goods and all coproduct outputs retained. Each included process contributes its whole employment, cost, and value added; nothing is allocated to individual coproducts.
• realized graphs start from configurations whose level exceeds tolerance (default 1e-9). allowed graphs start from configurations not disabled by the era cap, PM bans, exact building-group bans, or building limits at or below tolerance. Both are restricted to the terminal's upstream graph and attached coproducts. Allowed membership does not prove LP feasibility.
• processes(view), goods(view), and flows(view) use the requested graph, defaulting to realized. Even in allowed, quantities use the same solved levels; inactive alternatives do not get hypothetical production.
• Input/output coefficients include the compiled fixed throughput multiplier. It scales goods, not employment, construction, land, or infrastructure. Coefficients split production-table net flows into positive input/output magnitudes; flows already offset in that table cannot be recovered.
• Prices are state prices: base prices for the public nominal analyzer. The market automobile example at the end of this page has a separate schema.

Summary

result.summary() is a Series with these fixed fields and the resource-building column families below, named after the terminal good. Sweeps turn them into columns. On a successful analyzer result, all fields are present. Let Q denote whole-system terminal net output, G_chain chain weekly value added, and L_chain chain employment. Safe division returns zero only for an exactly zero denominator.

Column	Type / unit	Meaning / calculation	Presence / edge cases
scenario	String / label	Scenario.display_name().	Uses explicit scenario name or the generated display name.
terminal_good	String / identifier	The single good being analyzed.	Has positive aggregated exports in the scenario.
objective	String / category	Nominal objective: gdp, employment, automation, or construction_cost.	gdp_per_capita is rejected by the analyzer.
solver_status	Integer / status code	HiGHS result status.	Successful solves normally report 0; sweep failures use synthetic -1.
solver_message	String / diagnostic text	HiGHS message retained from the solve.	Sweep failures instead retain the exception message.
objective_value	Float / objective units	Whole-system attained objective: weekly GDP (£/week), employment (workers), automation (workers), or construction points.	Sign restored for maximized GDP/employment; not necessarily a chain-only metric.
target_quantity	Float / goods units per week	Aggregated terminal export orders.	Strictly positive; imports may cover some or all demand.
target_base_value	Float / £ per week	Target quantity times terminal state price.	Equals base value in nominal analysis; sweep preserves effective terminal export value on failure.
terminal_base_price	Float / £ per goods unit	Terminal price from the state.	Base price in nominal analysis; not an annual rate.
terminal_gross_output	Float / goods units per week	Whole-system terminal building_goods_output.	Before subtracting terminal consumption by processes.
terminal_net_output	Float / goods units per week	Q = levels @ net_matrix[:, terminal].	Whole-system output minus process inputs; not reduced by exports or population orders.
terminal_surplus	Float / goods units per week	terminal_net_output + terminal imports - terminal exports - min_goods_balance.	Slack above the household-excluding balance floor; can be slightly negative within solver tolerance.
system_gdp_weekly	Float / £ per week	Sum of system output minus input quantities valued at state prices.	Excludes payroll; imports/exports are not separately added to GDP.
system_gdp_annual	Float / £ per year	52 * system_gdp_weekly.	Annualization only, not a second solve.
chain_gdp_weekly	Float / £ per week	G_chain = sum(realized process value_added).	Includes entire coproduct values of included processes.
chain_gdp_annual	Float / £ per year	52 * chain_gdp_weekly.	Same process scope as weekly chain GDP.
system_employment	Float / workers	Sum of all state employment.	Model population means employed workers.
chain_employment	Float / workers	L_chain = sum(realized process employment).	No allocation to individual output goods.
system_construction_cost	Float / construction points	Whole-system levels dotted with construction coefficients.	A stock of required construction, not a weekly cost.
chain_construction_cost	Float / construction points	Sum of realized process construction costs.	Can be zero for zero-cost configurations.
system_arable_land	Float / arable-land units	Whole-system levels dotted with land coefficients.	Uses the immediate agricultural group rule on the economy dictionary page.
chain_arable_land	Float / arable-land units	Sum of realized process arable land.	Zero when no included process consumes modeled arable land.
system_net_infrastructure_usage	Float / infrastructure points	Whole-system levels dotted with net infrastructure coefficients.	Negative means net infrastructure provision.
chain_net_infrastructure_usage	Float / infrastructure points	Sum of realized process net infrastructure usage.	Also signed; not an absolute consumption total.
employment_per_terminal_unit	Float / workers per unit/week of output capacity	L_chain / Q.	Safe division; denominator is net system output, not the target or gross output.
construction_per_terminal_unit	Float / construction points per unit/week of output capacity	chain_construction_cost / Q.	Safe division; construction is not annualized.
gdp_per_terminal_unit	Float / £ per terminal goods unit	G_chain / Q.	Both GDP and net production are weekly; their time factors cancel.
output_per_10k_workers	Float / goods units per week per 10,000 workers	Q * 10000 / L_chain.	Safe division; zero if chain employment is exactly zero.
realized_process_count	Integer / nodes	Number of configurations in the realized graph.	Counts configurations, not summed building levels.
realized_good_count	Integer / nodes	Number of good nodes in the realized graph.	Includes terminal and attached coproduct goods.
realized_edge_count	Integer / edges	Number of directed goods/process edges in the realized graph.	Input and output edges count separately.
allowed_process_count	Integer / nodes	Number of configurations in the allowed graph.	Includes inactive alternatives upstream of the terminal.
allowed_good_count	Integer / nodes	Number of good nodes in the allowed graph.	Does not count every good in the economy.
allowed_edge_count	Integer / edges	Number of directed edges in the allowed graph.	Not the number of diagram edges after visualization pruning.
cyclic_component_count	Integer / components	Realized strongly connected components with more than one node, or a self-loop.	Zero for an acyclic graph.
cyclic_node_count	Integer / nodes	Number of realized nodes participating in a directed cycle.	Counts both goods and process nodes.
condensation_depth	Integer / edges	Longest directed-path length after each realized strongly connected component becomes one node.	Zero for an empty or single-component condensation; counts bipartite graph edges.
extractive_process_count	Integer / configurations	Count of realized processes marked is_extractive.	Uses immediate parent bg_extraction or either resource flag.
resource_process_count	Integer / configurations	Count with discoverable_resource or depletable_resource.	A process with both flags counts once; not resource capacity or remaining stock.
chain_gdp_share	Float / signed ratio	chain_gdp_weekly / system_gdp_weekly.	Safe division; negative or offsetting contributions can put the ratio outside [0, 1].
chain_employment_share	Float / fraction	chain_employment / system_employment.	Safe division.
chain_construction_share	Float / fraction	chain_construction_cost / system_construction_cost.	Safe division.
chain_arable_land_share	Float / fraction	chain_arable_land / system_arable_land.	Safe division.
chain_infrastructure_share	Float / signed ratio	chain_net_infrastructure_usage / system_net_infrastructure_usage.	Safe division; cancellation and negative values mean it need not lie in [0, 1].
system_gdp_per_capita_weekly	Float / £ per person per week	system_gdp_weekly * workforce_ratio / system_employment.	Safe division; includes uniformly inferred dependents.
system_gdp_per_capita_annual	Float / £ per person per year	52 * system_gdp_per_capita_weekly.	Included in every summary and sweep, including the video export.
chain_gdp_per_capita_weekly	Float / £ per person per week	chain_gdp_weekly * workforce_ratio / chain_employment.	Safe division; includes whole-process coproduct contributions.
chain_gdp_per_capita_annual	Float / £ per person per year	52 * chain_gdp_per_capita_weekly.	Same realized-chain scope.
system_wage_normalized_employment	Float / private-wage-equivalent workers	Sum of system levels times production-table wage_normalized_employment.	NaN if the column or any contributing coefficient is unavailable/nonfinite; inactive missing coefficients are ignored.
chain_wage_normalized_employment	Float / private-wage-equivalent workers	Sum of realized-chain levels times production-table wage_normalized_employment.	Same missing-data policy, restricted to contributing chain processes.
system_gdp_per_wage_normalized_employment_weekly	Float / £ per wage-equivalent worker per week	system_gdp_weekly / system_wage_normalized_employment.	Safe division; NaN when wage coefficients are unavailable.
system_gdp_per_wage_normalized_employment_annual	Float / £ per wage-equivalent worker per year	52 * system_gdp_per_wage_normalized_employment_weekly.	NaN when wage coefficients are unavailable.
chain_gdp_per_wage_normalized_employment_weekly	Float / £ per wage-equivalent worker per week	chain_gdp_weekly / chain_wage_normalized_employment.	Safe division; NaN when contributing chain wage coefficients are unavailable.
chain_gdp_per_wage_normalized_employment_annual	Float / £ per wage-equivalent worker per year	52 * chain_gdp_per_wage_normalized_employment_weekly.	Same realized-chain scope.
system_gdp_per_construction_cost_weekly	Float / £ per construction point per week	system_gdp_weekly / system_construction_cost.	Safe division; construction is a stock, not a weekly expense.
system_gdp_per_construction_cost_annual	Float / £ per construction point per year	52 * system_gdp_per_construction_cost_weekly.	Annualizes GDP only.
chain_gdp_per_construction_cost_weekly	Float / £ per construction point per week	chain_gdp_weekly / chain_construction_cost.	Safe division; includes all construction of realized-chain processes.
chain_gdp_per_construction_cost_annual	Float / £ per construction point per year	52 * chain_gdp_per_construction_cost_weekly.	Annualizes GDP only.

Graph counts use the complete retained graphs. They are independent of the top-producer pruning used by a plotted or Mermaid allowed view.

Resource building levels

These summary column families use full building identifiers, for example chain_levels_building_coal_mine and chain_levels_per_10k_pop_building_coal_mine. Extraction membership uses the same rule as is_extractive: immediate parent bg_extraction, or a discoverable or depletable resource flag. This includes ordinary mines, logging camps, fishing wharves, whaling stations, rubber plantations, and oil rigs. Farms, ranches, and other arable-land plantations remain represented by arable land.

Types come from the full economy production table, including banned and unused types. Columns follow the fixed summary fields, sorted by building identifier; within each type, system raw/normalized levels precede chain raw/normalized levels. All summaries from one economy share this schema, even empty sweeps and all-failure sweeps. Successful unused types are zero; failed solves are NaN.

Column family	Type / unit	Meaning / calculation	Presence / edge cases
system_levels_<building>	Float / building levels	Sum of all solved configuration levels for this extraction building type.	Fractional levels are retained; no aggregate resource-level total is added.
system_levels_per_10k_pop_<building>	Float / levels per 10,000 workers	system_levels_<building> * 10000 / system_employment.	Safe division; excludes dependents and unemployed people.
chain_levels_<building>	Float / building levels	Sum of realized-chain configuration levels for this extraction building type.	Multiple PM configurations of one building type are combined.
chain_levels_per_10k_pop_<building>	Float / levels per 10,000 workers	chain_levels_<building> * 10000 / chain_employment.	Safe division; 5 levels and 2,000 chain workers give 25 levels per 10,000 workers.

Processes

result.processes(view) has one row per configuration in the requested graph, in production-table row order. Let x be its solved level, m its throughput multiplier, I its adjusted input value per level, and O its adjusted output value per level. Unit economics retain all output goods.

Here, process means a building with its complete PM combination, rather than the building type alone. The guide explains the naming rationale and parallels with Brightway.

Column	Type / unit	Meaning / calculation	Presence / edge cases
building	String / identifier	Source building key.	Multiple configurations may share it.
production_method	String / compound identifiers	Complete chosen method combination.	Methods are joined by +.
building_group	String / identifier	Immediate source building group.	Missing metadata defaults to empty string.
parent_group	String / identifier	Immediate parent metadata of that group.	Missing metadata defaults to empty string.
land_usage	String / category	Source resolved land-use metadata.	Missing metadata defaults to empty string; not the land coefficient itself.
era	Integer / era number	Source configuration era.	Required by the reporting implementation.
allowed	Boolean / membership flag	Process node is in the terminal's allowed graph.	Graph membership, not independent LP feasibility.
active	Boolean / solved flag	x > tolerance.	May be false in the allowed view.
in_terminal_chain	Boolean / membership flag	Configuration appears in the realized terminal graph.	An active process elsewhere in the system can still be false.
level	Float / building levels	x, from the whole solved state.	Fractional or zero levels are valid.
throughput_multiplier	Float / multiplier	Compiled fixed goods multiplier m.	1 means unchanged; excludes level-dependent economy of scale.
unit_input_cost	Float / £ per week per level	I = adjusted input row @ state prices.	Includes throughput; zero with no modeled inputs.
unit_output_value	Float / £ per week per level	O = adjusted output row @ state prices.	Includes every coproduct.
unit_value_added	Float / £ per week per level	O - I.	Can be negative; excludes wages.
input_cost	Float / £ per week	I * x.	Zero for a zero-level allowed alternative.
output_value	Float / £ per week	O * x.	Zero for a zero-level allowed alternative.
value_added	Float / £ per week	(O - I) * x.	Entire process contribution, not allocated to one good.
employment	Float / workers	Per-level employment coefficient times x.	Not multiplied by m.
construction_cost	Float / construction points	Per-level construction coefficient times x.	Not multiplied by m.
arable_land	Float / arable-land units	Per-level land coefficient times x.	Uses immediate group membership; not multiplied by m.
net_infrastructure_usage	Float / infrastructure points	Per-level net infrastructure coefficient times x.	Signed; not multiplied by m.
is_extractive	Boolean / classification flag	Immediate parent_group == "bg_extraction" or either resource flag in allowed-graph metadata.	Defaults false if no matching metadata; does not traverse ancestors.
is_subsistence	Boolean / source flag	Source subsistence metadata.	Missing metadata defaults false.
discoverable_resource	Boolean / source flag	Source discoverable-resource metadata.	Missing metadata defaults false.
depletable_resource	Boolean / source flag	Source depletable-resource metadata.	Missing metadata defaults false; depletion is not simulated.

Goods

result.goods(view) has one row per good node in the requested graph, in economy goods order. Production and consumption sum only processes in that view, while imports, exports, and population orders come from the whole state. This mix of scopes means these balances need not equal whole-system market balances. Producer counts use the two retained graphs irrespective of the requested view.

Column	Type / unit	Meaning / calculation	Presence / edge cases
good	String / identifier	Good key.	Includes attached coproduct goods.
terminal	Boolean / flag	Whether this is the analyzed terminal good.	True only for the terminal key.
price	Float / £ per goods unit	State price for this good.	Base price in nominal analysis.
production	Float / goods units per week	View-process levels times adjusted output coefficients.	Zero is valid for an input-only good or inactive allowed production.
consumption	Float / goods units per week	View-process levels times adjusted input coefficients.	Positive magnitude, not a negative signed flow.
imports	Float / goods units per week	Whole-state imported orders.	Zero when no imports supplied.
exports	Float / goods units per week	Whole-state exported orders.	Zero when no exports supplied.
pop_needs	Float / goods units per week	Whole-state calculated household demand.	Zero for nominal supply-chain solves.
sell_orders	Float / goods units per week	View production + imports.	Combines view and state scope.
buy_orders	Float / goods units per week	View consumption + exports + pop_needs.	Combines view and state scope.
net_balance	Float / goods units per week	sell_orders - buy_orders.	Signed; differs from terminal net production.
output_value	Float / £ per week	View production * price.	Imports are not included.
input_value	Float / £ per week	View consumption * price.	Exports/population orders are not included.
input_cost_share	Float / fraction	input_value / total input value of all view processes.	Safe division; zero when total input value is zero.
allowed_producers	Integer / configurations	Process predecessors of this good in the allowed graph.	Restricted graph count, not every producer in the economy.
active_producers	Integer / configurations	Process predecessors of this good in the realized graph.	Count unaffected by view; does not sum their levels.

Flows

result.flows(view) has one row per directed graph edge, sorted by the source and target node strings. Inputs point from goods to processes; outputs point from processes to goods. Rates and amounts are positive magnitudes.

Column	Type / unit	Meaning / calculation	Presence / edge cases
source	String / node identifier	Edge source: good:<key> or process:<config>.	Direction depends on role.
target	String / node identifier	Edge target: process:<config> or good:<key>.	Direction depends on role.
role	String / category	input or output.	Distinguishes flow direction without a signed rate.
good	String / identifier	Good carried by this edge.	Same key as the goods report.
process	String / configuration identifier	Configuration key, without the process: prefix.	Same key as process config.
unit_rate	Float / goods units per week per level	Adjusted input or output coefficient.	Includes throughput; edges are created only for coefficients > 0.
realized_amount	Float / goods units per week	unit_rate * solved process level.	Can be zero for an allowed alternative.
price	Float / £ per goods unit	State price for the edge's good.	Base price in nominal analysis.
realized_value	Float / £ per week	realized_amount * price.	Zero for an inactive flow.
active	Boolean / flow flag	realized_amount > tolerance.	Amount threshold, distinct from process level threshold.
in_cycle	Boolean / endpoint flag	Both endpoints are cyclic nodes in the requested graph.	Does not require them to belong to the same cyclic component; do not interpret it as an exact cycle-edge test.

Workforce

result.workforce() uses only realized-chain configurations, in row order then profession order. It omits process/profession employment at or below tolerance.

Column	Type / unit	Meaning / calculation	Presence / edge cases
config	String / configuration identifier	Configuration employing the profession.	Matches process building + "+" + production_method.
building	String / identifier	Configuration's source building.	Multiple rows can share it.
profession	String / identifier	Profession key from the economy's profession index.	No row for employment at or below tolerance.
employment	Float / workers	State employment cell for this configuration/profession.	Per-level profession coefficient times solved level; throughput does not change it.

Constraints and bottlenecks

result.constraints() exposes compiled LP rows, inequalities first and equalities second, in their named slice order. Constraints apply to the whole system. Inequalities are stored in canonical form A_ub @ levels <= b_ub; for example the goods-balance floor uses negative net-production coefficients and a bound of imports - exports - min_goods_balance. Interpret the compiled sign before converting a marginal to a gameplay change. Variable bounds are not rows in this report.

Column	Type / unit	Meaning / calculation	Presence / edge cases
kind	String / category	inequality or equality.	Determines slack/residual interpretation.
family	String / constraint identifier	Named compiled constraint slice, such as min_goods_balance or building_level_caps.	Only families present in the compiled problem appear.
label	String / row label	Good/building/group key for applicable families; PM bans use their +-joined keys; other rows use the family name.	If label count differs from row count, uses family[i].
realized_value	Float / compiled row units	A_row @ solved levels.	Signed canonical left-hand side, not always the physical quantity named by the family.
bound	Float / compiled row units	Stored right-hand side b_row.	A lower-bound target can be represented as a negative bound.
slack	Float / compiled row units	For inequalities, bound - realized_value.	NaN for equalities; slightly negative values can occur within solver tolerance.
residual	Float / compiled row units	For equalities, realized_value - bound.	Exactly 0 in this column for inequalities.
binding	Boolean / threshold flag	Inequality slack <= tolerance; equality abs(residual) <= tolerance.	Numerical criterion, not proof that relaxing the row improves the objective.
solver_marginal	Float / solver objective units per compiled bound unit	Raw HiGHS derivative of its minimized objective with respect to this right-hand side.	NaN if diagnostics unavailable; GDP/employment objectives are negated for minimization.
improvement_per_unit	Float / solver objective units per compiled bound unit	For a finite inequality marginal, max(0, -solver_marginal).	NaN for equalities or unavailable/nonfinite marginals; local sensitivity to increasing the canonical bound.

For minimized construction/automation, positive improvement means lowering the minimized cost. For maximized GDP/employment, it means increasing the reported objective. It measures one unit of the compiled RHS, so relaxing a negative production constraint means lowering its requested physical minimum. It is local LP sensitivity and is not a guaranteed finite-change benefit.

result.bottlenecks() returns the same columns, restricted to binding inequalities with improvement_per_unit > tolerance, sorted by improvement descending, then family and label. It has no separate schema.

Alternatives

result.alternatives(good) lists producer predecessors of the specified good in the terminal's allowed graph. With no argument it covers all goods in that graph. It sorts by good, selected first, unit output descending, then config. Unit costs and values describe the entire process, not only the requested good. “Selected” means membership in the realized chain, not simply a positive level anywhere in the system.

Column	Type / unit	Meaning / calculation	Presence / edge cases
good	String / identifier	Good for which this process is an alternative producer.	A coproduct process can appear once per output good.
config	String / configuration identifier	Allowed producer configuration key.	No row if the good has no producer in the allowed graph.
building	String / identifier	Source building key.	Same metadata as the process report.
production_method	String / compound identifiers	Complete method combination.	+-joined.
era	Integer / era number	Configuration's required era.	Same metadata as the process report.
unit_output	Float / goods units per week per level	Adjusted output coefficient for this good.	Includes throughput, unlike the all-goods monetary unit fields.
unit_input_cost	Float / £ per week per level	Whole adjusted process input row valued at state prices.	Includes all process inputs.
unit_output_value	Float / £ per week per level	Whole adjusted process output row valued at state prices.	Includes every coproduct.
unit_value_added	Float / £ per week per level	unit_output_value - unit_input_cost.	Excludes wages; can be negative.
selected	Boolean / membership flag	Configuration occurs in the realized terminal graph.	False for alternatives outside that chain, even if active elsewhere.
level	Float / building levels	Whole-state solved level of the configuration.	Zero allowed for an inactive alternative.
parent_group	String / identifier	Parent metadata from the allowed graph.	Missing metadata defaults to empty string.
land_usage	String / category	Land-use metadata from the allowed graph.	Missing metadata defaults to empty string.
is_extractive	Boolean / classification flag	Immediate parent bg_extraction or either resource flag.	No ancestor traversal.
is_subsistence	Boolean / source flag	Source subsistence flag.	Missing metadata defaults false.
discoverable_resource	Boolean / source flag	Source discoverable-resource flag.	Missing metadata defaults false.
depletable_resource	Boolean / source flag	Source depletable-resource flag.	Missing metadata defaults false.

Sweep summary and failures

sweep_supply_chains() generates one scenario per requested good, preserving the requested order (by default economy.producible_goods order). Each quantity is target_value / base_price; every requested base price must be finite and positive. sweep.summary contains these fields followed by all summary fields. sweep.failures is the failed subset with the same columns and a reset index, including when empty.

Column	Type / unit	Meaning / calculation	Presence / edge cases
good	String / identifier	Good selected for this sweep row.	Matches terminal_good; retained on failures.
status	String / category	success or failure.	Failed solves do not stop subsequent goods.
error_type	String / exception name	Type of caught analysis exception.	Empty on success; currently ValueError on recorded failures.
error_message	String / diagnostic text	Caught exception message.	Empty on success; retained on failure.

On failure, scenario, terminal good, objective, effective terminal export quantity and base value, and terminal base price remain populated. solver_status becomes -1 and solver_message becomes the exception text. All remaining summary fields are NaN, including counts and objective_value; they do not mean zero. Sweep input-validation errors raise before an applicable result rather than becoming failure rows. For each terminal good, exports are the higher of aggregated template exports and target_value / base_price; other exports and all imports are retained.

sweep.df_building_levels is a float matrix with one row per production unit in economy.building_index order and one column per requested terminal good in selection order. Its index is named config and uses the existing building+production_method keys, keeping PM combinations separate. Each successful column contains the complete solved state.building_levels vector, including production outside the realized terminal chain, fractional levels, and zero levels for unused units. Failed columns contain NaN throughout. An empty sweep retains every production-unit row with zero columns. This report is available with keep_results=False and returns a defensive copy.

Example CSV additions

The nominal example exports retain every sweep field, including its failure policy, and add these labels. The comparison summary.csv has one row per configuration/good pair; video summary.csv files have one row per variant/good pair. Detailed results live under analyses/supply_chains/<configuration>/<good>/; cross-good summary.csv files live at configuration level. Cross-configuration comparisons live under analyses/supply_chains/comparison/summary.csv, alongside the population, construction, and annual GDP per capita charts pop.svg, construction_points.svg, and gdp_per_capita.svg. Labels describe presets, not additional economic metrics.

The comparison example also writes analyses/supply_chains/<configuration>/building_levels.csv beside each configuration's summary. Its first column is config, followed by one column per requested good; failed solves are blank cells. Empty selections produce only the config column with every production unit retained.

Column	Type / unit	Meaning / source	Presence / edge cases
configuration	String / preset label	Scenario preset in examples/supply_chain_compare.py.	In comparison and preset summary CSVs; rows retain it on failure.
variant	String / preset label	base or a named product-specific variant in examples/supply_chain_cangshulun.py.	Only in video summary CSVs; base covers all producible goods, named variants selected goods.
ban_config	String / preset label	Ban-policy label supplied by the example.	Retained on failure; an empty label in the comparison export does not mean no default building-group bans.

See the supply-chain API for report calls and the showcase for example downloads and commands.