fix(cluster): honor --resolution on the community split passes

[keith-oak]

_split_community() calls _partition() without forwarding resolution, so it always re-partitions at the default 1.0. cluster(resolution=...) (the --resolution flag added in 2d783e5) therefore applies to the top-level partition but is silently dropped by both split passes:

• the oversized-community split (cluster.py:214)
• the cohesion re-split (cluster.py:223)

So graphify build --resolution 0.5 partitions the graph at 0.5, then re-partitions its largest communities at 1.0 — the opposite granularity to the one requested.

Change

Thread cluster()'s resolution through to both call sites and give _split_community() a resolution: float = 1.0 parameter.

Default behaviour is unchanged — the default is 1.0 either way, which is what the split passes were already hardcoding. Only runs that pass an explicit --resolution change.

Tests

Two tests in tests/test_cluster.py, both verified to fail without the fix:

• test_split_community_forwards_resolution_to_partition — asserts the kwarg reaches _partition
• test_cluster_forwards_resolution_to_split_passes — end-to-end through cluster() on a graph that trips the oversized-split path

Full suite: 3904 passed, 36 skipped. Two failures on this branch are pre-existing on a clean v8 checkout and unrelated (test_ollama_retry_cap.py needs the openai extra; test_labeling.py::test_label_communities_batches_when_over_batch_size asserts a batch completion order).

[graphify-labs]

Looks safe to merge — no coupling regressions and no blocking issues, checked against the code graph (not a self-assessment).

Graphify reviewed this change.

Looks safe to merge — no coupling regressions and no blocking issues, checked against the code graph (not a self-assessment).

---

Graphify review — findings

This PR threads the resolution parameter through the community-splitting logic in graphify/cluster.py. Previously _split_community always ran its secondary Leiden partition with the default resolution of 1.0; now cluster() passes its own resolution value down to both the oversized-community and low-cohesion split passes, and _split_community forwards it to _partition. Two new tests verify that the resolution is forwarded from _split_community to _partition and from cluster into the split passes.

No blocking issues surfaced.

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 363 functions depend on the 38 functions this change touches.

Health — this change adds coupling hotspots:

• worse: cluster() — 61 callers, 3 callees

Verification — 363 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 210 function(s) in the blast radius were not formally verified this run

· 1 more finding(s) on lines outside this diff (see the check run).