Codecov Report

❌ Patch coverage is 98.93048% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 97.59%. Comparing base (51e49f5) to head (7ffed3c).
⚠️ Report is 1 commits behind head on main.

@@            Coverage Diff             @@
##             main     #727      +/-   ##
==========================================
+ Coverage   97.58%   97.59%   +0.01%     
==========================================
  Files         411      413       +2     
  Lines       51196    51383     +187     
==========================================
+ Hits        49958    50149     +191     
+ Misses       1238     1234       -4     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Implementation Summary

Changes

• Added the new fixed graph model BottleneckTravelingSalesman with brute-force evaluation based on Hamiltonian-cycle edge subsets and bottleneck objective minimization.
• Added dedicated unit tests for construction, evaluation, solver behavior, serialization, and the paper example.
• Registered the model across graph/model/prelude exports and the graph canonical example chain.
• Wired pred create support for explicit instances, random instances, and --example in the CLI.
• Added a new paper entry and bibliography support for the Bottleneck Traveling Salesman problem, including the K5 canonical example and the BTSP-versus-TSP contrast.

Deviations from Plan

• The first Batch 1 implementation pass drifted on two test details (it used a smaller construction fixture and missed the disconnected-subtour invalid case). The spec-review gate caught that, and the tests were corrected before final verification.
• The quality-review subagent did not return a usable verdict in time, so I replaced that gate with a manual code-quality review plus full verification (make paper, make test, make clippy).

Open Questions

• None.

Agentic Review Report

Structural Check

Structural Review: model BottleneckTravelingSalesman

Structural Completeness

Build Status

• make test: PASS
• make clippy: PASS
• make paper: PASS (extra verification beyond the skill minimum)

Semantic Review

• evaluate() correctness: OK — it rejects wrong-length or non-Hamiltonian edge selections and computes the bottleneck as the maximum selected edge weight.
• dims() correctness: OK — vec![2; num_edges] matches the binary edge-selection encoding.
• Size getter consistency: ISSUE — the model defines num_vertices()/num_edges() in src/models/graph/bottleneck_traveling_salesman.rs:74 but never declares them to the registry for an isolated problem, so pred show/pred inspect expose no size fields for this model.
• Weight handling: OK — weights are managed with inherent methods and the model stays fixed to SimpleGraph + i32, matching the issue.
• Paper entry: OK — make paper completed, so the new Typst block is syntactically valid and wired to the example DB.

Issue Compliance (if linked issue found)

Summary

• 17/17 structural checks passed
• 8/8 issue compliance checks passed
• ISSUE: src/models/graph/bottleneck_traveling_salesman.rs:74 defines the expected size getters, but the model never registers explicit size fields, so CLI/MCP metadata for this isolated problem is incomplete (pred inspect reports "size_fields": []).
• Note: the implementation packet reported Whitelist: fail; after manual diff review I did not identify a separate blocker beyond the metadata gap above.

Quality Check

Quality Review

Design Principles

• DRY: OK — the new model mirrors TravelingSalesman intentionally and keeps the duplicated surface area limited to model-specific wiring.
• KISS: OK — the implementation is straightforward: reuse the shared Hamiltonian-cycle checker, then swap the objective from sum to max.
• HC/LC: OK — model logic, CLI integration, tests, and paper entry stay separated by module responsibility.

HCI (if CLI/MCP changed)

• Error messages: OK — pred create BottleneckTravelingSalesman --graph 0-1,1-2 --edge-weights 1 returns Error: Expected 2 edge weight values but got 1.
• Discoverability: OK — the model appears in pred list, pred show BottleneckTravelingSalesman, and pred create --help.
• Consistency: OK — CLI help and create handling match the existing edge-weighted graph-model pattern in problemreductions-cli/src/commands/create.rs:1448 and problemreductions-cli/src/cli.rs:217.
• Least surprise: ISSUE — unlike analogous models such as TravelingSalesman, the new model shows no Size fields section in pred show and no size-field metadata in pred inspect, because the model file only registers schema metadata and omits explicit size-field registration.
• Feedback: OK — successful create/solve flows report the instance path and solution cleanly.

Test Quality

• Naive test detection: ISSUE
  • src/unit_tests/models/graph/bottleneck_traveling_salesman.rs:28 covers construction, evaluation, brute-force solving, serialization, and the paper example, but it never checks declared size fields or pred inspect metadata. That gap allowed the missing registry wiring for num_vertices/num_edges to slip through even though the linked issue explicitly called out size fields.

Issues

Critical (Must Fix)

None.

Important (Should Fix)

• src/models/graph/bottleneck_traveling_salesman.rs:12: add explicit ProblemSizeFieldEntry { fields: &["num_vertices", "num_edges"] } for this isolated model. Without it, pred show/pred inspect expose incomplete metadata until some future reduction happens to mention those fields.
• src/unit_tests/models/graph/bottleneck_traveling_salesman.rs:28: add a regression test for declared size fields (or CLI inspect output) so this user-facing metadata regression is covered.

Minor (Nice to Have)

None.

Summary

• ISSUE: missing explicit size-field registration makes CLI/MCP metadata for the new model incomplete.
• ISSUE: tests do not cover the size-field metadata path, so the regression is currently invisible to the suite.

Agentic Feature Tests

Agentic Feature Tests

• Feature: BottleneckTravelingSalesman
• Role: downstream CLI user trying to discover, create, inspect, and solve the new model from docs/help alone.
• Use case: find the new model in the catalog, inspect its shape, generate the canonical example, and solve it with the documented CLI workflow.
• Verdict: partial — core functionality works, but metadata discoverability is incomplete.

What I tried

• pred list | rg BottleneckTravelingSalesman
• pred show BottleneckTravelingSalesman
• pred create --example BottleneckTravelingSalesman -o <tmp>/btsp.json
• pred inspect <tmp>/btsp.json
• pred solve <tmp>/btsp.json --solver brute-force
• pred create BottleneckTravelingSalesman --graph ... --edge-weights ... | pred solve - --solver brute-force

Confirmed findings

• confirmed / important: pred inspect on the canonical example reports "size_fields": [], and pred show BottleneckTravelingSalesman omits the Size fields section entirely. This is reproducible in the current PR worktree and makes the new model less self-describing than analogous graph models such as TravelingSalesman.

Not reproducible in current worktree

• None.

Recommended fix

• Register num_vertices and num_edges explicitly for BottleneckTravelingSalesman, then add a regression test that exercises either declared_size_fields() or the pred inspect output.

Generated by review-pipeline

Agentic Review Report

Structural Check

Structural Review: model BottleneckTravelingSalesman

Structural Completeness

Build Status

• make test: PASS
• make clippy: PASS

Semantic Review

• evaluate(): OK — checks config length, reuses the shared Hamiltonian-cycle validator, and returns the correct bottleneck objective (src/models/graph/bottleneck_traveling_salesman.rs:111). I independently enumerated the 12 Hamiltonian tours of the K5 paper/example instance and confirmed the unique optimum has bottleneck 4.
• dims(): OK — one binary variable per edge via vec![2; num_edges] (src/models/graph/bottleneck_traveling_salesman.rs:107).
• Size getter consistency: OK — num_vertices() / num_edges() exist and match the declared complexity string "num_vertices^2 * 2^num_vertices" (src/models/graph/bottleneck_traveling_salesman.rs:74, src/models/graph/bottleneck_traveling_salesman.rs:79, src/models/graph/bottleneck_traveling_salesman.rs:168).
• Weight handling: OK — weights are exposed and updated through inherent methods (weights, set_weights, is_weighted) rather than extra traits (src/models/graph/bottleneck_traveling_salesman.rs:53).

Issue Compliance

Summary

• 17/18 structural checks passed
• 8/8 issue compliance checks passed
• FAIL — files outside expected model whitelist: problemreductions-cli/src/cli.rs, problemreductions-cli/src/commands/create.rs, src/lib.rs

Quality Check

Quality Review

Design Principles

• DRY: OK — the new model reuses traveling_salesman::is_hamiltonian_cycle instead of duplicating Hamiltonian-cycle feasibility logic (src/models/graph/bottleneck_traveling_salesman.rs:95).
• KISS: OK — the implementation mirrors TravelingSalesman and changes only the objective aggregation from sum to max, which is the minimal model-specific delta (src/models/graph/bottleneck_traveling_salesman.rs:121).
• HC/LC: OK — the new file stays focused on one model; the only cross-module dependency is the shared Hamiltonian-cycle helper, which is appropriate (src/models/graph/bottleneck_traveling_salesman.rs:95).

HCI (if CLI/MCP changed)

• Error messages: OK — explicit create-path usage errors remain actionable (problemreductions-cli/src/commands/create.rs:1450).
• Discoverability: ISSUE — pred create BottleneckTravelingSalesman prints an example using a 4-vertex path graph, which has no Hamiltonian cycle, so the most obvious “copy the example then solve it” user flow fails immediately (problemreductions-cli/src/commands/create.rs:539).
• Consistency: OK — the model is listed consistently across pred list, pred show, explicit create, random create, and help text (problemreductions-cli/src/cli.rs:217, problemreductions-cli/src/commands/create.rs:652, problemreductions-cli/src/commands/create.rs:5079).
• Least surprise: ISSUE — following the help example through pred solve yields Error: No solution found, which makes the newly added model look broken even though the canonical example works (problemreductions-cli/src/commands/create.rs:539).
• Feedback: OK — pred show, pred create --example BottleneckTravelingSalesman, pred solve, and pred evaluate all return sensible outputs in the current PR worktree.

Test Quality

• Naive test detection: ISSUE
  • The PR adds new CLI/help/create branches in problemreductions-cli/src/commands/create.rs:539, problemreductions-cli/src/commands/create.rs:1448, and problemreductions-cli/src/commands/create.rs:5079, but the only added automated coverage is the model unit-test file src/unit_tests/models/graph/bottleneck_traveling_salesman.rs:28. There is no CLI/integration test exercising those user-facing paths, and the infeasible help example above slipped through as a result.

Issues

Critical (Must Fix)

• None.

Important (Should Fix)

• pred create BottleneckTravelingSalesman advertises --graph 0-1,1-2,2-3 --edge-weights 1,1,1 as its example instance (problemreductions-cli/src/commands/create.rs:539), but that graph has no Hamiltonian cycle. Reproduced by piping the example into pred solve, which returns Error: No solution found.
• No automated CLI/integration test covers the new create/help/example wiring for BottleneckTravelingSalesman (problemreductions-cli/src/commands/create.rs:539, problemreductions-cli/src/commands/create.rs:1448, problemreductions-cli/src/commands/create.rs:5079), so the user-facing regression above is not caught by the added test suite.

Minor (Nice to Have)

• None.

Summary

• ISSUE — misleading help/example flow for pred create BottleneckTravelingSalesman; the documented sample instance is infeasible and immediately fails under pred solve.
• ISSUE — new CLI surface area has no automated coverage, despite multiple new branches in problemreductions-cli/src/commands/create.rs.

Agentic Feature Tests

Feature Test Report: BottleneckTravelingSalesman

Project type: CLI + Rust library
Feature tested: New model BottleneckTravelingSalesman
Use Case: Downstream user discovers the new model from the catalog, inspects it, creates a canonical example, and solves/evaluates it from the CLI.
Expected Outcome: The model is discoverable and the documented/example flows produce a valid instance and a valid optimum.
Verdict: fail
Critical Issues: 0

Test Results

• pred list: PASS — BottleneckTravelingSalesman appears in the catalog with complexity O(num_vertices^2 * 2^num_vertices).
• pred show BottleneckTravelingSalesman: PASS — description, fields, and complexity render correctly.
• pred create --example BottleneckTravelingSalesman -o /tmp/btsp-example.json: PASS — emits the canonical K5 instance with type: "BottleneckTravelingSalesman" and variant: {}.
• pred solve /tmp/btsp-example.json --solver brute-force: PASS — returns solution [0,1,1,0,1,0,1,0,0,1] with evaluation: "Valid(4)".
• pred evaluate /tmp/btsp-example.json --config 0,1,1,0,1,0,1,0,0,1: PASS — returns result: "Valid(4)".
• pred create BottleneckTravelingSalesman (help/example discovery): ISSUE — advertises pred create BottleneckTravelingSalesman --graph 0-1,1-2,2-3 --edge-weights 1,1,1 as the example command.
• pred create BottleneckTravelingSalesman --graph 0-1,1-2,2-3 --edge-weights 1,1,1 | pred solve - --solver brute-force: CONFIRMED ISSUE — returns Error: No solution found.

Findings

• Confirmed / Important — the help/example flow for the new model is misleading. The canonical example path works, but the problem-specific example shown by pred create BottleneckTravelingSalesman is infeasible and makes the first obvious user journey fail (problemreductions-cli/src/commands/create.rs:539).

Summary

• The catalog/show/canonical-example/solve flow works correctly in the current PR worktree.
• The ad hoc help-example flow does not: the advertised sample instance has no Hamiltonian tour, so a downstream user following the displayed example will hit Error: No solution found.

Generated by review-pipeline

Agentic Review Report

Structural Check

Structural Review: model BottleneckTravelingSalesman

Structural Completeness

Build Status

• make test: PASS
• make clippy: PASS

Semantic Review

• evaluate() correctness: OK — reuses the existing Hamiltonian-cycle validator and returns the maximum selected edge weight for feasible tours.
• dims() correctness: OK — binary edge encoding via vec![2; num_edges] matches the issue and implementation.
• Size getter consistency: OK — num_vertices and num_edges match the complexity string num_vertices^2 * 2^num_vertices.
• Weight handling: OK — weights are managed through inherent getters/setters rather than trait machinery.

Issue Compliance

Summary

• 17/18 structural checks passed
• 8/8 issue compliance checks passed
• FAIL — deterministic whitelist check flagged extra scope in problemreductions-cli/src/cli.rs, problemreductions-cli/src/commands/create.rs, and src/lib.rs

Quality Check

Quality Review

Design Principles

• DRY: OK — the new model mirrors TravelingSalesman where expected, but the changed objective and fixed schema justify a separate implementation.
• KISS: OK — the implementation stays straightforward by reusing the existing Hamiltonian-cycle predicate and swapping only the objective aggregator.
• HC/LC: OK — model logic, CLI wiring, and paper/example updates remain separated by layer.

HCI (CLI changed)

• Error messages: OK — the agentic test confirmed the default pred solve failure path suggests --solver brute-force.
• Discoverability: OK — pred list, pred show, and pred create --example expose the new model without extra handholding.
• Consistency: OK — CLI edge-weight handling matches neighboring graph problems such as TravelingSalesman, MaxCut, and MaximumMatching.
• Least surprise: OK — the fixed empty variant for BottleneckTravelingSalesman is reflected consistently in catalog and creation flows.
• Feedback: OK — create/solve flows report enough information to understand what happened.

Test Quality

• Naive test detection: OK
  • The added tests cover creation/getters, valid and invalid evaluation, disconnected subtours, optimization direction, brute-force optimum, serialization, and the paper example.

Issues

Critical (Must Fix)

None.

Important (Should Fix)

None.

Minor (Nice to Have)

• There is still no BottleneckTravelingSalesman-specific quickstart example in README.md or docs/src/cli.md, although the generic CLI discovery flow is sufficient in the current worktree.

Summary

• No correctness, design, or test-quality regressions were reproduced in the current worktree.
• Minor: the feature is discoverable through generic CLI flows, but not yet called out explicitly in the top-level docs.

Agentic Feature Tests

• Verdict: PASS
• Feature: BottleneckTravelingSalesman
• Role: downstream CLI user building from source and trying the new model from docs/self-discovery
• Use case: discover the new model, create its canonical example, and solve it locally
• Expected outcome: the model should appear in the catalog, explain its schema clearly, create a valid example instance, and solve successfully with the documented brute-force fallback

• What I tried:

  • make cli
  • pred list | rg BottleneckTravelingSalesman
  • pred show BottleneckTravelingSalesman
  • pred create --example BottleneckTravelingSalesman -o <tmp>/btsp.json
  • pred solve <tmp>/btsp.json --solver brute-force
  • pred create --example BottleneckTravelingSalesman | pred solve - --solver brute-force
  • pred create BottleneckTravelingSalesman
  • pred solve <tmp>/btsp.json

• Expected vs Actual: matched. The feature was discoverable from the existing CLI/documentation flow, the canonical example materialized correctly, and solving the example produced the expected bottleneck value 4.

• Blocked steps: none

• Confirmed issues: none

• Not reproducible in current worktree: none

• Friction points / doc suggestions:

  • Minor: there is no BottleneckTravelingSalesman-specific quickstart example in README.md or docs/src/cli.md; the feature is still usable via generic pred list / pred show / pred create --example flows.
  • Positive: the default pred solve failure path is well handled because it explicitly tells the user to retry with --solver brute-force.

Generated by review-pipeline

Addendum after the main review comment:

• Important: the problem-specific pred create BottleneckTravelingSalesman help example is currently misleading. In problemreductions-cli/src/commands/create.rs, the advertised sample is --graph 0-1,1-2,2-3 --edge-weights 1,1,1, but that 4-vertex path cannot contain a Hamiltonian cycle, so it is an infeasible BTSP instance by construction. This is a user-facing dead end even though pred create --example BottleneckTravelingSalesman works correctly.
• Important: the new model file defines num_vertices() and num_edges() but does not register a ProblemSizeFieldEntry. That leaves registry/inspect-style size-field metadata empty for BottleneckTravelingSalesman, despite the linked issue and paper entry both treating those as the canonical size fields.

These are additional review findings; the rest of the previously-posted review stands.