refactor: move the extraction fix out (split into a patch PR)

CHANGELOG.md

@@ -293,18 +293,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   instead of `%C3%A9`. An accented cwd reached the server as a different
   path, and Cursor events and the session-start handoff lookup both use the
   query `cwd`. (#877)
-- Link extraction no longer mints a permanently unresolved row from a
-  directory target. A `relations:` value whose final component is empty
-  (`sessions/`) had the extension appended to nothing and was stored as the
-  literal `sessions/.md`; the same target in a body link or wikilink
-  (`[notes/](notes/)`) stayed extension-less. No page path can match either —
-  page paths carry `.md` and `latest_page_id_for_link` matches exactly — so
-  both sat in `links` with `to_page_id = NULL` and were visible only as
-  `unresolved:` in `ai-memory status`. Both routes now skip a directory
-  target, or a stem-less `.md`, with the existing warning. Unresolved
-  same-project links are reported by `memory_lint` now too: it only knew
-  cross-project dangling edges, so a broken internal link stayed invisible
-  outside the counter. (#911)
+- `memory_lint` now reports unresolved same-project links. Only cross-project
+  dangling edges (`DanglingCrossLink`) were reported, so a link to a missing
+  page in the same project stayed invisible outside the status counter;
+  `ReaderPool::dangling_internal_links` now feeds the same `broken_link`
+  findings. (#911)
 
 ## [2.4.0] - 2026-09-21
 

crates/ai-memory-wiki/src/markdown.rs

@@ -190,18 +190,6 @@ fn log_bounded(value: &str) -> String {
     ai_memory_core::truncate_utf8_bytes(value, RELATION_LOG_FIELD_MAX_BYTES)
 }
 
-/// Whether a link target's final path segment can name a page.
-///
-/// A directory target (trailing `/`, so an empty last segment) and a
-/// stem-less `.md` cannot: the first used to normalize to the literal
-/// `notes/.md` in `relations:` frontmatter, the second to an
-/// extension-less path — both permanently unresolved `links` rows that no
-/// page write could ever repoint.
-fn last_segment_names_a_page(target: &str) -> bool {
-    let last = target.rsplit_once('/').map_or(target, |(_, s)| s);
-    !last.is_empty() && last != ".md"
-}
-
 /// Extract typed relation edges from a page's `relations:` frontmatter
 /// (2.0 item 3):
 ///
@@ -237,18 +225,17 @@ pub fn extract_relation_links(frontmatter: &serde_json::Value) -> Vec<LinkTarget
                 },
             };
             // Same terminal normalization as wikilinks: extension-less
-            // targets gain `.md`; a directory target, a stem-less `.md`, and
-            // anything with a non-md extension are not pages and are skipped.
+            // targets gain `.md`; anything with a non-md extension is
+            // not a page and is skipped.
             let raw_path = raw_path.trim();
             let last = raw_path.rsplit_once('/').map_or(raw_path, |(_, s)| s);
-            let names_a_page = last_segment_names_a_page(raw_path)
-                && (!last.contains('.') || raw_path.ends_with(".md"));
-            if !names_a_page {
-                tracing::warn!(target = %log_bounded(target), "relation target is not a page; skipping");
-                continue;
-            }
             let normalized = if last.contains('.') {
-                raw_path.to_string()
+                if raw_path.ends_with(".md") {
+                    raw_path.to_string()
+                } else {
+                    tracing::warn!(target = %log_bounded(target), "relation target is not a page; skipping");
+                    continue;
+                }
             } else {
                 format!("{raw_path}.md")
             };
@@ -381,20 +368,13 @@ fn normalize_link_target(raw: &str, page_path: &PagePath, wikilink: bool) -> Opt
         return None;
     }
 
-    // A directory target (`notes/`) is not a page: kept, it minted an
-    // extension-less `to_path` (or, for a wikilink, the literal
-    // `notes/.md`) that no page write could ever resolve.
-    if !last_segment_names_a_page(target) {
-        return None;
-    }
-
     let mut target = target.to_string();
     let last_segment = target.rsplit_once('/').map_or(target.as_str(), |(_, s)| s);
     if last_segment.contains('.') {
         if !target.ends_with(".md") {
             return None;
         }
-    } else {
+    } else if wikilink || !last_segment.is_empty() {
         target.push_str(".md");
     }
 
@@ -516,21 +496,6 @@ mod tests {
         assert_eq!(links[0].path.as_str(), "notes/ok.md");
     }
 
-    #[test]
-    fn directory_and_stemless_targets_are_not_pages() {
-        // `sessions/` used to normalize to the literal `sessions/.md` — a
-        // link no page write could ever resolve; the same trailing-slash
-        // form in the body stayed extension-less, which cannot match a page
-        // path either. Both are dropped now.
-        let fm = serde_json::json!({
-            "relations": {"fixes": ["sessions/", "sessions/.md"]}
-        });
-        assert!(extract_relation_links(&fm).is_empty());
-
-        assert!(extract_links("- [notes/](notes/)\n", &page()).is_empty());
-        assert!(extract_links("- [[notes/]]\n", &page()).is_empty());
-    }
-
     #[test]
     fn pages_without_relations_extract_nothing() {
         assert!(extract_relation_links(&serde_json::json!({})).is_empty());

docs/typed-edges.md

@@ -66,8 +66,7 @@ gain `.md`.
 ## Who writes them
 
 - **You**, in any page's frontmatter (the wiki files are plain
-  markdown — edit them and let the watcher reindex the page; `ai-memory
-  reindex` only rebuilds a clean store from the markdown).
+  markdown — edit and `reindex`, or let the watcher pick it up).
 - **The consolidator**, sparingly: both single-page consolidation and
   `memory_consolidate` with `multi_page=true` can preserve a relation
   when the session's evidence states it plainly (a fix