fix(wiki): link index rows to their own article when labels collide (#3032, #3033)

[rohit-jsfreaky]

Fixes #3032
Fixes #3033

What

index.md now links every Communities row and every God Nodes row to the article that row was actually written to. Before this, when two entries shared a label, both rows pointed at the first article and the second file (<Label>_2.md) was written but unreachable.

Both issues are the same defect in two adjacent loops of wiki.py, so they are fixed together — #3032 is the community loop, #3033 the god-node loop.

Why

to_wiki already assigns every article a unique filename via _unique_slug, and keeps them in community_slugs (cid -> slug) and god_articles ((node_id, slug)). _index_md then discarded those and re-derived each link target from resolver, a label -> slug map built with setdefault, so the first label registered wins.

A label -> slug map structurally cannot represent two articles that share a label. Both rows resolved to the same slug, and the correctly-written _2.md article had nothing linking to it — so anything crawling from index.md (the documented entry point) could never reach it.

This is the residual half of #497: that fix stopped the files from overwriting each other; the index linking still went by label.

How

_index_md takes community_slugs and god_articles instead of resolver, and links by the slug each article was actually written under. Since it no longer receives the resolver, the index cannot fall back to a label lookup.

• _md_link_to(label, slug) renders a link from a slug that is already known; _md_link keeps its behaviour and delegates to it, so article prose links are unchanged.
• A god node with no article of its own (id not in G — test_to_wiki_skips_missing_god_node_ids) is absent from the slug map and renders as plain text, matching test_wiki_links_to_nodes_without_articles_are_plain_text, rather than linking to a same-labelled article that isn't it.
• Slug assignment order is untouched, so the case-collision dedup those tests cover is unaffected.

Testing

Both issues' repro scripts, run verbatim, now print what each issue lists as "Expected":

#3032

files: ['Shared_Name.md', 'Shared_Name_2.md', 'index.md']
    - [Shared Name](Shared_Name.md) — 3 nodes
    - [Shared Name](Shared_Name_2.md) — 3 nodes     <- was Shared_Name.md

#3033

files: ['GameState.md', 'GameState_2.md', 'Only_Community.md', 'index.md']
    - [GameState](GameState.md) — 2 connections
    - [GameState](GameState_2.md) — 2 connections   <- was GameState.md

Three regression tests added to tests/test_wiki.py — one per issue, plus a general "every article written is reachable from the index" guard. On v8 without the source change all three fail; with it all pass.

Full suite: 12 failed, 5035 passed, 43 skipped. Those 12 are pre-existing failures on my machine (Windows) in test_hooks / test_install* / test_non_regular_files / test_uninstall_scope / test_watch — none import graphify.wiki, and re-running them with this change reverted gives the same 12. ruff check is clean on both files.

[graphify-labs]

Graphify reviewed this change.

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

Formal verification. 1 change(s) tested, no difference found (not proven).

---

Graphify review — findings

Fixes index rows pointing at the wrong article when two communities or god nodes share a label: _index_md now links each row by the slug the article was actually written under (community_slugs and god_articles) rather than re-deriving it from the label, which collapsed duplicates onto the first-registered file and orphaned the rest (#3032/#3033). Adds _md_link_to to render a link from a known slug, still falling back to plain text when no article exists. Backs this with regression tests asserting every written article is reachable from the index.

No blocking issues surfaced. 3 lower-confidence candidates did not survive cross-model review.

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 245 functions depend on the 77 functions this change touches.

Health — this change adds coupling hotspots:

• new: to_wiki() — 44 callers, 7 callees
• new: dispatch_command() — 2 callers, 122 callees
• new: test_poisoned_manifest_is_healed() — 0 callers, 6 callees

Verification — 245 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: 95 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify \_index\_md.

The verifier did not have enough to check \_index\_md, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: signature changed: no shared positional arity

No difference found (not proven): No behavior difference found in \_md\_link (not a proof).

The verifier ran both versions of \_md\_link on many inputs and saw identical behavior every time. Strong evidence the change is safe, but evidence, not a proof.

Guarantee: Empirical: differential testing (both versions run on many generated inputs). A divergence on an untested input remains possible, so this is 'no counterexample found', not 'proven equivalent'.

Note: An input the sampler did not try could still differ.

Could not verify: Could not verify to\_wiki.

The verifier did not have enough to check to\_wiki, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `output_dir` is annotated `str | Path` — outside the synthesizable primitive/collection set

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