cuopt-agent: multi-objective supply-vs-cost what-if + cost-cap eval

[cafzal]

What

A new what-if scenario (scenario_4.md) and eval case (max_supply_4) for the cuopt-agent's max-supply model — bringing multi-objective tradeoff exploration to the agent.

Why

The agent ships a multi-period MILP whose cost data (item_costs.csv, resource_costs.csv) is unused, and it only ever runs single-objective; the cuopt-multi-objective-exploration skill (NVIDIA/cuopt#1355) is available but no scenario exercises it. scenario_4 activates it as a supply-vs-cost tradeoff with no agreed weighting, framed to test judgment rather than prescribe the method (it surfaces the MILP-has-no-duals and 10000:1-weight traps without naming ε-constraint). max_supply_4 adds a numeric check graded by the existing cuopt_objective evaluator: cap total cost at 9,149.8 and maximize supply, ground truth 2,660,000, computed on cuOpt (Tesla T4).

User testing

Run on cuOpt (Tesla T4):

• The reference solve traces a well-posed supply-vs-cost frontier — supply buys in at ~288 weighted-units/$, then collapses to ~7/$ past the FG1-saturation knee (full sweep below).
• Before/after, same LLM: without the skill the agent collapses to a self-weighted blend (maximize supply − λ·cost); with scenario_4 + the skill it traces the frontier by ε-constraint, differences adjacent points for the rate (correctly noting a MILP has no duals), reports interpretable units, flags the knee, and leaves the pick to finance — and its solve hit the eval ground truth.

Reference frontier — max weighted supply vs. cost cap (cuOpt, Tesla T4); unconstrained max 3,450,061 at cost 15,249.6

total cost ≤ C	max weighted obj	FG1	FG2	Δobj/Δ$
3,812	1,140,000	114	0	—
5,168	1,530,000	153	0	288
6,523	1,920,000	192	0	288
7,879	2,300,000	230	0	280
9,235	2,690,000	269	0	288
10,590	3,020,000	302	0	243
11,946	3,300,001	330	1	207
13,301	3,440,006	344	6	103
14,657	3,450,063	345	63	7
16,012	3,460,062	346	62	7

Toy sample data — exercises the multi-objective method on the agent, not a planning study.

[cafzal]

@rgsl888prabhu cuopt-agent what-if + eval – activates the multi-objective skill in the agent (supply-vs-cost, no agreed weighting). GPU-validated; before/after in the description. Ready when you have a cycle.