Align priceQuality implementation with API docs

[fleupold]

Description

The API docs state that "optimal" quotes should only care about the out amount, without attempting or taking verification into account. However, the current logic uses the same behavior for optimal and verified quotes. This PR changes this.

Note, that verification can still be turned off server side by setting verification-mode unverified.

Streamed quotes continue to use the verified quoting logic (as per docs)

Changes

• Separate optimal and verified quoter (the former not having any verification configured) in the quote
• Route requests to the respective quoter
• Update documentation to be more explicit
• Small refactor on pre-cached estimator entries (fast vs. optimal now becomes verified vs. unverified)

How to test

Adjusted e2e test

[github-actions]

Reminder: Please consider backward compatibility when modifying the API specification.
If breaking changes are unavoidable, ensure:

• You explicitly pointed out breaking changes.
• You communicate the changes to affected teams (at least Frontend team and SAFE team).
• You provide proper versioning and migration mechanisms.

---

Caused by:

• crates/orderbook/openapi.yml

[claude]

Claude finished @fleupold's task in 3m 33s —— View job

---

PR Review: Align priceQuality implementation with API docs

• Load pr-review skill
• Gather context (diff, related code)
• Analyze changes
• Post review

The change is well-structured and correctly implements the documented split between optimal (best price, no simulation) and verified (simulated + preferring verified quotes). I traced the wiring end-to-end and it holds up:

• factory.rs — the verified/unverified entry rename is accurate: unverified is the raw instrumented estimator, verified wraps it in the trade verifier (falling back to unverified when none is configured). unverified_price_estimator correctly omits .with_verification(...), so the CompetitionEstimator defaults to QuoteVerificationMode::Unverified (no simulation, no preference for verified quotes) — exactly the intended "optimal" behavior. Estimator entries stay cached in the HashMap, so no extra driver load is introduced.
• run.rs — order creation (order_validator) and streaming both correctly route to verified_quoter, preserving prior behavior for those paths; only the one-shot optimal path now uses the unverified quoter.
• quoter.rs — routing per PriceQuality is correct, and the default verified_quoter field falling back to the base quoter (when the builder isn't called) is sensible.
• model/quote.rs — moving #[default] from Optimal to Verified keeps default request behavior effectively unchanged, since the old default (Optimal) previously behaved as verified.

One maintainability note (inline): the price_estimation_drivers mapping block is now duplicated three times in run.rs — worth extracting into a single binding.

No correctness issues found. The behavior change for clients explicitly requesting optimal (previously simulated, now not) is intentional and already called out in the PR description / the compatibility bot comment.
· optimal_quotes_without_verification

[claude]

This price_estimation_drivers.iter().map(...).collect() block is now repeated three times verbatim (here plus the price_estimator and fast_price_estimator calls above). Consider extracting it into a small local binding once, e.g.:

let price_estimation_solvers: Vec<_> = config
    .order_quoting
    .price_estimation_drivers
    .iter()
    .map(|d| configs::native_price_estimators::ExternalSolver {
        name: d.name.clone(),
        url: d.url.clone(),
    })
    .collect();

and pass &price_estimation_solvers to all three. Reduces the risk of the three copies drifting.

Fix this →

[MartinquaXD]

Looks alright to me.

[fleupold]

Given this is a breaking change, I'm holding off merging until we have figured out which existing integration partners should change their priceQuality parameter to receive the optimal experience.

[jmg-duarte]

nit

Suggested change

	Estimates do not get verified by
	simulation.
	Optimal: The price estimate is chosen among all price estimates, ranked
	purely by the promised price. Estimates do not get verified by
	simulation.
	Verified: All price estimates get verified by simulation whenever
	possible and verified estimates are preferred over unverified ones,
	even when an unverified estimate promises a better price. The
	response's `verified` flag indicates whether the returned estimate
	was actually verified.
	Estimates do not get verified by simulation.
	Optimal: The price estimate is chosen among all price estimates, ranked
	purely by the promised price. Estimates do not get verified by simulation.
	Verified: All price estimates get verified by simulation whenever
	possible and verified estimates are preferred over unverified ones,
	even when an unverified estimate promises a better price. The
	response's `verified` flag indicates whether the returned estimate
	was actually verified.