You are a constitutional council ranking individual git commits for ownership allocation.

Compare these two commits. Decide which contributed more lasting value to the project.

Judge substance, not spectacle:
- Prefer correct, lasting design and real bugfixes over churn, formatting, renames, or generated noise.
- Prefer clarity and necessity over sheer line count. A small precise change can beat a large diffuse one.
- Do not favor a side merely because its patch is longer or noisier.
- Weight what the change does for the project, not the contributor's name.

Return ONLY a JSON object: {"winner": "A" or "B", "ratio": "N:M", "explanation": "..."}
The explanation must cite concrete differences in the patches (1-3 sentences).

Side A — contributor: tommy-mor
Side A — commit message:
[3b3d5873] item refactor

Side A — unified diff (full patch):
diff --git a/plan.md b/plan.md
deleted file mode 100644
index 00d6867a1e0ed144a16a020ea037f685ce646c73..0000000000000000000000000000000000000000
--- a/plan.md
+++ /dev/null
@@ -1,155 +0,0 @@
-# Plan: `ItemId` + `RouteContext` (identity vs hrefs)
-
-This document is for **the next agent** to continue the refactor without re-deriving context from chat. It supersedes ad-hoc notes: treat it as the checklist of record until the work lands and this file is deleted or trimmed.
-
-## Goal
-
-- **Identity** (what lives in the reducer graph, votes, indexes) becomes a **structural `ItemId` enum** in `slug-types`, not a canonical `String` / `CanonicalItemUrl` newtype.
-- **Presentation** (tilde / dash display, breadcrumbs) derives from `ItemId` via explicit methods, not string stripping.
-- **Routing** (browser `href`s for public vs room) goes through **`RouteContext`** (started in `server/src/html/routing.rs`) so Maud/handlers do not stitch `/r/…` vs `/~` ad hoc.
-
-**Non-goals for v1 of the migration:** backward-compatible JSONL or dual-read of old canonical strings in the event log (project has accepted breaking changes). If you reintroduce compat, document it here.
-
-## Current state (as of this plan)
-
-- **`CanonicalItemUrl`** (`types/src/paths.rs`): newtype around `String`; `parse` / `parent` / `display_path` / `tilde_tail` / etc. Reducer `ContentState`, `VoteData`, ranking, RPC, search, garden, breadcrumbs all use it or `String` keys derived from it.
-- **`ThreadNav`** (`server/src/html/forum/nav.rs`): encodes scope prefixes for threads and garden URLs; **`RouteContext`** now wraps `ThreadNav` (`server/src/html/routing.rs`, re-exported from `server/src/html/mod.rs`) but **most HTML still takes `&ThreadNav` directly** — migration incomplete.
-- **URL normalization** lives in `types/src/url_normalize.rs` + `canonicalize_item` / `finalize_external_identity_url` in `paths.rs` (YouTube, sorted query params, room path `room_route_segment` in `paths.rs`).
-- **Room HTTP paths** are `/r/{short}{slug}` (fused segment); wire **`room_id`** remains `short/slug` for RPC/events.
-
-## Target architecture
-
-### `ItemId` (types)
-
-Suggested shape (adjust after profiling `Ord` / `Hash` / serde size):
-
-```text
-ItemId::Root                      — tilde ontology root (today `SLUG_TILDE_ONTOLOGY_ROOT`)
-ItemId::Local { segments }        — slug.social ~/… path as Vec<String> (lowercase segments, non-empty for non-root)
-ItemId::External { url: Url }    — normalized `url::Url` (crate `url` already in `slug-types`)
-```
-
-**API surface (minimum):**
-
-- `ItemId::parse(&str) -> Option<ItemId>` — single entry from DSL / user input / legacy wire (internally may call `canonicalize_item` + structured split).
-- `ItemId::to_wire_url(&self) -> String` — only for **external** boundaries if needed (HTTP fetch, rare assertions); avoid using as the primary key once maps use `ItemId`.
-- `parent`, `display_path`, `tilde_tail` / `tilde_http_tail`, `tilde_segments`, `last_segment`, `normalized_storage` — port from `CanonicalItemUrl`.
-- **`Ord` + `Hash` + `Eq`** stable for `BTreeSet` / `HashMap` (see `write_actor` scope-rank snapshots).
-- **`Serialize` / `Deserialize`** — decide **tagged JSON** for any persisted or API-carried structs (e.g. `VoteData` in tests). If RPC must stay stringy for clients, use a **DTO layer** that converts `ItemId` ↔ wire at the boundary only.
-
-**Remove:** `CanonicalItemUrl` type and all `path_types::CanonicalItemUrl` / `slug_types::paths::CanonicalItemUrl` exports once call sites are migrated. **`Borrow<str>`** on the old newtype goes away; update `nav!` / any code that assumed map keys borrowed as `str`.
-
-### `RouteContext` (server HTML)
-
-- **File:** `server/src/html/routing.rs` — **`RouteContext(ThreadNav)`** with `item_href`, `item_href_raw`, `thread_url`, `garden_root_url`, `room_url`, `From`/`Into` `ThreadNav`.
-- **Direction:** new code and refactored Maud should take **`&RouteContext`** (or owned where appropriate) instead of `&ThreadNav` when building links. Long term, **`item_href(&ItemId)`** should not parse strings — it should pattern-match `ItemId` and append tilde tail or `/-/…` external tail using the same rules as today’s `ThreadNav::garden_item_url`.
-
-### Axum / garden routes
-
-- **No** single catch-all route (explicit decision): keep the existing router layout in `server/src/lib.rs`.
-- Room routes stay **`/r/:room_key/...`** with `room_key` fused; parsing via `slug_types::room_id_from_route_segment` / `room_route_segment` in `paths.rs`.
-
-## Phased execution (recommended order)
-
-### Phase 0 — Preconditions (quick)
-
-1. Read **`AGENTS.md`** (UI contract, durability matrix, `RpcCommand` vs `HtmlUiAction`).
-2. Run **`cargo test --workspace`** and **`./scripts/clj-test.sh`** on clean `main` before large diffs; repeat after each phase.
-
-### Phase 1 — `ItemId` in `slug-types` (no server yet)
-
-1. Add **`ItemId`** (new file e.g. `types/src/item_id.rs` **or** inline at bottom of `paths.rs` — see **Module cycle** below).
-2. Implement **`ItemId::parse`** using existing **`canonicalize_item`** + normalization; port **`CanonicalItemUrl`** methods to **`ItemId`** with tests ported from `paths.rs` `#[cfg(test)] mod tests`.
-3. **`GardenItemUrl::from_stored(&ItemId, room_wire)`** (and thread helpers) — build absolute hrefs from structure, not from re-parsing a canonical string.
-4. **`TildeHttpPathTail::to_item_id`** (rename from `to_canonical`) / **`tilde_http_path_to_item_id`**.
-5. **`TildeOntologyPath::from_stored(&ItemId)`**.
-6. Export **`ItemId`** from **`types/src/lib.rs`**; update **`server/src/path_types.rs`** re-exports.
-7. **Delete `CanonicalItemUrl`** and fix all **in-crate** references in `types` only until `cargo test` passes for `slug-types`.
-
-**Module cycle trap:** `item_id.rs` must not `use crate::paths::{...}` if `paths.rs` also imports `ItemId` for `GardenItemUrl` in the same module. **Fix one of:**
-
-- **A)** Put `ItemId` **inside `paths.rs`** below `canonicalize_item` / helpers (simplest, large file), or  
-- **B)** Split **`canonicalize_item`** (+ dash host helpers + `finalize_external_identity_url`) into **`types/src/item_wire.rs`**, then `paths.rs` + `item_id.rs` both depend on `item_wire` only (cleaner, more files).
-
-### Phase 2 — Reducer + ranking (server core)
-
-1. **`server/src/reducer.rs`**: `ContentState` / `GroupState` / **`VoteData`** — replace **`CanonicalItemUrl`** with **`ItemId`** on all maps, sets, deques, vectors.
-2. **`apply_vote`**: normalize `a`/`b` via **`ItemId::parse`** or **`ItemId`**-aware logic (remove string round-trip).
-3. **`apply_ingest_to_content`**: **`dsl`** still yields strings for item titles in statements; normalize to **`ItemId`** at ingest boundary via **`ItemId::parse`** once per item.
-4. **`server/src/ranking.rs`**, **`server/src/scope_rank.rs`**, **`server/src/api/write_actor.rs`** (including **`BTreeSet`** ordering), **`server/src/api/validate.rs`**, **`server/src/api/helpers.rs`** — propagate **`ItemId`**.
-5. **`server/tests/basic.rs`** and any reducer tests constructing **`VoteData`** — use **`ItemId::parse(...).unwrap()`** or helpers.
-
-### Phase 3 — RPC + search + external resolver
-
-1. **`server/src/api/rpc.rs`**: rank/pair/matchup/search payloads; today many paths use **`GardenItemUrl::from_storage_str(item.as_str(), …)`** — switch to **`ItemId`** + **`GardenItemUrl::from_stored(&item_id, …)`** (or equivalent).
-2. **`server/src/html/search.rs`**: scoring uses item path strings — derive from **`ItemId::display_path`** / **`to_wire_url`** only at the scoring boundary if needed.
-3. **`server/src/external_resolver.rs`**: take **`&ItemId`** or **`ItemId::external_url()`** instead of **`&CanonicalItemUrl`**.
-
-### Phase 4 — HTML / Maud
-
-1. **`ThreadNav::garden_item_url`**: overload or replace with **`garden_item_href(&self, item: &ItemId)`** (no `CanonicalItemUrl::parse` inside).
-2. **`RouteContext`**: extend **`item_href(&ItemId)`**; migrate call sites from **`ThreadNav`** to **`RouteContext`** where only link-building is needed (keep **`ThreadNav`** where scope / auth helpers need the full struct).
-3. **`server/src/html/garden.rs`**, **`breadcrumb_path.rs`**, **`forum/*`**, **`editor.rs`**: replace **`CanonicalItemUrl`** with **`ItemId`**; breadcrumbs should walk **`ItemId::parent`** without string `rsplit`.
-4. **`types` JSON types** (`RankRow`, etc.): decide whether **`GardenItemUrl`** stays string for JSON or becomes a structured field; keep **one** wire format for the public API.
-
-### Phase 5 — Cleanup + docs
-
-1. Remove dead **`canonical_path`** / **`breadcrumb_path`** string logic if fully superseded.
-2. Update **`AGENTS.md`** if durability, `POST /ui`, or command surfaces change.
-3. Delete or shrink **`plan.md`** when done.
-
-## File / symbol checklist (non-exhaustive — grep-driven)
-
-Run periodically:
-
-```bash
-rg "CanonicalItemUrl" -g'*.rs'
-rg "path_types::CanonicalItemUrl" -g'*.rs'
-rg "tilde_http_path_to_canonical" -g'*.rs'
-```
-
-**High-touch files (from prior exploration):**
-
-| Area | Files |
-|------|--------|
-| Types | `types/src/paths.rs`, `types/src/lib.rs`, `types/src/url_normalize.rs`, (optional) `types/src/item_id.rs`, `types/src/item_wire.rs` |
-| Server re-exports | `server/src/path_types.rs`, `server/src/canonical_path.rs` |
-| Reducer / ingest | `server/src/reducer.rs`, `server/src/dsl.rs` (parse output types if changed) |
-| Ranking | `server/src/ranking.rs`, `server/src/scope_rank.rs` |
-| Writer / RPC | `server/src/api/write_actor.rs`, `server/src/api/rpc.rs`, `server/src/api/helpers.rs`, `server/src/api/validate.rs` |
-| HTML | `server/src/html/garden.rs`, `server/src/html/breadcrumb_path.rs`, `server/src/html/forum/nav.rs`, `server/src/html/routing.rs`, `server/src/html/search.rs`, `server/src/html/editor.rs`, `server/src/html/forum/ingest.rs`, … |
-| Tests | `server/tests/basic.rs`, `server/tests/integration.rs`, `types/src/paths.rs` tests, Clojure under `test/` if URLs/assertions mention canonical shapes |
-
-## Events / JSONL
-
-- **`Ingest`** events store **`raw` DSL** only — no change required for item identity inside the event.
-- If any future event type stores item ids as strings, migrate to **structured `ItemId` serde** or accept string only at the event boundary with immediate parse into **`ItemId`** on `apply_event`.
-
-## `nav!` macro (`server/src/paths.rs`)
-
-- Macros use **`keypath($key)`** with **`.clone()`** — **`ItemId`** must be **`Clone`** (already for enums). Remove any reliance on **`Borrow<str>`** for map keys.
-
-## Testing gate
-
-After each phase:
-
-```bash
-cargo test --workspace
-./scripts/clj-test.sh
-```
-
-## Risks / gotchas
-
-1. **`Ord` on `ItemId`**: must match prior **`CanonicalItemUrl`** / `String` ordering wherever **`BTreeSet`** is used (e.g. deterministic scope-rank snapshots in **`write_actor`**).
-2. **External `ItemId`**: **`Url`** equality / hashing — normalization is already centralized in **`url_normalize`**; ensure **`ItemId::parse`** always inserts normalized **`Url`** into **`External`**.
-3. **Fake parent URLs** in garden (e.g. **`https://.`** for external root ranking): find all **`parse("https://.")`** style hacks and express as **`ItemId`** or a dedicated sentinel.
-4. **Serde**: tests and any RPC clients that snapshot JSON may need expectation updates if **`VoteData`** shape changes.
-
-## Optional follow-ups (not blocking `ItemId`)
-
-- More **domain normalizers** in **`url_normalize.rs`** (e.g. `music.youtube.com`, Spotify, etc.).
-- **Room wire** vs **HTTP segment** helpers already in **`paths.rs`** (`ROOM_SHORT_ID_LEN`, `room_route_segment`, `room_id_from_route_segment`).
-
----
-
-**End state criteria:** `rg CanonicalItemUrl` returns nothing; reducer maps use **`ItemId`**; HTML link generation for items goes through **`RouteContext` + `ItemId`**; tests and Kaocha green.
diff --git a/server/src/api/helpers.rs b/server/src/api/helpers.rs
index 1b291db83df7364a026f2e147e0a29a70a399371..8cd23a02fa5219d6aa375e766e6b2bd2c7bb7dfb 100644
--- a/server/src/api/helpers.rs
+++ b/server/src/api/helpers.rs
@@ -4,7 +4,8 @@ use axum::{
     Json,
 };
 use sha2::{Digest, Sha256};
-use slug_types::paths::{CanonicalItemUrl, GardenItemUrl};
+use slug_types::paths::GardenItemUrl;
+use slug_types::ItemId;
 use slug_types::*;
 use std::collections::HashMap;
 
@@ -31,12 +32,12 @@ pub fn now_ms() -> i64 {
 }
 
 /// Resolve DSL/user input to a stored canonical item id.
-pub fn resolve_item(item: &str) -> Result<CanonicalItemUrl, String> {
+pub fn resolve_item(item: &str) -> Result<ItemId, String> {
     let canonical = canonicalize_item(item);
     if canonical.is_empty() {
         return Err(format!("empty item path: `{}`", item));
     }
-    Ok(CanonicalItemUrl(canonical))
+    ItemId::parse(&canonical).ok_or_else(|| format!("invalid item path: `{}`", item))
 }
 
 pub fn parse_parent_specs(parent: Option<&String>) -> Vec<String> {
@@ -94,7 +95,7 @@ pub fn paginate_rankings(
     (out_components, out_unranked)
 }
 
-pub fn pick_random_distinct_canonical(items: &[CanonicalItemUrl]) -> Option<(CanonicalItemUrl, CanonicalItemUrl)> {
+pub fn pick_random_distinct_canonical(items: &[ItemId]) -> Option<(ItemId, ItemId)> {
     use rand::seq::SliceRandom;
     if items.len() < 2 {
         return None;
@@ -115,15 +116,15 @@ pub fn pick_random_distinct_canonical(items: &[CanonicalItemUrl]) -> Option<(Can
 }
 
 pub fn is_pair_voted(group: &crate::reducer::GroupState, a: &str, b: &str) -> bool {
-    let a_key = CanonicalItemUrl(a.to_string());
-    let b_key = CanonicalItemUrl(b.to_string());
+    let a_key = ItemId::parse(a).unwrap_or_else(|| ItemId::opaque(a.to_string()));
+    let b_key = ItemId::parse(b).unwrap_or_else(|| ItemId::opaque(b.to_string()));
     let Some(&a_idx) = group.item_to_idx.get(&a_key) else { return false; };
     let Some(&b_idx) = group.item_to_idx.get(&b_key) else { return false; };
     let (i, j) = if a_idx < b_idx { (a_idx, b_idx) } else { (b_idx, a_idx) };
     group.voted_pairs.contains(&(i, j))
 }
 
-pub fn compute_connectivity_stats(group: &crate::reducer::GroupState, pool: &[CanonicalItemUrl]) -> ConnectivityStats {
+pub fn compute_connectivity_stats(group: &crate::reducer::GroupState, pool: &[ItemId]) -> ConnectivityStats {
     let n = pool.len();
 
     let global_idxs: Vec<Option<usize>> = pool
diff --git a/server/src/api/rpc.rs b/server/src/api/rpc.rs
index 0ff2d701dd1e8661f58abd55672cd91280e9491e..079d96eb6f1d717c203b1d4aec09f3916384b919 100644
--- a/server/src/api/rpc.rs
+++ b/server/src/api/rpc.rs
@@ -17,7 +17,7 @@ use crate::{
     dsl,
     events::{Event, Ingest, ThreadCapability},
     identity::{parse_agent, parse_username},
-    path_types::CanonicalItemUrl,
+    path_types::ItemId,
     ranking::{connected_components_from_voted_pairs, ranked_items_subset},
     reducer::{scope_from_room_wire, ReducerState, ScopeId},
     state::{AppState, InviteState},
@@ -150,7 +150,7 @@ fn build_rank_response_for_content(
 
     if !is_global && !specs.is_empty() {
         let none_exist = specs.iter().all(|spec| {
-            let Some(canon) = CanonicalItemUrl::parse(spec) else { return true };
+            let Some(canon) = ItemId::parse(spec) else { return true };
             !content.items.contains(&canon) && !content.item_children.contains_key(&canon)
         });
         if none_exist {
@@ -163,10 +163,10 @@ fn build_rank_response_for_content(
 
     let depth = depth.max(1);
     let rankings = if is_global {
-        let all_items: Vec<CanonicalItemUrl> = content.items.iter().cloned().collect();
+        let all_items: Vec<ItemId> = content.items.iter().cloned().collect();
         crate::scope_rank::build_rankings_for_item_set(content, &all_items)
     } else if specs.is_empty() {
-        crate::scope_rank::build_children_rankings(content, &CanonicalItemUrl::ontology_root())
+        crate::scope_rank::build_children_rankings(content, &ItemId::ontology_root())
     } else if depth > 1 {
         let items = crate::scope_rank::resolve_scope_recursive(content, &specs, depth);
         crate::scope_rank::build_rankings_for_item_set(content, &items)
@@ -359,8 +359,8 @@ async fn rpc_check(
     let mut simulated = { reduced_arc.read().await.clone() };
     simulated.apply_event(event);
 
-    let voted_parents: Vec<CanonicalItemUrl> = {
-        let mut parents: HashSet<CanonicalItemUrl> = HashSet::new();
+    let voted_parents: Vec<ItemId> = {
+        let mut parents: HashSet<ItemId> = HashSet::new();
         for s in &v.doc.statements {
             if let dsl::Stmt::Vote { item1, item2, .. } = s {
                 if let (Ok(a), Ok(b)) = (resolve_item(item1), resolve_item(item2)) {
@@ -369,7 +369,7 @@ async fn rpc_check(
                 }
             }
         }
-        let mut out: Vec<CanonicalItemUrl> = parents.into_iter().collect();
+        let mut out: Vec<ItemId> = parents.into_iter().collect();
         out.sort();
         out
     };
@@ -695,7 +695,7 @@ fn rpc_search(reduced: &ReducerState, q: &str, limit: usize, principal: Option<&
 async fn rpc_get_pair(state: &AppState, room: String, parent_path: String) -> Result<RpcResult, RpcErr> {
     let scope = scope_from_room_wire(&room);
     let reduced_arc = state.reduced.clone();
-    let pool: Vec<CanonicalItemUrl> = {
+    let pool: Vec<ItemId> = {
         let reduced = reduced_arc.read().await;
         let content = content_for_room(&reduced, &room);
         let tmp = if parent_path.trim().is_empty() {
@@ -716,7 +716,7 @@ async fn rpc_get_pair(state: &AppState, room: String, parent_path: String) -> Re
             Some("add items via ingest".into()),
         ));
     }
-    let selected: Option<(CanonicalItemUrl, CanonicalItemUrl)> = {
+    let selected: Option<(ItemId, ItemId)> = {
         let mut reduced = reduced_arc.write().await;
         let content = reduced.content.entry(scope.clone()).or_default();
         let group = &mut content.ranking_group;
@@ -729,16 +729,16 @@ async fn rpc_get_pair(state: &AppState, room: String, parent_path: String) -> Re
                 .filter_map(|it| group.item_to_idx.get(it).copied())
                 .collect();
             let ranked = ranked_items_subset(group, &idxs, 10000, 1e-8);
-            let ranked_set: HashSet<CanonicalItemUrl> = ranked.iter().map(|r| r.item.clone()).collect();
-            let unsorted: Vec<CanonicalItemUrl> = pool
+            let ranked_set: HashSet<ItemId> = ranked.iter().map(|r| r.item.clone()).collect();
+            let unsorted: Vec<ItemId> = pool
                 .iter()
                 .filter(|it| !ranked_set.contains(*it))
                 .cloned()
                 .collect();
-            let mut pick: Option<(CanonicalItemUrl, CanonicalItemUrl)> = None;
+            let mut pick: Option<(ItemId, ItemId)> = None;
             if !unsorted.is_empty() {
                 if let Some(left) = unsorted.choose(&mut rng).cloned() {
-                    let mut candidates: Vec<CanonicalItemUrl> = if !ranked.is_empty() {
+                    let mut candidates: Vec<ItemId> = if !ranked.is_empty() {
                         ranked.iter().map(|r| r.item.clone()).collect()
                     } else {
                         pool.clone()
@@ -865,7 +865,7 @@ pub async fn handle_rpc_batch(
                 } else {
                 let content = content_for_room(&reduced, &room);
                 let item_str = canonicalize_item(&item_path);
-                let item = CanonicalItemUrl(item_str.clone());
+                let item = ItemId::parse(&item_str).unwrap_or_else(|| ItemId::opaque(item_str.clone()));
                 if !content.items.contains(&item) {
                     line_err(
                         "item not found",
@@ -1212,7 +1212,7 @@ pub async fn handle_rpc_batch(
                 }
 
                 let ranked_total = ranked.len();
-                let mut unranked: Vec<CanonicalItemUrl> = content
+                let mut unranked: Vec<ItemId> = content
                     .items
                     .iter()
                     .filter(|it| !group.item_to_idx.contains_key(*it))
@@ -1263,7 +1263,7 @@ pub async fn handle_rpc_batch(
                 } else {
                 let content = content_for_room(&reduced, &room);
                 let item_str = canonicalize_item(&item_path);
-                let item = CanonicalItemUrl(item_str.clone());
+                let item = ItemId::parse(&item_str).unwrap_or_else(|| ItemId::opaque(item_str.clone()));
                 let limit = limit.unwrap_or(50).clamp(1, 200);
                 if !content.items.contains(&item) {
                     line_err(
@@ -1304,7 +1304,7 @@ pub async fn handle_rpc_batch(
                 let content = content_for_room(&reduced, &room);
                 let scope = scope_from_room_wire(&room);
                 let item_str = canonicalize_item(&item_path);
-                let item = CanonicalItemUrl(item_str.clone());
+                let item = ItemId::parse(&item_str).unwrap_or_else(|| ItemId::opaque(item_str.clone()));
                 let entries = content.rank_history.get(&item).cloned().unwrap_or_default();
                 let history: Vec<RankHistoryRow> = entries.iter().map(|e| {
                     let caused_by: Vec<VoteRow> = reduced.ingests_by_id.get(&e.post_id)
@@ -1361,7 +1361,11 @@ pub async fn handle_rpc_batch(
                     line_err(e, h)
                 } else {
                 let content = content_for_room(&reduced, &room);
-                let parents: HashSet<&str> = content.item_children.keys().map(|s| s.as_str()).collect();
+                let parents: HashSet<String> = content
+                    .item_children
+                    .keys()
+                    .map(|k| k.to_storage_string())
+                    .collect();
                 let mut paths: Vec<GardenItemUrl> = content
                     .items
                     .iter()
@@ -1380,11 +1384,11 @@ pub async fn handle_rpc_batch(
                 let content = content_for_room(&reduced, &room);
                 let out: Vec<PathSummary> = content
                     .item_children
-                    .get(&CanonicalItemUrl::ontology_root())
+                    .get(&ItemId::ontology_root())
                     .map(|roots| {
                         let mut v: Vec<PathSummary> = roots.iter()
                             .map(|path| {
-                                let children = content.item_children.get(path.as_str()).map(|s| s.len()).unwrap_or(0);
+                                let children = content.item_children.get(path).map(|s| s.len()).unwrap_or(0);
                                 PathSummary {
                                     path: TildeOntologyPath::from_stored(path),
                                     children,
diff --git a/server/src/api/validate.rs b/server/src/api/validate.rs
index 3c7aa80fd485cf247231fe5c5c7e6fea62f22534..3f657a88ef9328a31e3e29c3bc3bd32bec6fd7a1 100644
--- a/server/src/api/validate.rs
+++ b/server/src/api/validate.rs
@@ -4,7 +4,7 @@ use std::collections::HashSet;
 use crate::{
     canonical_path::canonicalize_tag,
     dsl,
-    path_types::CanonicalItemUrl,
+    path_types::ItemId,
     reducer::{ReducerState, ScopeId},
 };
 use slug_types::paths::GardenItemUrl;
@@ -32,11 +32,11 @@ pub fn validate_ingest_document(
         ScopeId::Public => None,
         _ => reduced.content_for_scope(scope),
     };
-    let item_exists = |key: &CanonicalItemUrl| {
+    let item_exists = |key: &ItemId| {
         scoped_content.map(|c| c.items.contains(key)).unwrap_or(false)
             || public_content.items.contains(key)
     };
-    let body_exists = |key: &CanonicalItemUrl| {
+    let body_exists = |key: &ItemId| {
         scoped_content.map(|c| c.item_bodies.contains_key(key)).unwrap_or(false)
             || public_content.item_bodies.contains_key(key)
     };
@@ -52,7 +52,7 @@ pub fn validate_ingest_document(
     };
 
     let ts = super::helpers::now_ms();
-    let mut defined_in_doc: HashSet<CanonicalItemUrl> = HashSet::new();
+    let mut defined_in_doc: HashSet<ItemId> = HashSet::new();
 
     for s in &doc.statements {
         match s {
diff --git a/server/src/api/write_actor.rs b/server/src/api/write_actor.rs
index f9c3b8bd3fbf8fcb9c035e1a1572fef0b08fa8a9..3e96b015def54921962f74b4cd884808ff6c3bcc 100644
--- a/server/src/api/write_actor.rs
+++ b/server/src/api/write_actor.rs
@@ -10,7 +10,7 @@ use crate::{
     events::{AgentBound, Event, GrantAdded, Ingest, PostRedacted, RoomDeleted, UserRegistered},
     html::JsBuilder,
     identity::parse_agent,
-    path_types::CanonicalItemUrl,
+    path_types::ItemId,
     reducer::{scope_from_room_wire, ReducerState, ScopeId},
     state::AppState,
     write_cmd::WriteCmd,
@@ -90,7 +90,7 @@ async fn broadcast_web_refresh(state: &AppState, room_key: &str, thread_id: &str
 }
 
 fn compute_scope_rank_changes(
-    parent: &CanonicalItemUrl,
+    parent: &ItemId,
     before: &crate::scope_rank::ChildrenRankings,
     after: &crate::scope_rank::ChildrenRankings,
     room_wire: &str,
@@ -99,7 +99,7 @@ fn compute_scope_rank_changes(
     use std::collections::BTreeSet;
     fn build_positions(
         rankings: &crate::scope_rank::ChildrenRankings,
-    ) -> HashMap<CanonicalItemUrl, Option<RankPosition>> {
+    ) -> HashMap<ItemId, Option<RankPosition>> {
         let mut map = HashMap::new();
         for comp in &rankings.component_rankings {
             let total = comp.ranked.len();
@@ -116,7 +116,7 @@ fn compute_scope_rank_changes(
     let before_pos = build_positions(before);
     let after_pos = build_positions(after);
 
-    let all_items: BTreeSet<CanonicalItemUrl> = before_pos
+    let all_items: BTreeSet<ItemId> = before_pos
         .keys()
         .cloned()
         .chain(after_pos.keys().cloned())
@@ -287,8 +287,8 @@ pub async fn writer_actor(mut rx: mpsc::Receiver<WriteCmd>, state: AppState) {
                         .map(|d| reduced.agent_bindings.get(d).is_none())
                         .unwrap_or(false);
 
-                    let voted_parent_scopes: Vec<CanonicalItemUrl> = {
-                        let mut parents: HashSet<CanonicalItemUrl> = HashSet::new();
+                    let voted_parent_scopes: Vec<ItemId> = {
+                        let mut parents: HashSet<ItemId> = HashSet::new();
                         for s in &v.doc.statements {
                             if let dsl::Stmt::Vote { item1, item2, .. } = s {
                                 if let (Ok(a), Ok(b)) = (resolve_item(item1), resolve_item(item2)) {
@@ -301,12 +301,12 @@ pub async fn writer_actor(mut rx: mpsc::Receiver<WriteCmd>, state: AppState) {
                                 }
                             }
                         }
-                        let mut out: Vec<CanonicalItemUrl> = parents.into_iter().collect();
+                        let mut out: Vec<ItemId> = parents.into_iter().collect();
                         out.sort();
                         out
                     };
 
-                    let pre_rankings: HashMap<CanonicalItemUrl, crate::scope_rank::ChildrenRankings> =
+                    let pre_rankings: HashMap<ItemId, crate::scope_rank::ChildrenRankings> =
                         if !voted_parent_scopes.is_empty() {
                             let content = content_for_room(&reduced, &room_key);
                             voted_parent_scopes
diff --git a/server/src/external_resolver.rs b/server/src/external_resolver.rs
index 87409c4cab5f1930d2cd075f3417fa824fd791c7..6f5c982a627f9f620f0e3be0ba0cb92a3d7d4bb7 100644
--- a/server/src/external_resolver.rs
+++ b/server/src/external_resolver.rs
@@ -1,6 +1,6 @@
 use async_trait::async_trait;
 
-use crate::path_types::CanonicalItemUrl;
+use crate::path_types::ItemId;
 
 #[async_trait]
 pub trait ExternalResolver: Send + Sync {
@@ -11,7 +11,7 @@ pub trait ExternalResolver: Send + Sync {
     fn normalize(&self, path: &str) -> String;
 
     /// Fetches body when missing; GitHub hook lands here in a follow-up.
-    async fn fetch_body(&self, canonical_url: &CanonicalItemUrl) -> Result<String, String>;
+    async fn fetch_body(&self, canonical_url: &ItemId) -> Result<String, String>;
 }
 
 /// Placeholder until domain-specific resolvers exist.
@@ -27,7 +27,7 @@ impl ExternalResolver for DefaultExternalResolver {
         path.to_string()
     }
 
-    async fn fetch_body(&self, _canonical_url: &CanonicalItemUrl) -> Result<String, String> {
+    async fn fetch_body(&self, _canonical_url: &ItemId) -> Result<String, String> {
         Err("external fetch not implemented".to_string())
     }
 }
diff --git a/server/src/html/breadcrumb_path.rs b/server/src/html/breadcrumb_path.rs
index c8a3937923161a6ff248bd77e87eff0dc6fe9ab0..5743a98753f579c70e10961278469afd7cb9ddcf 100644
--- a/server/src/html/breadcrumb_path.rs
+++ b/server/src/html/breadcrumb_path.rs
@@ -1,8 +1,8 @@
-use crate::path_types::{tilde_http_path_to_canonical, CanonicalItemUrl};
+use crate::path_types::{tilde_http_path_to_item_id, ItemId};
 
 /// Semantic view of an ontology path for rendering and routing decisions.
 pub(super) struct OntologyPath {
-    canonical: CanonicalItemUrl,
+    canonical: ItemId,
     /// Breadcrumb segments: for `~/a/b` this is `["a", "b"]` (leading `~` rendered separately).
     segments: Vec<String>,
 }
@@ -11,11 +11,11 @@ impl OntologyPath {
     /// Path is the `*path` segment from `/~/*path` (e.g. `topic/a`). Always treat it as under `~/`
     /// so it canonicalizes to `https://slug.social/~/…`, not the non-tilde site path.
     pub(super) fn from_input(path: &str) -> Self {
-        let canonical = tilde_http_path_to_canonical(path);
+        let canonical = tilde_http_path_to_item_id(path);
         Self::from_canonical(canonical)
     }
 
-    pub(super) fn from_canonical(canonical: CanonicalItemUrl) -> Self {
+    pub(super) fn from_canonical(canonical: ItemId) -> Self {
         // tilde_segments() returns ["~", "a", "b"] but bc_path() renders "~" itself,
         // so we skip the leading "~" segment here.
         let segments = canonical
@@ -28,7 +28,7 @@ impl OntologyPath {
     }
 
     pub(super) fn root() -> Self {
-        Self::from_canonical(CanonicalItemUrl::ontology_root())
+        Self::from_canonical(ItemId::ontology_root())
     }
 
     pub(super) fn is_root(&self) -> bool {
@@ -56,7 +56,7 @@ impl OntologyPath {
 
 /// External `https://host/…` items addressed as `/-/host/…` in the URL bar.
 pub(super) struct ExternalOntologyPath {
-    canonical: CanonicalItemUrl,
+    canonical: ItemId,
     /// e.g. `["github.com", "org", "repo", "issues"]`
     segments: Vec<String>,
 }
@@ -71,13 +71,13 @@ impl ExternalOntologyPath {
         } else {
             format!("-/{}", p.trim_start_matches('/'))
         };
-        let Some(canonical) = CanonicalItemUrl::parse(&raw) else {
-            return Self::from_canonical(CanonicalItemUrl("https://.".to_string()));
+        let Some(canonical) = ItemId::parse(&raw) else {
+            return Self::from_canonical(ItemId::opaque("https://.".to_string()));
         };
         Self::from_canonical(canonical)
     }
 
-    pub(super) fn from_canonical(canonical: CanonicalItemUrl) -> Self {
+    pub(super) fn from_canonical(canonical: ItemId) -> Self {
         let s = canonical.as_str();
         let rest = s
             .strip_prefix("https://")
diff --git a/server/src/html/editor.rs b/server/src/html/editor.rs
index 80657ac0df9b4bb67e4cdacca296f70267840bd1..c76417acd7a164fda23537cd4e7fb9f1f1dcd895 100644
--- a/server/src/html/editor.rs
+++ b/server/src/html/editor.rs
@@ -97,7 +97,7 @@ pub async fn editor_check(
             simulated.apply_event(event);
 
             // Collect voted parent scopes.
-            let voted_parents: Vec<crate::path_types::CanonicalItemUrl> = {
+            let voted_parents: Vec<crate::path_types::ItemId> = {
                 let mut parents = std::collections::HashSet::new();
                 for s in &v.doc.statements {
                     if let crate::dsl::Stmt::Vote { item1, item2, .. } = s {
@@ -107,7 +107,7 @@ pub async fn editor_check(
                         }
                     }
                 }
-                let mut out: Vec<crate::path_types::CanonicalItemUrl> = parents.into_iter().collect();
+                let mut out: Vec<crate::path_types::ItemId> = parents.into_iter().collect();
                 out.sort();
                 out
             };
diff --git a/server/src/html/forum/nav.rs b/server/src/html/forum/nav.rs
index 48fe11e46731670874ff8b6b05baa6f09ae0b7e4..7ca5b491f38104ec8c81db43a95c59da00abd8c4 100644
--- a/server/src/html/forum/nav.rs
+++ b/server/src/html/forum/nav.rs
@@ -52,9 +52,14 @@ impl ThreadNav {
     }
 
     pub(crate) fn garden_item_url(&self, item: &str) -> String {
-        let Some(c) = crate::path_types::CanonicalItemUrl::parse(item) else {
+        let Some(c) = crate::path_types::ItemId::parse(item) else {
             return format!("{}/{}", self.garden_path_prefix, canonicalize_item(item));
         };
+        self.garden_item_href(&c)
+    }
+
+    /// Relative href for a structured [`crate::path_types::ItemId`] in this scope’s garden.
+    pub(crate) fn garden_item_href(&self, c: &crate::path_types::ItemId) -> String {
         if let Some(tail) = c.tilde_tail().map(str::to_owned) {
             format!("{}/{}", self.garden_path_prefix, tail)
         } else if c.as_str().starts_with("http://") || c.as_str().starts_with("https://") {
@@ -63,7 +68,7 @@ impl ThreadNav {
             let ext_prefix = format!("{}-", self.garden_path_prefix.trim_end_matches('~'));
             format!("{}/{}", ext_prefix, rest)
         } else {
-            format!("{}/{}", self.garden_path_prefix, canonicalize_item(item))
+            format!("{}/{}", self.garden_path_prefix, canonicalize_item(c.as_str()))
         }
     }
 
diff --git a/server/src/html/garden.rs b/server/src/html/garden.rs
index e615dd356bcf634232d85610c0a26235ead125fd..2af9ac5bfdee1630883e6f8257883baeafcb5a44 100644
--- a/server/src/html/garden.rs
+++ b/server/src/html/garden.rs
@@ -10,7 +10,7 @@ use crate::{
     api::optional_principal,
     canonical_path::canonicalize_item,
     events::ThreadCapability,
-    path_types::CanonicalItemUrl,
+    path_types::ItemId,
     reducer::{ContentState, ReducerState, ScopeId},
     ranking::{connected_components_from_voted_pairs, ranked_items_subset},
     scope_rank::{build_children_rankings, ChildrenRankings},
@@ -27,7 +27,7 @@ use super::{
 
 /// Display path for an item: `~/…` or `-/…` form.
 fn item_display_path(item: &str) -> String {
-    CanonicalItemUrl::parse(item)
+    ItemId::parse(item)
         .map(|c| c.display_path())
         .unwrap_or_else(|| canonicalize_item(item))
 }
@@ -159,7 +159,7 @@ pub async fn garden_index(
     let nav = ThreadNav::public();
     let child_rankings = {
         let reduced = state.reduced.read().await;
-        build_children_rankings(reduced.public(), &CanonicalItemUrl::ontology_root())
+        build_children_rankings(reduced.public(), &ItemId::ontology_root())
     };
 
     let page = layout(
@@ -236,7 +236,7 @@ pub async fn external_garden_index(
 ) -> impl IntoResponse {
     let nav = ThreadNav::public();
     let ext_path = ExternalOntologyPath::from_input("");
-    let parent = CanonicalItemUrl::parse("https://.").unwrap();
+    let parent = ItemId::parse("https://.").unwrap();
     let child_rankings = {
         let reduced = state.reduced.read().await;
         build_children_rankings(reduced.public(), &parent)
@@ -365,7 +365,7 @@ pub async fn room_external_garden_index(
         return room_not_found_page(&jar, &uri).into_response();
     }
     let ext_path = ExternalOntologyPath::from_input("");
-    let parent = CanonicalItemUrl::parse("https://.").unwrap();
+    let parent = ItemId::parse("https://.").unwrap();
     let child_rankings = build_children_rankings(
         content_for_garden_view(&reduced, &nav.scope()),
         &parent,
@@ -509,13 +509,13 @@ struct ItemPageViewModel {
 fn build_sibling_rank(
     reduced: &crate::reducer::ReducerState,
     scope: &ScopeId,
-    item: &CanonicalItemUrl,
+    item: &ItemId,
 ) -> Option<SiblingRank> {
     let item = item.clone().normalized_storage();
     let content = content_for_garden_view(reduced, scope);
     let group = &content.ranking_group;
     let parent = item.parent()?.normalized_storage();
-    let siblings: Vec<CanonicalItemUrl> = content
+    let siblings: Vec<ItemId> = content
         .item_children
         .get(&parent)
         .map(|s| s.iter().cloned().collect())
@@ -574,7 +574,7 @@ fn build_rank_history(
     item: &str,
 ) -> Vec<RankHistoryEntryView> {
     let content = content_for_garden_view(reduced, scope);
-    let item_key = CanonicalItemUrl(item.to_string());
+    let item_key = ItemId::parse(item).unwrap_or_else(|| ItemId::opaque(item.to_string()));
     let entries = match content.rank_history.get(&item_key) {
         None => return vec![],
         Some(e) => e,
@@ -592,8 +592,8 @@ fn build_rank_history(
                         if a_str == item || b_str == item {
                             Some(crate::reducer::VoteData {
                                 ts: e.ts,
-                                a: CanonicalItemUrl(a_str),
-                                b: CanonicalItemUrl(b_str),
+                                a: ItemId::parse(&a_str).unwrap_or_else(|| ItemId::opaque(a_str)),
+                                b: ItemId::parse(&b_str).unwrap_or_else(|| ItemId::opaque(b_str)),
                                 ratio_left, ratio_right,
                                 body: explanation,
                                 principal: reduced.ingests_by_id.get(&e.post_id)
@@ -629,8 +629,8 @@ fn build_item_page_view_model(
     item: &str,
 ) -> ItemPageViewModel {
     let content = content_for_garden_view(reduced, scope);
-    let item_key = CanonicalItemUrl::parse(item)
-        .unwrap_or_else(|| CanonicalItemUrl::parse("~/").unwrap())
+    let item_key = ItemId::parse(item)
+        .unwrap_or_else(|| ItemId::parse("~/").unwrap())
         .normalized_storage();
     let item_has_parent = item_key.parent().is_some();
     let child_rankings = build_children_rankings(content, &item_key);
@@ -925,10 +925,10 @@ mod tests {
             .map(|r| r.item.as_str())
             .collect();
         assert_eq!(names, vec!["https://slug.social/~/topic/a", "https://slug.social/~/topic/b"]);
-        use crate::path_types::CanonicalItemUrl;
+        use crate::path_types::ItemId;
         assert!(
-            model.child_rankings.unranked_items.contains(&CanonicalItemUrl("https://slug.social/~/topic/kid1".to_string()))
-                || model.child_rankings.unranked_items.contains(&CanonicalItemUrl("https://slug.social/~/topic/kid2".to_string()))
+            model.child_rankings.unranked_items.contains(&ItemId::parse("https://slug.social/~/topic/kid1").unwrap())
+                || model.child_rankings.unranked_items.contains(&ItemId::parse("https://slug.social/~/topic/kid2").unwrap())
         );
     }
 
@@ -941,8 +941,8 @@ mod tests {
             "9ab12cd/my-room",
             "@00000000-0000-0000-0000-000000000000:test:local/test\n~/t1 {a}\n~/t2 {b}\n",
         );
-        use crate::path_types::CanonicalItemUrl;
-        let root = CanonicalItemUrl::ontology_root();
+        use crate::path_types::ItemId;
+        let root = ItemId::ontology_root();
         let model = build_item_page_view_model(
             &reduced,
             &ScopeId::Room("9ab12cd/my-room".to_string()),
@@ -971,8 +971,8 @@ mod tests {
             "@00000000-0000-0000-0000-000000000000:test:local/test\n\
              ~/a {a}\n~/b {b}\n~/a 2:1 ~/b {because}\n",
         );
-        use crate::path_types::CanonicalItemUrl;
-        let root = CanonicalItemUrl::ontology_root();
+        use crate::path_types::ItemId;
+        let root = ItemId::ontology_root();
         let model = build_item_page_view_model(
             &reduced,
             &ScopeId::Room("9ab12cd/my-room".to_string()),
diff --git a/server/src/html/routing.rs b/server/src/html/routing.rs
index 13e9ae0cbe32d454e34a7737edf8234d2a558502..df545a8195aa97484ef0012cff31d802ac67d9df 100644
--- a/server/src/html/routing.rs
+++ b/server/src/html/routing.rs
@@ -4,7 +4,7 @@
 //! Prefer `RouteContext::item_href` / [`RouteContext::thread_url`] in new Maud over stitching
 //! `/r/…` vs `/~` manually. Call sites can migrate incrementally from passing `&ThreadNav`.
 
-use crate::path_types::CanonicalItemUrl;
+use crate::path_types::ItemId;
 
 use super::forum::ThreadNav;
 
@@ -32,9 +32,9 @@ impl RouteContext {
         self.0
     }
 
-    /// Relative path for a stored canonical item in this scope’s garden.
-    pub fn item_href(&self, item: &CanonicalItemUrl) -> String {
-        self.0.garden_item_url(item.as_str())
+    /// Relative path for a stored item in this scope’s garden.
+    pub fn item_href(&self, item: &ItemId) -> String {
+        self.0.garden_item_href(item)
     }
 
     /// Same as [`Self::item_href`] but parses `item` first (raw DSL / user paste).
diff --git a/server/src/path_types.rs b/server/src/path_types.rs
index 361a9d446c9043cbdea5f060db2e8633c2fd9bf9..a6e0b90a1aaead0bc1f7e0f4c5296506edf3f7ac 100644
--- a/server/src/path_types.rs
+++ b/server/src/path_types.rs
@@ -1,5 +1,6 @@
-//! Re-exports — implementations live in `slug-types` (`paths` module).
+//! Re-exports — implementations live in `slug-types` (`paths` / `item_id` modules).
 
+pub use slug_types::ItemId;
 pub use slug_types::paths::{
-    tilde_http_path_to_canonical, CanonicalItemUrl, RelativePath, TildeHttpPathTail, TildePath,
+    tilde_http_path_to_item_id, RelativePath, TildeHttpPathTail, TildePath,
 };
diff --git a/server/src/ranking.rs b/server/src/ranking.rs
index b24e80ddd5f34b4e454bb81b541e8f54bd97a541..3710c9f64437f5bef3b2121905b6f3bcb7611047 100644
--- a/server/src/ranking.rs
+++ b/server/src/ranking.rs
@@ -1,11 +1,11 @@
 use std::collections::HashMap;
 
-use crate::path_types::CanonicalItemUrl;
+use crate::path_types::ItemId;
 use crate::reducer::GroupState;
 
 #[derive(Debug, Clone)]
 pub struct RankedItem {
-    pub item: CanonicalItemUrl,
+    pub item: ItemId,
     pub score: f64,
 }
 
@@ -239,7 +239,7 @@ pub fn group_summary_scores(
     group: &mut GroupState,
     max_iters: usize,
     tol: f64,
-) -> HashMap<CanonicalItemUrl, f64> {
+) -> HashMap<ItemId, f64> {
     ranked_items(group, max_iters, tol)
         .into_iter()
         .map(|r| (r.item, r.score))
@@ -256,11 +256,11 @@ mod tests {
     }
 
     fn vote(ts: i64, a: &str, b: &str, l: i32, r: i32) -> VoteData {
-        use crate::path_types::CanonicalItemUrl;
+        use crate::path_types::ItemId;
         VoteData {
             ts,
-            a: CanonicalItemUrl(a.to_string()),
-            b: CanonicalItemUrl(b.to_string()),
+            a: ItemId::parse(a).unwrap(),
+            b: ItemId::parse(b).unwrap(),
             ratio_left: l,
             ratio_right: r,
             body: "because".to_string(),
diff --git a/server/src/reducer.rs b/server/src/reducer.rs
index a8651994f2c062af25b6ad792a014dcc083296f9..cd11a5391de8e36b1c3bb6c9b8067ceba0878e31 100644
--- a/server/src/reducer.rs
+++ b/server/src/reducer.rs
@@ -5,7 +5,7 @@ use serde::{Deserialize, Serialize};
 use crate::canonical_path::canonicalize_tag;
 use crate::dsl;
 use crate::events::{Event, Ingest, ThreadCapability};
-use crate::path_types::CanonicalItemUrl;
+use crate::path_types::ItemId;
 
 #[derive(Debug, Clone, Hash, PartialEq, Eq, PartialOrd, Ord)]
 pub enum ScopeId {
@@ -27,8 +27,8 @@ pub fn scope_from_room_wire(room: &str) -> ScopeId {
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
 pub struct VoteData {
     pub ts: i64,
-    pub a: CanonicalItemUrl,
-    pub b: CanonicalItemUrl,
+    pub a: ItemId,
+    pub b: ItemId,
     pub ratio_left: i32,
     pub ratio_right: i32,
     pub body: String,
@@ -40,8 +40,8 @@ pub struct VoteData {
 
 #[derive(Debug, Clone)]
 pub struct GroupState {
-    pub item_to_idx: HashMap<CanonicalItemUrl, usize>,
-    pub idx_to_item: Vec<CanonicalItemUrl>,
+    pub item_to_idx: HashMap<ItemId, usize>,
+    pub idx_to_item: Vec<ItemId>,
 
     /// Aggregated directed edge weights: (src_idx, dst_idx) -> weight.
     pub edges: HashMap<(usize, usize), f64>,
@@ -72,7 +72,7 @@ impl GroupState {
         }
     }
 
-    fn ensure_item(&mut self, item: &CanonicalItemUrl) -> usize {
+    fn ensure_item(&mut self, item: &ItemId) -> usize {
         if let Some(&idx) = self.item_to_idx.get(item) {
             return idx;
         }
@@ -85,11 +85,11 @@ impl GroupState {
 
     /// Public test helper: insert an item into the group without a vote (for unit tests).
     pub fn ensure_item_pub(&mut self, item: &str) -> usize {
-        if let Some(canon) = CanonicalItemUrl::parse(item) {
+        if let Some(canon) = ItemId::parse(item) {
             self.ensure_item(&canon)
         } else {
             // Fallback: treat as raw canonical string
-            let canon = CanonicalItemUrl(item.to_string());
+            let canon = ItemId::opaque(item.to_string());
             self.ensure_item(&canon)
         }
     }
@@ -103,8 +103,8 @@ impl GroupState {
     }
 
     pub fn apply_vote(&mut self, mut vote: VoteData) {
-        vote.a = CanonicalItemUrl::parse(vote.a.as_str()).unwrap_or(vote.a);
-        vote.b = CanonicalItemUrl::parse(vote.b.as_str()).unwrap_or(vote.b);
+        vote.a = ItemId::parse(vote.a.as_str()).unwrap_or_else(|| vote.a.clone());
+        vote.b = ItemId::parse(vote.b.as_str()).unwrap_or_else(|| vote.b.clone());
         vote.thread_tag = canonicalize_tag(&vote.thread_tag);
         if vote.ratio_left < 0 {
             vote.ratio_left = 0;
@@ -210,18 +210,18 @@ impl Default for ForumThreadState {
 #[derive(Debug, Clone, Default)]
 pub struct ContentState {
     pub ranking_group: GroupState,
-    pub items: HashSet<CanonicalItemUrl>,
-    pub item_bodies: HashMap<CanonicalItemUrl, String>,
+    pub items: HashSet<ItemId>,
+    pub item_bodies: HashMap<ItemId, String>,
     /// Parent canonical URL -> direct children.
-    pub item_children: HashMap<CanonicalItemUrl, HashSet<CanonicalItemUrl>>,
+    pub item_children: HashMap<ItemId, HashSet<ItemId>>,
     /// Per-item vote history (most recent first).
-    pub item_votes: HashMap<CanonicalItemUrl, VecDeque<VoteData>>,
+    pub item_votes: HashMap<ItemId, VecDeque<VoteData>>,
     /// Per-item ingest references (most recent first).
-    pub item_snippets: HashMap<CanonicalItemUrl, VecDeque<String>>,
+    pub item_snippets: HashMap<ItemId, VecDeque<String>>,
     /// Item path -> threads that mention or vote on this item.
-    pub item_threads: HashMap<CanonicalItemUrl, HashSet<String>>,
+    pub item_threads: HashMap<ItemId, HashSet<String>>,
     /// Per-item rank history, oldest first.
-    pub rank_history: HashMap<CanonicalItemUrl, Vec<RankHistoryEntry>>,
+    pub rank_history: HashMap<ItemId, Vec<RankHistoryEntry>>,
 }
 
 #[derive(Debug, Clone)]
@@ -351,7 +351,7 @@ impl ReducerState {
     /// For `a/b/c/d` this creates: `a/b/c→d`, `a/b→a/b/c`, `a→a/b`, `""→a`.
     /// Stops early when an intermediate is already registered (its ancestors must be too).
     /// @e2bdefa9-a6fa-4725-b0a2-c0b09d95bb20:claudecode:anthropic/claude-opus-4
-    fn add_child_edge(content: &mut ContentState, item: &CanonicalItemUrl) {
+    fn add_child_edge(content: &mut ContentState, item: &ItemId) {
         let mut child = item.clone();
         loop {
             let Some(parent) = child.parent() else { break };
@@ -366,16 +366,16 @@ impl ReducerState {
     }
 
     /// Resolve an item path as a first-class canonical path.
-    fn normalize_item(item: &str) -> Option<CanonicalItemUrl> {
-        CanonicalItemUrl::parse(item)
+    fn normalize_item(item: &str) -> Option<ItemId> {
+        ItemId::parse(item)
     }
 
     /// 1-indexed rank of `item` within its connected component in the parent scope.
     /// 0 if the item has no votes connecting it to siblings (unranked).
     fn scope_rank_of(
         group: &GroupState,
-        item: &CanonicalItemUrl,
-        item_children: &HashMap<CanonicalItemUrl, HashSet<CanonicalItemUrl>>,
+        item: &ItemId,
+        item_children: &HashMap<ItemId, HashSet<ItemId>>,
     ) -> usize {
         let scope = match item.parent() {
             Some(p) => p,
@@ -421,7 +421,7 @@ impl ReducerState {
     /// 1-indexed position of `item` in the component-aware global flat list.
     /// Components sorted largest-first; items ranked within each component.
     /// 0 if the item is not in the ranking group.
-    fn global_rank_of(group: &GroupState, item: &CanonicalItemUrl) -> usize {
+    fn global_rank_of(group: &GroupState, item: &ItemId) -> usize {
         if !group.item_to_idx.contains_key(item) {
             return 0;
         }
@@ -448,7 +448,7 @@ impl ReducerState {
         let doc = dsl::parse_full(&ing.raw).map_err(|_| ())?;
         let canonical_thread = canonicalize_tag(&ing.thread_tag);
 
-        let voted_items: Vec<CanonicalItemUrl> = doc
+        let voted_items: Vec<ItemId> = doc
             .statements
             .iter()
             .filter_map(|s| {
@@ -467,7 +467,7 @@ impl ReducerState {
         let principal = ing.principal.clone();
         let delegate = ing.delegate.clone();
 
-        let before: HashMap<CanonicalItemUrl, (usize, usize)> = if !voted_items.is_empty() {
+        let before: HashMap<ItemId, (usize, usize)> = if !voted_items.is_empty() {
             crate::ranking::compute_group_ranking(&mut content.ranking_group, 10000, 1e-8);
             voted_items
                 .iter()
@@ -485,7 +485,7 @@ impl ReducerState {
             HashMap::new()
         };
 
-        let mut ingest_items: HashSet<CanonicalItemUrl> = HashSet::new();
+        let mut ingest_items: HashSet<ItemId> = HashSet::new();
 
         for stmt in doc.statements {
             match stmt {
diff --git a/server/src/scope_rank.rs b/server/src/scope_rank.rs
index d656b2f0a0a3623ca6b234b746beaaca8ae6017d..4fdaae150c74ec5162c2accc06d4cc778ad79914 100644
--- a/server/src/scope_rank.rs
+++ b/server/src/scope_rank.rs
@@ -3,7 +3,7 @@
 
 use std::collections::{HashMap, HashSet};
 
-use crate::path_types::CanonicalItemUrl;
+use crate::path_types::ItemId;
 use crate::ranking::{connected_components_from_voted_pairs, ranked_items_subset, RankedItem};
 use crate::reducer::ContentState;
 
@@ -17,12 +17,12 @@ pub struct ScopedComponent {
 pub struct ChildrenRankings {
     pub component_rankings: Vec<ScopedComponent>,
     /// Items in scope with no rank (no votes connecting them to others in this scope).
-    pub unranked_items: Vec<CanonicalItemUrl>,
+    pub unranked_items: Vec<ItemId>,
 }
 
 /// Resolve one scope spec (literal path) to direct children of that parent. No wildcards.
-fn resolve_one_scope(content: &ContentState, spec: &str) -> HashSet<CanonicalItemUrl> {
-    let Some(parent) = CanonicalItemUrl::parse(spec.trim()) else {
+fn resolve_one_scope(content: &ContentState, spec: &str) -> HashSet<ItemId> {
+    let Some(parent) = ItemId::parse(spec.trim()) else {
         return HashSet::new();
     };
     content
@@ -36,12 +36,12 @@ fn resolve_one_scope(content: &ContentState, spec: &str) -> HashSet<CanonicalIte
 
 /// Resolve multiple scope specs (literal paths) to a single merged, deduplicated list of item paths.
 /// Used for rank/pair with multiple parents (explicit merge, e.g. rank ~/models ~/ai-models).
-pub fn resolve_scope(content: &ContentState, specs: &[String]) -> Vec<CanonicalItemUrl> {
+pub fn resolve_scope(content: &ContentState, specs: &[String]) -> Vec<ItemId> {
     let mut set = HashSet::new();
     for spec in specs {
         set.extend(resolve_one_scope(content, spec));
     }
-    let mut out: Vec<CanonicalItemUrl> = set.into_iter().collect();
+    let mut out: Vec<ItemId> = set.into_iter().collect();
     out.sort();
     out
 }
@@ -49,18 +49,18 @@ pub fn resolve_scope(content: &ContentState, specs: &[String]) -> Vec<CanonicalI
 /// Resolve scope specs recursively up to `depth` levels deep.
 /// depth=1 is equivalent to resolve_scope (direct children only).
 /// depth=2 includes grandchildren, etc.
-pub fn resolve_scope_recursive(content: &ContentState, specs: &[String], depth: usize) -> Vec<CanonicalItemUrl> {
+pub fn resolve_scope_recursive(content: &ContentState, specs: &[String], depth: usize) -> Vec<ItemId> {
     if depth == 0 {
         return vec![];
     }
-    let mut visited: HashSet<CanonicalItemUrl> = HashSet::new();
-    let mut frontier: Vec<CanonicalItemUrl> = specs
+    let mut visited: HashSet<ItemId> = HashSet::new();
+    let mut frontier: Vec<ItemId> = specs
         .iter()
-        .filter_map(|s| CanonicalItemUrl::parse(s))
+        .filter_map(|s| ItemId::parse(s))
         .collect();
 
     for _level in 0..depth {
-        let mut next_frontier: Vec<CanonicalItemUrl> = Vec::new();
+        let mut next_frontier: Vec<ItemId> = Vec::new();
         for parent in &frontier {
             if let Some(children) = content.item_children.get(parent) {
                 for child in children {
@@ -76,16 +76,16 @@ pub fn resolve_scope_recursive(content: &ContentState, specs: &[String], depth:
         frontier = next_frontier;
     }
 
-    let mut out: Vec<CanonicalItemUrl> = visited.into_iter().collect();
+    let mut out: Vec<ItemId> = visited.into_iter().collect();
     out.sort();
     out
 }
 
 /// Build connected-component rankings for an explicit set of item paths.
 /// Use this when scope comes from multiple parents (resolve_scope).
-pub fn build_rankings_for_item_set(content: &ContentState, items_in_scope: &[CanonicalItemUrl]) -> ChildrenRankings {
+pub fn build_rankings_for_item_set(content: &ContentState, items_in_scope: &[ItemId]) -> ChildrenRankings {
     let group = &content.ranking_group;
-    let mut items_in_scope: Vec<CanonicalItemUrl> = items_in_scope.to_vec();
+    let mut items_in_scope: Vec<ItemId> = items_in_scope.to_vec();
     items_in_scope.sort();
 
     let scoped_idxs: Vec<usize> = items_in_scope
@@ -127,7 +127,7 @@ pub fn build_rankings_for_item_set(content: &ContentState, items_in_scope: &[Can
         })
         .collect();
 
-    let mut unranked_items: Vec<CanonicalItemUrl> = isolate_local_idxs
+    let mut unranked_items: Vec<ItemId> = isolate_local_idxs
         .into_iter()
         .filter_map(|li| local_to_global.get(li).copied())
         .filter_map(|idx| group.idx_to_item.get(idx).cloned())
@@ -148,9 +148,9 @@ pub fn build_rankings_for_item_set(content: &ContentState, items_in_scope: &[Can
 
 /// Build connected-component rankings for direct children of parent_scope.
 /// Matches the HTML garden view: multiple components, isolates, no-vote items.
-pub fn build_children_rankings(content: &ContentState, parent: &CanonicalItemUrl) -> ChildrenRankings {
+pub fn build_children_rankings(content: &ContentState, parent: &ItemId) -> ChildrenRankings {
     let parent = parent.clone().normalized_storage();
-    let items: Vec<CanonicalItemUrl> = content
+    let items: Vec<ItemId> = content
         .item_children
         .get(&parent)
         .map(|s| s.iter().cloned().collect())
@@ -164,10 +164,13 @@ mod tests {
     use std::collections::{HashMap, HashSet};
 
     fn content_with_children(edges: &[(&str, &[&str])]) -> ContentState {
-        let mut item_children: HashMap<CanonicalItemUrl, HashSet<CanonicalItemUrl>> = HashMap::new();
+        let mut item_children: HashMap<ItemId, HashSet<ItemId>> = HashMap::new();
         for (parent, children) in edges {
-            let parent = CanonicalItemUrl((*parent).to_string());
-            let set: HashSet<CanonicalItemUrl> = children.iter().map(|s| CanonicalItemUrl((*s).to_string())).collect();
+            let parent = ItemId::parse(parent).unwrap();
+            let set: HashSet<ItemId> = children
+                .iter()
+                .map(|s| ItemId::parse(s).unwrap())
+                .collect();
             item_children.insert(parent, set);
         }
         ContentState {
@@ -189,8 +192,8 @@ mod tests {
         ]);
         let out = resolve_one_scope(&content, "models");
         assert_eq!(out.len(), 2);
-        assert!(out.contains(&CanonicalItemUrl("https://slug.social/models/x".to_string())));
-        assert!(out.contains(&CanonicalItemUrl("https://slug.social/models/y".to_string())));
+        assert!(out.contains(&ItemId::parse("https://slug.social/models/x").unwrap()));
+        assert!(out.contains(&ItemId::parse("https://slug.social/models/y").unwrap()));
     }
 
     #[test]
@@ -201,9 +204,9 @@ mod tests {
         ]);
         let out = resolve_scope(&content, &["a".into(), "b".into()]);
         assert_eq!(out.len(), 4);
-        assert!(out.contains(&CanonicalItemUrl("https://slug.social/a/1".to_string())));
-        assert!(out.contains(&CanonicalItemUrl("https://slug.social/a/2".to_string())));
-        assert!(out.contains(&CanonicalItemUrl("https://slug.social/b/1".to_string())));
-        assert!(out.contains(&CanonicalItemUrl("https://slug.social/b/2".to_string())));
+        assert!(out.contains(&ItemId::parse("https://slug.social/a/1").unwrap()));
+        assert!(out.contains(&ItemId::parse("https://slug.social/a/2").unwrap()));
+        assert!(out.contains(&ItemId::parse("https://slug.social/b/1").unwrap()));
+        assert!(out.contains(&ItemId::parse("https://slug.social/b/2").unwrap()));
     }
 }
diff --git a/server/tests/basic.rs b/server/tests/basic.rs
index 17db86894f18d6b542d2943f4cedb0572b247171..c8efea2976e20b443d78a5d4c3b244bc4f2fd6da 100644
--- a/server/tests/basic.rs
+++ b/server/tests/basic.rs
@@ -7,8 +7,15 @@ use slugsocial_server::{
 };
 
 
+use slugsocial_server::path_types::ItemId;
+
 use tempfile::TempDir;
 
+#[inline]
+fn item_id(s: &str) -> ItemId {
+    ItemId::parse(s).unwrap()
+}
+
 fn ingest_event(ts: i64, raw: &str) -> Event {
     Event::Ingest(Ingest {
         ts,
@@ -47,7 +54,7 @@ fn reducer_external_namespace_ranking() {
     let g = &content.ranking_group;
     assert_eq!(g.idx_to_item.len(), 2);
 
-    let parent = slugsocial_server::path_types::CanonicalItemUrl::parse("https://github.com/iss").unwrap();
+    let parent = slugsocial_server::path_types::ItemId::parse("https://github.com/iss").unwrap();
     let children = slugsocial_server::scope_rank::build_children_rankings(content, &parent);
     assert_eq!(children.component_rankings.len(), 1);
     let names: Vec<&str> = children.component_rankings[0]
@@ -77,9 +84,9 @@ fn reducer_and_ranking_linear_chain() {
     let mut group = state.public().ranking_group.clone();
     let ranked = ranked_items(&mut group, 20000, 1e-9);
     assert_eq!(ranked.len(), 3);
-    assert_eq!(ranked[0].item, "https://slug.social/~/t/a");
-    assert_eq!(ranked[1].item, "https://slug.social/~/t/b");
-    assert_eq!(ranked[2].item, "https://slug.social/~/t/c");
+    assert_eq!(ranked[0].item.as_str(), "https://slug.social/~/t/a");
+    assert_eq!(ranked[1].item.as_str(), "https://slug.social/~/t/b");
+    assert_eq!(ranked[2].item.as_str(), "https://slug.social/~/t/c");
 }
 
 #[test]
@@ -112,15 +119,15 @@ fn reducer_handles_item_and_body_from_ingest() {
     ));
 
     let content = state.public();
-    assert!(content.items.contains("https://slug.social/~/t/test-item"));
+    assert!(content.items.contains(&item_id("https://slug.social/~/t/test-item")));
     assert_eq!(
-        content.item_bodies.get("https://slug.social/~/t/test-item"),
+        content.item_bodies.get(&item_id("https://slug.social/~/t/test-item")),
         Some(&"Description here".to_string())
     );
     assert!(content
         .item_children
-        .get("https://slug.social/~/t")
-        .map(|c| c.contains("https://slug.social/~/t/test-item"))
+        .get(&item_id("https://slug.social/~/t"))
+        .map(|c| c.contains(&item_id("https://slug.social/~/t/test-item")))
         .unwrap_or(false));
 }
 
@@ -136,12 +143,12 @@ fn reducer_indexes_item_threads_and_vote_thread() {
     state.apply_event(Event::Ingest(ev));
 
     let content = state.public();
-    let threads_for_insertion = content.item_threads.get("https://slug.social/~/sorts/insertion").unwrap();
+    let threads_for_insertion = content.item_threads.get(&item_id("https://slug.social/~/sorts/insertion")).unwrap();
     assert!(threads_for_insertion.contains("sorting-hat"));
-    let threads_for_mergesort = content.item_threads.get("https://slug.social/~/sorts/mergesort").unwrap();
+    let threads_for_mergesort = content.item_threads.get(&item_id("https://slug.social/~/sorts/mergesort")).unwrap();
     assert!(threads_for_mergesort.contains("sorting-hat"));
 
-    let vote = content.item_votes.get("https://slug.social/~/sorts/insertion").unwrap().front().unwrap();
+    let vote = content.item_votes.get(&item_id("https://slug.social/~/sorts/insertion")).unwrap().front().unwrap();
     assert_eq!(vote.thread_tag, "sorting-hat");
 }
 
@@ -155,8 +162,8 @@ fn reducer_aggregates_multiple_votes() {
     }
 
     let group = &state.public().ranking_group;
-    let a_idx = group.item_to_idx["https://slug.social/~/t/a"];
-    let b_idx = group.item_to_idx["https://slug.social/~/t/b"];
+    let a_idx = group.item_to_idx[&item_id("https://slug.social/~/t/a")];
+    let b_idx = group.item_to_idx[&item_id("https://slug.social/~/t/b")];
 
     // Should have accumulated edge weights in both directions.
     assert!(group.edges.contains_key(&(a_idx, b_idx)));
@@ -232,7 +239,7 @@ fn ranking_dominant_item_wins() {
     let mut group = state.public().ranking_group.clone();
     let ranked = ranked_items(&mut group, 20000, 1e-9);
 
-    assert_eq!(ranked[0].item, "https://slug.social/~/t/champion");
+    assert_eq!(ranked[0].item.as_str(), "https://slug.social/~/t/champion");
     assert!(ranked[0].score > ranked[1].score);
 }
 
@@ -379,7 +386,7 @@ async fn full_workflow_reducer_and_ranking() {
     let ranked = ranked_items(&mut group, 20000, 1e-9);
 
     assert_eq!(ranked.len(), 2);
-    assert_eq!(ranked[0].item, "https://slug.social/~/langs/rust"); // Should win
+    assert_eq!(ranked[0].item.as_str(), "https://slug.social/~/langs/rust"); // Should win
     assert!(ranked[0].score > ranked[1].score);
 }
 
@@ -395,27 +402,27 @@ fn reducer_materializes_ancestor_path_segments() {
 
     // The intermediate path "https://slug.social/~/ai-models/anthropic" should appear as a child of "https://slug.social/~/ai-models".
     let content = state.public();
-    let ai_models_children = content.item_children.get("https://slug.social/~/ai-models").expect("ai-models should have children");
+    let ai_models_children = content.item_children.get(&item_id("https://slug.social/~/ai-models")).expect("ai-models should have children");
     assert!(
-        ai_models_children.contains("https://slug.social/~/ai-models/anthropic"),
+        ai_models_children.contains(&item_id("https://slug.social/~/ai-models/anthropic")),
         "ai-models/anthropic should be a child of ai-models"
     );
 
     // The leaf items should still be children of "https://slug.social/~/ai-models/anthropic".
-    let anthropic_children = content.item_children.get("https://slug.social/~/ai-models/anthropic").expect("ai-models/anthropic should have children");
-    assert!(anthropic_children.contains("https://slug.social/~/ai-models/anthropic/claude-opus"));
-    assert!(anthropic_children.contains("https://slug.social/~/ai-models/anthropic/claude-sonnet"));
+    let anthropic_children = content.item_children.get(&item_id("https://slug.social/~/ai-models/anthropic")).expect("ai-models/anthropic should have children");
+    assert!(anthropic_children.contains(&item_id("https://slug.social/~/ai-models/anthropic/claude-opus")));
+    assert!(anthropic_children.contains(&item_id("https://slug.social/~/ai-models/anthropic/claude-sonnet")));
 
     // Root should contain "https://slug.social/~/ai-models".
-    let root_children = content.item_children.get("https://slug.social/~").expect("root should have children");
-    assert!(root_children.contains("https://slug.social/~/ai-models"));
+    let root_children = content.item_children.get(&item_id("https://slug.social/~")).expect("root should have children");
+    assert!(root_children.contains(&item_id("https://slug.social/~/ai-models")));
 
     // The phantom intermediates should NOT be in the items set (they weren't explicitly created).
-    assert!(!content.items.contains("https://slug.social/~/ai-models"));
-    assert!(!content.items.contains("https://slug.social/~/ai-models/anthropic"));
+    assert!(!content.items.contains(&item_id("https://slug.social/~/ai-models")));
+    assert!(!content.items.contains(&item_id("https://slug.social/~/ai-models/anthropic")));
     // But the leaf items should be.
-    assert!(content.items.contains("https://slug.social/~/ai-models/anthropic/claude-opus"));
-    assert!(content.items.contains("https://slug.social/~/ai-models/anthropic/claude-sonnet"));
+    assert!(content.items.contains(&item_id("https://slug.social/~/ai-models/anthropic/claude-opus")));
+    assert!(content.items.contains(&item_id("https://slug.social/~/ai-models/anthropic/claude-sonnet")));
 }
 
 #[test]
@@ -469,9 +476,9 @@ fn ranking_repeated_votes_normalized() {
 
     // Same winner regardless of how many times voted.
     assert_eq!(ranked_once[0].item, ranked_many[0].item);
-    assert_eq!(ranked_once[0].item, "https://slug.social/~/norm/a");
+    assert_eq!(ranked_once[0].item.as_str(), "https://slug.social/~/norm/a");
     assert_eq!(ranked_once[1].item, ranked_many[1].item);
-    assert_eq!(ranked_once[1].item, "https://slug.social/~/norm/b");
+    assert_eq!(ranked_once[1].item.as_str(), "https://slug.social/~/norm/b");
 
     // Scores should be identical (normalization makes repeated votes idempotent).
     let eps = 1e-6;
@@ -506,8 +513,8 @@ fn reducer_zero_zero_vote_ratio_normalizes_to_one_one() {
         "~/t/a {a}\n~/t/b {b}\n~/t/a 0:0 ~/t/b {zero}\n",
     ));
     let group = &state.public().ranking_group;
-    let a_idx = group.item_to_idx["https://slug.social/~/t/a"];
-    let b_idx = group.item_to_idx["https://slug.social/~/t/b"];
+    let a_idx = group.item_to_idx[&item_id("https://slug.social/~/t/a")];
+    let b_idx = group.item_to_idx[&item_id("https://slug.social/~/t/b")];
     // 0:0 should normalize to 1:1 — both directions should have weight
     assert!(group.edges.contains_key(&(a_idx, b_idx)));
     assert!(group.edges.contains_key(&(b_idx, a_idx)));
@@ -523,8 +530,8 @@ fn reducer_negative_ratio_clamped_to_zero() {
     let mut group = GroupState::new();
     group.apply_vote(slugsocial_server::reducer::VoteData {
         ts: 1,
-        a: slugsocial_server::path_types::CanonicalItemUrl("https://slug.social/~/t/a".to_string()),
-        b: slugsocial_server::path_types::CanonicalItemUrl("https://slug.social/~/t/b".to_string()),
+        a: slugsocial_server::path_types::ItemId::parse("https://slug.social/~/t/a").unwrap(),
+        b: slugsocial_server::path_types::ItemId::parse("https://slug.social/~/t/b").unwrap(),
         ratio_left: -5,
         ratio_right: -3,
         body: "negative".to_string(),
@@ -534,8 +541,8 @@ fn reducer_negative_ratio_clamped_to_zero() {
     });
     assert_eq!(group.idx_to_item.len(), 2);
     // Both edges should exist (negatives clamped to 0, then 0:0 -> 1:1)
-    let a_idx = group.item_to_idx["https://slug.social/~/t/a"];
-    let b_idx = group.item_to_idx["https://slug.social/~/t/b"];
+    let a_idx = group.item_to_idx[&item_id("https://slug.social/~/t/a")];
+    let b_idx = group.item_to_idx[&item_id("https://slug.social/~/t/b")];
     assert!(group.edges.contains_key(&(a_idx, b_idx)));
     assert!(group.edges.contains_key(&(b_idx, a_idx)));
 }
@@ -553,24 +560,24 @@ fn reducer_deep_path_ancestor_materialization_four_levels() {
     let content = state.public();
     let tilde_scope = content
         .item_children
-        .get("https://slug.social/~")
+        .get(&item_id("https://slug.social/~"))
         .expect("~/ scope should have children");
-    assert!(tilde_scope.contains("https://slug.social/~/a"));
+    assert!(tilde_scope.contains(&item_id("https://slug.social/~/a")));
 
-    let a_children = content.item_children.get("https://slug.social/~/a").expect("a should have children");
-    assert!(a_children.contains("https://slug.social/~/a/b"));
+    let a_children = content.item_children.get(&item_id("https://slug.social/~/a")).expect("a should have children");
+    assert!(a_children.contains(&item_id("https://slug.social/~/a/b")));
 
-    let ab_children = content.item_children.get("https://slug.social/~/a/b").expect("a/b should have children");
-    assert!(ab_children.contains("https://slug.social/~/a/b/c"));
+    let ab_children = content.item_children.get(&item_id("https://slug.social/~/a/b")).expect("a/b should have children");
+    assert!(ab_children.contains(&item_id("https://slug.social/~/a/b/c")));
 
-    let abc_children = content.item_children.get("https://slug.social/~/a/b/c").expect("a/b/c should have children");
-    assert!(abc_children.contains("https://slug.social/~/a/b/c/d"));
+    let abc_children = content.item_children.get(&item_id("https://slug.social/~/a/b/c")).expect("a/b/c should have children");
+    assert!(abc_children.contains(&item_id("https://slug.social/~/a/b/c/d")));
 
     // Only the leaf should be in items set
-    assert!(content.items.contains("https://slug.social/~/a/b/c/d"));
-    assert!(!content.items.contains("https://slug.social/~/a"));
-    assert!(!content.items.contains("https://slug.social/~/a/b"));
-    assert!(!content.items.contains("https://slug.social/~/a/b/c"));
+    assert!(content.items.contains(&item_id("https://slug.social/~/a/b/c/d")));
+    assert!(!content.items.contains(&item_id("https://slug.social/~/a")));
+    assert!(!content.items.contains(&item_id("https://slug.social/~/a/b")));
+    assert!(!content.items.contains(&item_id("https://slug.social/~/a/b/c")));
 }
 
 // ============================================================================
@@ -610,7 +617,7 @@ fn ranking_convergence_tolerance_triggers_early_exit() {
     // Very tight tolerance but huge max_iters — should still converge fast
     let ranked = ranked_items(&mut group, 1_000_000, 1e-15);
     assert_eq!(ranked.len(), 2);
-    assert_eq!(ranked[0].item, "https://slug.social/~/t/a");
+    assert_eq!(ranked[0].item.as_str(), "https://slug.social/~/t/a");
 }
 
 // ============================================================================
@@ -624,14 +631,14 @@ fn test_item_body_overwrite() {
         1,
         "~/t/x {first}\n",
     ));
-    assert_eq!(state.public().item_bodies.get("https://slug.social/~/t/x"), Some(&"first".to_string()));
+    assert_eq!(state.public().item_bodies.get(&item_id("https://slug.social/~/t/x")), Some(&"first".to_string()));
 
     state.apply_event(ingest_event(
         2,
         "~/t/x {second}\n",
     ));
     assert_eq!(
-        state.public().item_bodies.get("https://slug.social/~/t/x"),
+        state.public().item_bodies.get(&item_id("https://slug.social/~/t/x")),
         Some(&"second".to_string()),
         "last writer should win for item bodies"
     );
@@ -645,9 +652,9 @@ fn test_empty_body_not_stored() {
         "~/t/blank {   }\n",
     ));
     let content = state.public();
-    assert!(content.items.contains("https://slug.social/~/t/blank"), "item should exist");
+    assert!(content.items.contains(&item_id("https://slug.social/~/t/blank")), "item should exist");
     assert!(
-        !content.item_bodies.contains_key("https://slug.social/~/t/blank"),
+        !content.item_bodies.contains_key(&item_id("https://slug.social/~/t/blank")),
         "whitespace-only body should not be stored"
     );
 }
@@ -664,7 +671,7 @@ fn test_duplicate_items_across_ingests() {
         "~/t/dup {second}\n",
     ));
     let content = state.public();
-    let count = content.items.iter().filter(|i| *i == "https://slug.social/~/t/dup").count();
+    let count = content.items.iter().filter(|i| i.as_str() == "https://slug.social/~/t/dup").count();
     assert_eq!(count, 1, "items set should deduplicate across ingests");
 }
 
@@ -736,7 +743,7 @@ fn test_thread_id_is_used_for_votes_and_indexes() {
     let content = state.public();
     let vote = content
         .item_votes
-        .get("https://slug.social/~/t/a")
+        .get(&item_id("https://slug.social/~/t/a"))
         .unwrap()
         .front()
         .unwrap();
@@ -745,7 +752,7 @@ fn test_thread_id_is_used_for_votes_and_indexes() {
     assert!(state
         .public()
         .item_threads
-        .get("https://slug.social/~/t/a")
+        .get(&item_id("https://slug.social/~/t/a"))
         .is_some_and(|threads| threads.contains("first")));
 }
 
@@ -757,11 +764,11 @@ fn test_rank_history_created_for_voted_items() {
         "~/t/a {a}\n~/t/b {b}\n~/t/a 3:1 ~/t/b {reason}\n",
     ));
     assert!(
-        state.public().rank_history.contains_key("https://slug.social/~/t/a"),
+        state.public().rank_history.contains_key(&item_id("https://slug.social/~/t/a")),
         "rank_history should have entry for voted item a"
     );
     assert!(
-        state.public().rank_history.contains_key("https://slug.social/~/t/b"),
+        state.public().rank_history.contains_key(&item_id("https://slug.social/~/t/b")),
         "rank_history should have entry for voted item b"
     );
 }
@@ -774,7 +781,7 @@ fn test_rank_history_not_created_for_unvoted_items() {
         "~/t/c {just a definition}\n",
     ));
     assert!(
-        !state.public().rank_history.contains_key("https://slug.social/~/t/c"),
+        !state.public().rank_history.contains_key(&item_id("https://slug.social/~/t/c")),
         "rank_history should NOT have entry for item with no votes"
     );
 }
@@ -786,7 +793,7 @@ fn test_rank_history_first_entry_delta_zero() {
         1,
         "~/t/a {a}\n~/t/b {b}\n~/t/a 3:1 ~/t/b {reason}\n",
     ));
-    let history_a = state.public().rank_history.get("https://slug.social/~/t/a").unwrap();
+    let history_a = state.public().rank_history.get(&item_id("https://slug.social/~/t/a")).unwrap();
     assert_eq!(history_a.len(), 1);
     assert_eq!(
         history_a[0].scope_rank_delta, 0,
diff --git a/types/src/item_id.rs b/types/src/item_id.rs
new file mode 100644
index 0000000000000000000000000000000000000000..5fe9fad39c0d58e546c0a0162eba4348d64e0865
--- /dev/null
+++ b/types/src/item_id.rs
@@ -0,0 +1,225 @@
+//! Structural item identity for the reducer graph and ranking (vs presentation-only strings).
+
+use std::cmp::Ordering;
+use std::fmt;
+
+use serde::{Deserialize, Deserializer, Serialize, Serializer};
+
+use crate::item_wire::{
+    canonicalize_item, external_display_dash_prefix, normalize_slug_ontology_storage_url,
+    SLUG_TILDE_ONTOLOGY_ROOT,
+};
+
+/// Structural key for items in [`slug_types`] and the server reducer.
+///
+/// Wire / JSON uses the same single string as the former canonical item URL (via serde).
+#[derive(Debug, Clone, Hash, PartialEq, Eq)]
+pub enum ItemId {
+    /// Tilde ontology root (`~/`); storage [`SLUG_TILDE_ONTOLOGY_ROOT`].
+    Root,
+    /// `https://slug.social/~/…` (non-root; normalized trailing path).
+    Local(String),
+    /// Normalized `http(s)://…` storage form, including non-`~/` paths on `slug.social`.
+    Web(String),
+    /// Raw key material that did not round-trip through [`Self::parse`] (historical edge case).
+    Opaque(String),
+}
+
+impl ItemId {
+    pub fn parse(input: &str) -> Option<Self> {
+        let c = normalize_slug_ontology_storage_url(&canonicalize_item(input));
+        if c.is_empty() {
+            return None;
+        }
+        if c == SLUG_TILDE_ONTOLOGY_ROOT {
+            return Some(Self::Root);
+        }
+        if c.starts_with("https://slug.social/~/") && c != SLUG_TILDE_ONTOLOGY_ROOT {
+            return Some(Self::Local(c));
+        }
+        Some(Self::Web(c))
+    }
+
+    /// Same as the old `ensure_item` fallback: use `s` verbatim as the map key.
+    pub fn opaque(raw: String) -> Self {
+        Self::Opaque(raw)
+    }
+
+    pub fn ontology_root() -> Self {
+        Self::Root
+    }
+
+    /// Collapses legacy slug ontology root spellings so [`std::collections::HashMap`] keys match the graph.
+    pub fn normalized_storage(self) -> Self {
+        Self::parse(self.as_str()).unwrap_or(self)
+    }
+
+    pub fn as_str(&self) -> &str {
+        match self {
+            ItemId::Root => SLUG_TILDE_ONTOLOGY_ROOT,
+            ItemId::Local(s) | ItemId::Web(s) | ItemId::Opaque(s) => s,
+        }
+    }
+
+    pub fn to_storage_string(&self) -> String {
+        self.as_str().to_string()
+    }
+
+    pub fn tilde_tail(&self) -> Option<&str> {
+        match self {
+            ItemId::Root => Some(""),
+            ItemId::Local(s) => s.strip_prefix("https://slug.social/~/"),
+            ItemId::Web(s) | ItemId::Opaque(s) => {
+                if let Some(tail) = s.strip_prefix("https://slug.social/~/") {
+                    return Some(tail);
+                }
+                if s == SLUG_TILDE_ONTOLOGY_ROOT || s == "https://slug.social/~/" {
+                    return Some("");
+                }
+                None
+            }
+        }
+    }
+
+    /// HTTP garden tail after `~/` (empty at ontology root), or `None` if not under tilde ontology.
+    pub fn tilde_http_tail(&self) -> Option<String> {
+        self.tilde_tail().map(str::to_owned)
+    }
+
+    pub fn last_segment(&self) -> &str {
+        let s = self.as_str();
+        s.rsplit('/').find(|x| !x.is_empty()).unwrap_or(s)
+    }
+
+    pub fn parent(&self) -> Option<Self> {
+        match self {
+            ItemId::Root => None,
+            ItemId::Local(s) => {
+                if s == SLUG_TILDE_ONTOLOGY_ROOT || s == "https://slug.social/~/" {
+                    return None;
+                }
+                let last_slash = s.rfind('/')?;
+                let parent_str = &s[..last_slash];
+                if parent_str.is_empty() {
+                    None
+                } else {
+                    Self::parse(parent_str)
+                }
+            }
+            ItemId::Web(s) | ItemId::Opaque(s) => {
+                if s == SLUG_TILDE_ONTOLOGY_ROOT || s == "https://slug.social/~/" {
+                    return None;
+                }
+                if let Some(rest) = s.strip_prefix("https://slug.social/~/") {
+                    if rest.is_empty() {
+                        return None;
+                    }
+                    let last_slash = s.rfind('/')?;
+                    let parent_str = &s[..last_slash];
+                    Self::parse(parent_str)
+                } else if s.starts_with("https://") {
+                    let rest = s.strip_prefix("https://").unwrap();
+                    Self::parent_http_url("https://", rest)
+                } else if s.starts_with("http://") {
+                    let rest = s.strip_prefix("http://").unwrap();
+                    Self::parent_http_url("http://", rest)
+                } else {
+                    None
+                }
+            }
+        }
+    }
+
+    fn parent_http_url(scheme: &'static str, rest: &str) -> Option<Self> {
+        let (host, path) = rest.split_once('/').map_or((rest, ""), |(h, p)| (h, p));
+        let host = host.trim();
+        let path = path.trim_end_matches('/');
+        if path.is_empty() {
+            return None;
+        }
+        let parent_path = path.rsplit_once('/').map(|(p, _)| p).unwrap_or("");
+        if parent_path.is_empty() {
+            Self::parse(&format!("{scheme}{}", host))
+        } else {
+            Self::parse(&format!("{scheme}{}/{}", host, parent_path))
+        }
+    }
+
+    /// `-/` representation for external `https://…` items, `~/…` for slug ontology, else unchanged.
+    pub fn display_path(&self) -> String {
+        if let Some(tail) = self.tilde_tail() {
+            if tail.is_empty() {
+                return "~/".to_string();
+            }
+            return format!("~/{}", tail);
+        }
+        let s = self.as_str();
+        if let Some(tail) = s.strip_prefix("https://") {
+            if tail.starts_with("slug.social") {
+                return s.to_string();
+            }
+            return external_display_dash_prefix(tail);
+        }
+        if let Some(tail) = s.strip_prefix("http://") {
+            if tail.starts_with("slug.social") {
+                return s.to_string();
+            }
+            return external_display_dash_prefix(tail);
+        }
+        s.to_string()
+    }
+
+    pub fn tilde_segments(&self) -> Vec<&str> {
+        match self.tilde_tail() {
+            Some(tail) if !tail.is_empty() => std::iter::once("~")
+                .chain(tail.split('/').filter(|s| !s.is_empty()))
+                .collect(),
+            Some(_) => vec!["~"],
+            None => vec![],
+        }
+    }
+
+    /// Normalized URL string for HTTP fetch boundaries (external identities).
+    pub fn to_wire_url(&self) -> String {
+        self.to_storage_string()
+    }
+}
+
+impl PartialOrd for ItemId {
+    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
+        Some(self.cmp(other))
+    }
+}
+
+impl Ord for ItemId {
+    fn cmp(&self, other: &Self) -> Ordering {
+        self.as_str().cmp(other.as_str())
+    }
+}
+
+impl fmt::Display for ItemId {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        self.as_str().fmt(f)
+    }
+}
+
+impl Serialize for ItemId {
+    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
+    where
+        S: Serializer,
+    {
+        self.as_str().serialize(serializer)
+    }
+}
+
+impl<'de> Deserialize<'de> for ItemId {
+    fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
+    where
+        D: Deserializer<'de>,
+    {
+        let s = String::deserialize(deserializer)?;
+        ItemId::parse(&s).ok_or_else(|| {
+            serde::de::Error::custom(format!("invalid item id: {s:?}"))
+        })
+    }
+}
diff --git a/types/src/item_wire.rs b/types/src/item_wire.rs
new file mode 100644
index 0000000000000000000000000000000000000000..20e3e7df77a74713c146c7e7222a809173723c7d
--- /dev/null
+++ b/types/src/item_wire.rs
@@ -0,0 +1,184 @@
+//! Wire normalization for item identity strings (`canonicalize_item`, path segments).
+//! Split from `paths` so [`crate::item_id::ItemId`] can depend on this without import cycles.
+
+use crate::url_normalize::{host_preserves_dash_path_case, normalize_http_identity_url};
+
+/// Canonical absolute URL for the tilde ontology **root** (`~/` in UI).
+pub const SLUG_TILDE_ONTOLOGY_ROOT: &str = "https://slug.social/~";
+
+/// Collapse legacy or parser variants of the ontology root to [`SLUG_TILDE_ONTOLOGY_ROOT`].
+pub fn normalize_slug_ontology_storage_url(s: &str) -> String {
+    if s == "https://slug.social/~/" {
+        SLUG_TILDE_ONTOLOGY_ROOT.to_string()
+    } else {
+        s.to_string()
+    }
+}
+
+fn finalize_external_identity_url(s: String) -> String {
+    if s.starts_with("https://slug.social/") {
+        return s;
+    }
+    let normalized = normalize_http_identity_url(&s).unwrap_or_else(|| s.clone());
+    strip_redundant_root_slash(&normalized).unwrap_or(normalized)
+}
+
+/// `url::Url` serializes bare hosts with a `/` path; we keep host-only items slash-free for stable
+/// keys matching the pre-normalizer spellings.
+fn strip_redundant_root_slash(s: &str) -> Option<String> {
+    let u = url::Url::parse(s).ok()?;
+    if u.path() == "/" && u.query().is_none() && u.fragment().is_none() {
+        let scheme = u.scheme();
+        let host = u.host_str()?;
+        return Some(match u.port() {
+            Some(p) => format!("{scheme}://{host}:{p}"),
+            None => format!("{scheme}://{host}"),
+        });
+    }
+    None
+}
+
+/// Ontology item reference → canonical absolute URL on the slug host.
+pub fn canonicalize_item(input: &str) -> String {
+    let s = input.trim();
+    if s.is_empty() {
+        return String::new();
+    }
+
+    if let Some(rest) = s.strip_prefix("-/") {
+        let (host, tail) = rest
+            .split_once('/')
+            .map_or((rest, ""), |(h, t)| (h, t));
+        let host = host.trim().to_lowercase();
+        if host.is_empty() {
+            return String::new();
+        }
+        let preserve_case = host_preserves_dash_path_case(&host);
+        return if tail.is_empty() {
+            finalize_external_identity_url(format!("https://{}", host))
+        } else {
+            let path = tail
+                .trim_start_matches('/')
+                .trim_end_matches('/')
+                .split('/')
+                .filter_map(|seg| {
+                    let t = seg.trim();
+                    if t.is_empty() {
+                        None
+                    } else if preserve_case {
+                        Some(t.to_string())
+                    } else {
+                        Some(t.to_lowercase())
+                    }
+                })
+                .collect::<Vec<_>>()
+                .join("/");
+            finalize_external_identity_url(format!("https://{}/{}", host, path))
+        };
+    }
+
+    if let Some(rest) = s.strip_prefix("https://") {
+        let (host, tail) = rest.split_once('/').map_or((rest, ""), |(h, t)| (h, t));
+        let host = host.trim().to_lowercase();
+        return finalize_external_identity_url(if tail.is_empty() {
+            format!("https://{}", host)
+        } else {
+            format!("https://{}/{}", host, tail)
+        });
+    }
+    if let Some(rest) = s.strip_prefix("http://") {
+        let (host, tail) = rest.split_once('/').map_or((rest, ""), |(h, t)| (h, t));
+        let host = host.trim().to_lowercase();
+        return finalize_external_identity_url(if tail.is_empty() {
+            format!("http://{}", host)
+        } else {
+            format!("http://{}/{}", host, tail)
+        });
+    }
+
+    let is_tilde = s.starts_with("~/");
+    let rest = s.strip_prefix("~/").or_else(|| s.strip_prefix("/")).unwrap_or(s);
+
+    let tail = rest
+        .split('/')
+        .filter_map(|seg| {
+            let t = seg.trim();
+            if t.is_empty() {
+                None
+            } else {
+                Some(t.to_lowercase())
+            }
+        })
+        .collect::<Vec<_>>()
+        .join("/");
+
+    if is_tilde {
+        if tail.is_empty() {
+            return SLUG_TILDE_ONTOLOGY_ROOT.to_string();
+        }
+        format!("https://slug.social/~/{}", tail)
+    } else if tail.is_empty() {
+        "https://slug.social".to_string()
+    } else {
+        format!("https://slug.social/{}", tail)
+    }
+}
+
+pub fn item_path_segments(input: &str) -> Vec<String> {
+    let canonical = canonicalize_item(input);
+    if canonical.is_empty() {
+        return vec![];
+    }
+
+    if let Some(rest) = canonical.strip_prefix("https://") {
+        let (host, tail) = rest.split_once('/').map_or((rest, ""), |(h, t)| (h, t));
+        let mut out = vec![format!("https://{}", host)];
+        out.extend(tail.split('/').filter(|s| !s.is_empty()).map(|s| s.to_string()));
+        return out;
+    }
+    if let Some(rest) = canonical.strip_prefix("http://") {
+        let (host, tail) = rest.split_once('/').map_or((rest, ""), |(h, t)| (h, t));
+        let mut out = vec![format!("http://{}", host)];
+        out.extend(tail.split('/').filter(|s| !s.is_empty()).map(|s| s.to_string()));
+        return out;
+    }
+
+    canonical
+        .split('/')
+        .filter(|s| !s.is_empty())
+        .map(|s| s.to_string())
+        .collect()
+}
+
+pub fn item_parent_path(input: &str) -> Option<String> {
+    let segs = item_path_segments(input);
+    if segs.len() <= 1 {
+        return None;
+    }
+    Some(segs[..segs.len() - 1].join("/"))
+}
+
+pub(crate) fn external_display_dash_prefix(host_and_path: &str) -> String {
+    let (host, path) = host_and_path
+        .split_once('/')
+        .map_or((host_and_path, ""), |(h, p)| (h, p));
+    let host = host.trim().to_lowercase();
+    let path = path
+        .trim_end_matches('/')
+        .split('/')
+        .filter_map(|seg| {
+            let t = seg.trim();
+            if t.is_empty() {
+                None
+            } else {
+                Some(t.to_lowercase())
+            }
+        })
+        .collect::<Vec<_>>()
+        .join("/");
+    if path.is_empty() {
+        format!("-/{}", host)
+    } else {
+        format!("-/{}", format!("{}/{}", host, path))
+    }
+}
diff --git a/types/src/lib.rs b/types/src/lib.rs
index ceaa574cd32b7e7eb397c9a3191fcbc037b8c606..05bcfcfbf48f1dc7a4c05e32055bd12eb5cfd623 100644
--- a/types/src/lib.rs
+++ b/types/src/lib.rs
@@ -1,14 +1,20 @@
 use serde::{Deserialize, Serialize};
 
 pub mod url_normalize;
+pub mod item_wire;
+pub mod item_id;
 pub mod paths;
 pub mod timeago;
 
+pub use item_id::ItemId;
+pub use item_wire::{
+    canonicalize_item, item_parent_path, item_path_segments, normalize_slug_ontology_storage_url,
+    SLUG_TILDE_ONTOLOGY_ROOT,
+};
 pub use paths::{
-    canonicalize_item, canonicalize_tag, item_parent_path, item_path_segments, normalize_slug_ontology_storage_url,
-    CanonicalItemUrl, ForumThreadUrl, GardenItemUrl, RelativePath, SLUG_TILDE_ONTOLOGY_ROOT,
+    canonicalize_tag, ForumThreadUrl, GardenItemUrl, RelativePath,
     room_id_from_route_segment, room_route_segment, ROOM_SHORT_ID_LEN,
-    TildeHttpPathTail, TildeOntologyPath, TildePath, tilde_http_path_to_canonical,
+    TildeHttpPathTail, TildeOntologyPath, TildePath, tilde_http_path_to_item_id,
 };
 pub use url_normalize::normalize_http_identity_url;
 
diff --git a/types/src/paths.rs b/types/src/paths.rs
index 5c60e762c2ca74d79c74237097d5bcc02ba74af2..a1d066d83e7fbcefc27707dd682e54dbcec1cc56 100644
--- a/types/src/paths.rs
+++ b/types/src/paths.rs
@@ -3,37 +3,23 @@
 //!
 //! ## String kinds (parse in this module only)
 //!
-//! - **[`canonicalize_item`] / [`CanonicalItemUrl`]** — graph storage key and DSL form; tilde
+//! - **[`canonicalize_item`] / [`crate::ItemId`]** — graph storage key and DSL form; tilde
 //!   ontology root is always [`SLUG_TILDE_ONTOLOGY_ROOT`] (no `…/~/` trailing slash only).
 //! - **[`TildeHttpPathTail`]** — capture from `GET /~/*path` or `…/r/{short}{slug}/~/…` (the `*path` segment).
 //! - **`-/…` wire form** — external items; see [`canonicalize_item`] dash branch.
 //! - **[`GardenItemUrl`], [`ForumThreadUrl`]** — JSON / browser href surfaces.
 //! - **[`ROOM_SHORT_ID_LEN`] / [`room_route_segment`]** — `/r/{short}{slug}` vs wire `short/slug`.
 
-use std::borrow::Borrow;
 use std::fmt;
 use std::ops::Deref;
 
 use serde::{Deserialize, Serialize};
 
-use crate::url_normalize::{host_preserves_dash_path_case, normalize_http_identity_url};
-
-// ---------------------------------------------------------------------------
-// Slug tilde ontology (single storage form for `~/`)
-// ---------------------------------------------------------------------------
-
-/// Canonical absolute URL for the tilde ontology **root** (`~/` in UI). Used as the
-/// `item_children` parent key for top-level items and must match [`CanonicalItemUrl::ontology_root`].
-pub const SLUG_TILDE_ONTOLOGY_ROOT: &str = "https://slug.social/~";
-
-/// Collapse legacy or parser variants of the ontology root to [`SLUG_TILDE_ONTOLOGY_ROOT`].
-pub fn normalize_slug_ontology_storage_url(s: &str) -> String {
-    if s == "https://slug.social/~/" {
-        SLUG_TILDE_ONTOLOGY_ROOT.to_string()
-    } else {
-        s.to_string()
-    }
-}
+use crate::item_id::ItemId;
+pub use crate::item_wire::{
+    canonicalize_item, item_parent_path, item_path_segments, normalize_slug_ontology_storage_url,
+    SLUG_TILDE_ONTOLOGY_ROOT,
+};
 
 // ---------------------------------------------------------------------------
 // Private room HTTP path (`/r/{short}{slug}`; wire id remains `short/slug`)
@@ -84,310 +70,10 @@ pub fn canonicalize_tag(input: &str) -> String {
     input.trim().trim_start_matches('#').to_lowercase()
 }
 
-fn finalize_external_identity_url(s: String) -> String {
-    if s.starts_with("https://slug.social/") {
-        return s;
-    }
-    let normalized = normalize_http_identity_url(&s).unwrap_or_else(|| s.clone());
-    strip_redundant_root_slash(&normalized).unwrap_or(normalized)
-}
-
-/// `url::Url` serializes bare hosts with a `/` path; we keep host-only items slash-free for stable
-/// keys matching the pre-normalizer spellings.
-fn strip_redundant_root_slash(s: &str) -> Option<String> {
-    let u = url::Url::parse(s).ok()?;
-    if u.path() == "/" && u.query().is_none() && u.fragment().is_none() {
-        let scheme = u.scheme();
-        let host = u.host_str()?;
-        return Some(match u.port() {
-            Some(p) => format!("{scheme}://{host}:{p}"),
-            None => format!("{scheme}://{host}"),
-        });
-    }
-    None
-}
-
-/// Ontology item reference → canonical absolute URL on the slug host.
-pub fn canonicalize_item(input: &str) -> String {
-    let s = input.trim();
-    if s.is_empty() {
-        return String::new();
-    }
-
-    // External scope: `-/host/path` is the universal alias for `https://host/path`.
-    if let Some(rest) = s.strip_prefix("-/") {
-        let (host, tail) = rest
-            .split_once('/')
-            .map_or((rest, ""), |(h, t)| (h, t));
-        let host = host.trim().to_lowercase();
-        if host.is_empty() {
-            return String::new();
-        }
-        let preserve_case = host_preserves_dash_path_case(&host);
-        return if tail.is_empty() {
-            finalize_external_identity_url(format!("https://{}", host))
-        } else {
-            let path = tail
-                .trim_start_matches('/')
-                .trim_end_matches('/')
-                .split('/')
-                .filter_map(|seg| {
-                    let t = seg.trim();
-                    if t.is_empty() {
-                        None
-                    } else if preserve_case {
-                        Some(t.to_string())
-                    } else {
-                        Some(t.to_lowercase())
-                    }
-                })
-                .collect::<Vec<_>>()
-                .join("/");
-            finalize_external_identity_url(format!("https://{}/{}", host, path))
-        };
-    }
-
-    if let Some(rest) = s.strip_prefix("https://") {
-        let (host, tail) = rest.split_once('/').map_or((rest, ""), |(h, t)| (h, t));
-        let host = host.trim().to_lowercase();
-        return finalize_external_identity_url(if tail.is_empty() {
-            format!("https://{}", host)
-        } else {
-            format!("https://{}/{}", host, tail)
-        });
-    }
-    if let Some(rest) = s.strip_prefix("http://") {
-        let (host, tail) = rest.split_once('/').map_or((rest, ""), |(h, t)| (h, t));
-        let host = host.trim().to_lowercase();
-        return finalize_external_identity_url(if tail.is_empty() {
-            format!("http://{}", host)
-        } else {
-            format!("http://{}/{}", host, tail)
-        });
-    }
-
-    let is_tilde = s.starts_with("~/");
-    let rest = s.strip_prefix("~/").or_else(|| s.strip_prefix("/")).unwrap_or(s);
-
-    let tail = rest
-        .split('/')
-        .filter_map(|seg| {
-            let t = seg.trim();
-            if t.is_empty() {
-                None
-            } else {
-                Some(t.to_lowercase())
-            }
-        })
-        .collect::<Vec<_>>()
-        .join("/");
-
-    if is_tilde {
-        if tail.is_empty() {
-            return SLUG_TILDE_ONTOLOGY_ROOT.to_string();
-        }
-        format!("https://slug.social/~/{}", tail)
-    } else if tail.is_empty() {
-        "https://slug.social".to_string()
-    } else {
-        format!("https://slug.social/{}", tail)
-    }
-}
-
-pub fn item_path_segments(input: &str) -> Vec<String> {
-    let canonical = canonicalize_item(input);
-    if canonical.is_empty() {
-        return vec![];
-    }
-
-    if let Some(rest) = canonical.strip_prefix("https://") {
-        let (host, tail) = rest.split_once('/').map_or((rest, ""), |(h, t)| (h, t));
-        let mut out = vec![format!("https://{}", host)];
-        out.extend(tail.split('/').filter(|s| !s.is_empty()).map(|s| s.to_string()));
-        return out;
-    }
-    if let Some(rest) = canonical.strip_prefix("http://") {
-        let (host, tail) = rest.split_once('/').map_or((rest, ""), |(h, t)| (h, t));
-        let mut out = vec![format!("http://{}", host)];
-        out.extend(tail.split('/').filter(|s| !s.is_empty()).map(|s| s.to_string()));
-        return out;
-    }
-
-    canonical
-        .split('/')
-        .filter(|s| !s.is_empty())
-        .map(|s| s.to_string())
-        .collect()
-}
-
-pub fn item_parent_path(input: &str) -> Option<String> {
-    let segs = item_path_segments(input);
-    if segs.len() <= 1 {
-        return None;
-    }
-    Some(segs[..segs.len() - 1].join("/"))
-}
-
-fn external_display_dash_prefix(host_and_path: &str) -> String {
-    let (host, path) = host_and_path
-        .split_once('/')
-        .map_or((host_and_path, ""), |(h, p)| (h, p));
-    let host = host.trim().to_lowercase();
-    let path = path
-        .trim_end_matches('/')
-        .split('/')
-        .filter_map(|seg| {
-            let t = seg.trim();
-            if t.is_empty() {
-                None
-            } else {
-                Some(t.to_lowercase())
-            }
-        })
-        .collect::<Vec<_>>()
-        .join("/");
-    if path.is_empty() {
-        format!("-/{}", host)
-    } else {
-        format!("-/{}", format!("{}/{}", host, path))
-    }
-}
-
 // ---------------------------------------------------------------------------
 // Storage + input path newtypes
 // ---------------------------------------------------------------------------
 
-/// Canonical item identifier as produced by [`canonicalize_item`].
-///
-/// Shared across all scopes; room is not embedded. Usually
-/// `https://slug.social/~/…` or an external `http(s)://…` URL item.
-#[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord, Hash, Serialize, Deserialize)]
-pub struct CanonicalItemUrl(pub String);
-
-impl CanonicalItemUrl {
-    pub fn parse(input: &str) -> Option<Self> {
-        let c = canonicalize_item(input);
-        if c.is_empty() {
-            None
-        } else {
-            Some(Self(normalize_slug_ontology_storage_url(&c)))
-        }
-    }
-
-    pub fn as_str(&self) -> &str {
-        &self.0
-    }
-
-    /// Collapses legacy slug ontology root spellings so [`HashMap`] keys match the reducer graph.
-    pub fn normalized_storage(self) -> Self {
-        Self(normalize_slug_ontology_storage_url(self.as_str()))
-    }
-
-    pub fn tilde_tail(&self) -> Option<&str> {
-        if let Some(tail) = self.0.strip_prefix("https://slug.social/~/") {
-            return Some(tail);
-        }
-        if self.0 == SLUG_TILDE_ONTOLOGY_ROOT || self.0 == "https://slug.social/~/" {
-            return Some("");
-        }
-        None
-    }
-
-    pub fn last_segment(&self) -> &str {
-        self.0
-            .rsplit('/')
-            .find(|s| !s.is_empty())
-            .unwrap_or(self.0.as_str())
-    }
-
-    pub fn ontology_root() -> Self {
-        Self(SLUG_TILDE_ONTOLOGY_ROOT.to_string())
-    }
-
-    pub fn parent(&self) -> Option<Self> {
-        if self.tilde_tail().is_some() {
-            if self.tilde_tail().map(|t| t.is_empty()).unwrap_or(true) {
-                return None;
-            }
-            let last_slash = self.0.rfind('/')?;
-            let parent_str = &self.0[..last_slash];
-            if parent_str.is_empty() {
-                None
-            } else {
-                Some(Self(parent_str.to_string()))
-            }
-        } else if let Some(rest) = self.0.strip_prefix("https://") {
-            Self::parent_http_url("https://", rest)
-        } else if let Some(rest) = self.0.strip_prefix("http://") {
-            Self::parent_http_url("http://", rest)
-        } else {
-            None
-        }
-    }
-
-    fn parent_http_url(scheme: &'static str, rest: &str) -> Option<Self> {
-        let (host, path) = rest.split_once('/').map_or((rest, ""), |(h, p)| (h, p));
-        let host = host.trim();
-        let path = path.trim_end_matches('/');
-        if path.is_empty() {
-            return None;
-        }
-        let parent_path = path.rsplit_once('/').map(|(p, _)| p).unwrap_or("");
-        if parent_path.is_empty() {
-            Some(Self(format!("{scheme}{}", host)))
-        } else {
-            Some(Self(format!("{scheme}{}/{}", host, parent_path)))
-        }
-    }
-
-    /// `-/` representation for external `https://…` items, `~/…` for slug ontology, else unchanged.
-    pub fn display_path(&self) -> String {
-        if let Some(tail) = self.tilde_tail() {
-            if tail.is_empty() {
-                return "~/".to_string();
-            }
-            return format!("~/{}", tail);
-        }
-        if let Some(tail) = self.0.strip_prefix("https://") {
-            if tail.starts_with("slug.social") {
-                self.0.clone()
-            } else {
-                external_display_dash_prefix(tail)
-            }
-        } else if let Some(tail) = self.0.strip_prefix("http://") {
-            if tail.starts_with("slug.social") {
-                self.0.clone()
-            } else {
-                external_display_dash_prefix(tail)
-            }
-        } else {
-            self.0.clone()
-        }
-    }
-
-    pub fn tilde_segments(&self) -> Vec<&str> {
-        match self.tilde_tail() {
-            Some(tail) if !tail.is_empty() => {
-                std::iter::once("~")
-                    .chain(tail.split('/').filter(|s| !s.is_empty()))
-                    .collect()
-            }
-            Some(_) => vec!["~"],
-            None => vec![],
-        }
-    }
-
-    /// `~/…` list label for ontology items (paths index, CLI).
-    pub fn tilde_list_label(&self) -> TildeOntologyPath {
-        TildeOntologyPath::from_stored(self)
-    }
-
-    /// Absolute href for JSON/RPC and browsers for this stored id in `room`.
-    pub fn json_href(&self, room_wire: &str) -> GardenItemUrl {
-        GardenItemUrl::from_stored(self, room_wire)
-    }
-}
-
 /// HTTP route capture: path segment after `~/` in `GET /~/*path` or `…/r/{short}{slug}/~/…` (empty = ontology root).
 #[derive(Debug, Clone, PartialEq, Eq, Hash)]
 pub struct TildeHttpPathTail(pub String);
@@ -401,13 +87,13 @@ impl TildeHttpPathTail {
         &self.0
     }
 
-    pub fn to_canonical(&self) -> CanonicalItemUrl {
-        tilde_http_path_to_canonical(self.as_str())
+    pub fn to_item_id(&self) -> ItemId {
+        tilde_http_path_to_item_id(self.as_str())
     }
 }
 
-/// Map the router's tilde tail (e.g. `topic/a`, or empty for root) to a [`CanonicalItemUrl`].
-pub fn tilde_http_path_to_canonical(path_segment: &str) -> CanonicalItemUrl {
+/// Map the router's tilde tail (e.g. `topic/a`, or empty for root) to an [`ItemId`].
+pub fn tilde_http_path_to_item_id(path_segment: &str) -> ItemId {
     let p = path_segment.trim_start_matches('/');
     let raw = if p.starts_with("http://") || p.starts_with("https://") {
         p.to_string()
@@ -416,37 +102,7 @@ pub fn tilde_http_path_to_canonical(path_segment: &str) -> CanonicalItemUrl {
     } else {
         format!("~/{}", p)
     };
-    CanonicalItemUrl::parse(&raw).unwrap_or_else(|| CanonicalItemUrl::ontology_root())
-}
-
-impl fmt::Display for CanonicalItemUrl {
-    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
-        self.0.fmt(f)
-    }
-}
-
-impl Borrow<str> for CanonicalItemUrl {
-    fn borrow(&self) -> &str {
-        &self.0
-    }
-}
-
-impl PartialEq<str> for CanonicalItemUrl {
-    fn eq(&self, other: &str) -> bool {
-        self.0 == other
-    }
-}
-
-impl PartialEq<&str> for CanonicalItemUrl {
-    fn eq(&self, other: &&str) -> bool {
-        self.0 == *other
-    }
-}
-
-impl PartialEq<String> for CanonicalItemUrl {
-    fn eq(&self, other: &String) -> bool {
-        &self.0 == other
-    }
+    ItemId::parse(&raw).unwrap_or_else(ItemId::ontology_root)
 }
 
 #[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord, Hash, Serialize, Deserialize)]
@@ -468,8 +124,8 @@ impl TildePath {
         &self.0
     }
 
-    pub fn canonicalize(&self) -> Option<CanonicalItemUrl> {
-        CanonicalItemUrl::parse(&self.0)
+    pub fn canonicalize(&self) -> Option<ItemId> {
+        ItemId::parse(&self.0)
     }
 }
 
@@ -496,7 +152,7 @@ impl RelativePath {
         &self.0
     }
 
-    pub fn join_under_ontology_root(&self, root: &CanonicalItemUrl) -> Option<CanonicalItemUrl> {
+    pub fn join_under_ontology_root(&self, root: &ItemId) -> Option<ItemId> {
         let base = root.tilde_tail()?;
         let joined = if base.is_empty() {
             if self.0.is_empty() {
@@ -509,7 +165,7 @@ impl RelativePath {
         } else {
             format!("~/{}/{}", base.trim_end_matches('/'), self.0)
         };
-        CanonicalItemUrl::parse(&joined)
+        ItemId::parse(&joined)
     }
 }
 
@@ -545,14 +201,17 @@ impl GardenItemUrl {
         self.0
     }
 
-    /// Stored canonical id + RPC `room` field (`"public"` or `"short/slug"`).
-    pub fn from_stored(stored: &CanonicalItemUrl, room_wire: &str) -> Self {
-        Self(garden_href_string(stored.as_str(), room_wire))
+    /// Stored item id + RPC `room` field (`"public"` or `"short/slug"`).
+    pub fn from_stored(stored: &ItemId, room_wire: &str) -> Self {
+        Self(garden_href_string(stored, room_wire))
     }
 
     /// Like [`Self::from_stored`] but accepts a string that may already be canonical.
     pub fn from_storage_str(stored: &str, room_wire: &str) -> Self {
-        Self(garden_href_string(stored, room_wire))
+        let Some(id) = ItemId::parse(stored) else {
+            return Self(api_path_or_url(stored));
+        };
+        Self(garden_href_string(&id, room_wire))
     }
 }
 
@@ -570,18 +229,15 @@ impl Deref for GardenItemUrl {
     }
 }
 
-fn garden_href_string(item: &str, room_wire: &str) -> String {
+fn garden_href_string(c: &ItemId, room_wire: &str) -> String {
     let room = room_wire.trim();
     if room.is_empty() || room == "public" {
-        return api_path_or_url(item);
+        return api_path_or_url(c.as_str());
     }
     let Some(room_seg) = room_route_segment(room) else {
-        return api_path_or_url(item);
-    };
-    let Some(c) = CanonicalItemUrl::parse(item) else {
-        return api_path_or_url(item);
+        return api_path_or_url(c.as_str());
     };
-    let root = CanonicalItemUrl::ontology_root();
+    let root = ItemId::ontology_root();
     let item_norm = c.as_str().trim_end_matches('/');
     let root_norm = root.as_str().trim_end_matches('/');
     if let Some(tail) = c.tilde_tail() {
@@ -600,7 +256,7 @@ fn garden_href_string(item: &str, room_wire: &str) -> String {
         let tail = tail.strip_prefix("-/").unwrap_or(tail.as_str());
         return format!("https://slug.social/r/{room_seg}/-/{tail}");
     }
-    api_path_or_url(item)
+    api_path_or_url(c.as_str())
 }
 
 /// Forum thread URL for JSON (`/t/…` or `/r/…/t/…` on slug.social).
@@ -650,7 +306,7 @@ impl Deref for ForumThreadUrl {
 pub struct TildeOntologyPath(pub String);
 
 impl TildeOntologyPath {
-    pub fn from_stored(c: &CanonicalItemUrl) -> Self {
+    pub fn from_stored(c: &ItemId) -> Self {
         Self(c.display_path())
     }
 
@@ -679,22 +335,22 @@ mod tests {
 
     #[test]
     fn canonical_parent_deep() {
-        let c = CanonicalItemUrl::parse("~/a/b/c").unwrap();
+        let c = ItemId::parse("~/a/b/c").unwrap();
         assert_eq!(c.parent().unwrap().as_str(), "https://slug.social/~/a/b");
     }
 
     #[test]
     fn canonical_parent_one_level() {
-        let c = CanonicalItemUrl::parse("~/a").unwrap();
+        let c = ItemId::parse("~/a").unwrap();
         assert_eq!(c.parent().unwrap().as_str(), "https://slug.social/~");
     }
 
     #[test]
     fn canonical_parent_root_is_none() {
-        let root = CanonicalItemUrl::parse("~/").unwrap();
+        let root = ItemId::parse("~/").unwrap();
         assert!(root.parent().is_none());
         assert_eq!(root.as_str(), SLUG_TILDE_ONTOLOGY_ROOT);
-        assert_eq!(root, CanonicalItemUrl::ontology_root());
+        assert_eq!(root, ItemId::ontology_root());
     }
 
     #[test]
@@ -705,49 +361,49 @@ mod tests {
             SLUG_TILDE_ONTOLOGY_ROOT.to_string()
         );
         assert_eq!(
-            CanonicalItemUrl::parse("https://slug.social/~/")
+            ItemId::parse("https://slug.social/~/")
                 .unwrap()
                 .as_str(),
             SLUG_TILDE_ONTOLOGY_ROOT
         );
-        let legacy = CanonicalItemUrl("https://slug.social/~/".to_string());
+        let legacy = ItemId::opaque("https://slug.social/~/".to_string());
         assert_eq!(legacy.normalized_storage().as_str(), SLUG_TILDE_ONTOLOGY_ROOT);
     }
 
     #[test]
     fn tilde_http_path_tail_maps_router_segment() {
         assert_eq!(
-            TildeHttpPathTail::new("").to_canonical(),
-            CanonicalItemUrl::ontology_root()
+            TildeHttpPathTail::new("").to_item_id(),
+            ItemId::ontology_root()
         );
         assert_eq!(
-            tilde_http_path_to_canonical("topic/x").as_str(),
+            tilde_http_path_to_item_id("topic/x").as_str(),
             "https://slug.social/~/topic/x"
         );
     }
 
     #[test]
     fn display_path_slug_ontology_root() {
-        let r = CanonicalItemUrl::ontology_root();
+        let r = ItemId::ontology_root();
         assert_eq!(r.display_path(), "~/");
         assert_eq!(r.tilde_tail(), Some(""));
     }
 
     #[test]
     fn tilde_segments_deep() {
-        let c = CanonicalItemUrl::parse("~/a/b").unwrap();
+        let c = ItemId::parse("~/a/b").unwrap();
         assert_eq!(c.tilde_segments(), vec!["~", "a", "b"]);
     }
 
     #[test]
     fn tilde_segments_root() {
-        let c = CanonicalItemUrl::parse("~/").unwrap();
+        let c = ItemId::parse("~/").unwrap();
         assert_eq!(c.tilde_segments(), vec!["~"]);
     }
 
     #[test]
     fn tilde_segments_non_ontology_is_empty() {
-        let c = CanonicalItemUrl::parse("https://example.com/foo").unwrap();
+        let c = ItemId::parse("https://example.com/foo").unwrap();
         assert_eq!(c.tilde_segments(), Vec::<&str>::new());
     }
 
@@ -831,28 +487,28 @@ mod tests {
     }
 
     #[test]
-    fn canonical_item_url_parent_external_strips_last_segment() {
-        let c = CanonicalItemUrl::parse("https://spotify.com/track/1").unwrap();
+    fn item_id_parent_external_strips_last_segment() {
+        let c = ItemId::parse("https://spotify.com/track/1").unwrap();
         assert_eq!(
             c.parent().unwrap().as_str(),
             "https://spotify.com/track"
         );
         assert_eq!(
-            CanonicalItemUrl::parse("https://github.com/iss/1")
+            ItemId::parse("https://github.com/iss/1")
                 .unwrap()
                 .parent()
                 .unwrap()
                 .as_str(),
             "https://github.com/iss"
         );
-        assert!(CanonicalItemUrl::parse("https://github.com").unwrap().parent().is_none());
+        assert!(ItemId::parse("https://github.com").unwrap().parent().is_none());
     }
 
     #[test]
     fn display_path_roundtrips_dash_and_tilde() {
-        let ext = CanonicalItemUrl::parse("https://GitHub.com/org/Issue").unwrap();
+        let ext = ItemId::parse("https://GitHub.com/org/Issue").unwrap();
         assert_eq!(ext.display_path(), "-/github.com/org/issue");
-        let tilde = CanonicalItemUrl::parse("~/Rust/Doc").unwrap();
+        let tilde = ItemId::parse("~/Rust/Doc").unwrap();
         assert_eq!(tilde.display_path(), "~/rust/doc");
     }
 


Side B — contributor: tommy-mor
Side B — commit message:
[2bc302c3] refactor

Side B — unified diff (full patch):
diff --git a/server/src/api/ui_html.rs b/server/src/api/ui_html.rs
index 06001212820101e0cc953d3687dea64f85e60787..d2f9769108def7ca2c5857aec8b4319426188a66 100644
--- a/server/src/api/ui_html.rs
+++ b/server/src/api/ui_html.rs
@@ -47,7 +47,7 @@ pub async fn post_ui_html(
             ratio_left,
             ratio_right,
             scope,
-            next,
+            vote_compare,
         } => {
             let parent = parent_from_scope(&scope);
             if let Err(e) = state
@@ -57,14 +57,12 @@ pub async fn post_ui_html(
                 return ui_js_warn(&e).into_response();
             }
             let tree = state.tree.read().await;
-            if !next.trim().is_empty() {
+            if vote_compare {
+                let left = parse_item_param(&a);
+                let right = parse_item_param(&b);
+                let morph = crate::html::vote::vote_recorded_morph(&tree, &parent, &left, &right);
                 drop(tree);
-                return JsBuilder::new()
-                    .raw(&format!(
-                        "window.location.href={};",
-                        js_string_literal(next.trim())
-                    ))
-                    .into_response();
+                return morph.into_response();
             }
             let empty = crate::reducer::NodeState::default();
             let node = tree.get(&parent).unwrap_or(&empty);
@@ -137,7 +135,7 @@ mod tests {
                 ratio_left: 3,
                 ratio_right: 1,
                 scope: String::new(),
-                next: String::new(),
+                vote_compare: false,
             }
         );
     }
diff --git a/server/src/html/mod.rs b/server/src/html/mod.rs
index 61cdbc094819ddedb755572c59456ec0d6617619..cd578a5fea46a9a1d49e238d08a80c2caf18708d 100644
--- a/server/src/html/mod.rs
+++ b/server/src/html/mod.rs
@@ -65,6 +65,15 @@ impl JsBuilder {
         self
     }
 
+    pub(crate) fn morph_inner_selector(mut self, selector: &str, markup: Markup) -> Self {
+        let html = js_string_literal(&markup.into_string());
+        self.snippets.push(format!(
+            "var __el = document.querySelector({sel}); if (__el) {{ Idiomorph.morph(__el, {html}, {{ morphStyle: 'innerHTML' }}); }}",
+            sel = js_string_literal(selector),
+        ));
+        self
+    }
+
     pub(crate) fn raw(mut self, js: &str) -> Self {
         if !js.is_empty() {
             self.snippets.push(js.to_string());
diff --git a/server/src/html/vote.rs b/server/src/html/vote.rs
new file mode 100644
index 0000000000000000000000000000000000000000..89ed621ad861214e147add2062224fc4137ff8ad
--- /dev/null
+++ b/server/src/html/vote.rs
@@ -0,0 +1,306 @@
+//! Pairwise vote UI — `/vote?parent=` with optional `left` / `right`.
+
+use axum::{
+    extract::{Query, State},
+    response::{Html, IntoResponse},
+};
+use maud::{html, Markup};
+use serde::Deserialize;
+
+use crate::{
+    form_template::template_json_compact,
+    html::JsBuilder,
+    pair::{children_of, resolve_pair, suggest_next_pair_in_pool},
+    path_types::ItemId,
+    reducer::{GlobalTree, GroupState, NodeState, VoteData},
+    state::{parse_item_param, AppState},
+    ui_action::UI_RPC_FIELD,
+};
+
+use super::{breadcrumb_path, item_href, layout};
+
+#[derive(Debug, Deserialize)]
+pub struct VoteQuery {
+    pub parent: String,
+    #[serde(default)]
+    pub left: Option<String>,
+    #[serde(default)]
+    pub right: Option<String>,
+}
+
+pub fn vote_href(parent: &ItemId) -> String {
+    format!(
+        "/vote?parent={}",
+        urlencoding::encode(parent.as_str())
+    )
+}
+
+fn vote_compare_href(parent: &ItemId, left: &ItemId, right: &ItemId) -> String {
+    format!(
+        "/vote?parent={}&left={}&right={}",
+        urlencoding::encode(parent.as_str()),
+        urlencoding::encode(left.as_str()),
+        urlencoding::encode(right.as_str()),
+    )
+}
+
+fn display_label(id: &ItemId) -> String {
+    id.segments()
+        .last()
+        .map_or("item".into(), |v| v.to_string())
+}
+
+fn child_title(tree: &GlobalTree, id: &ItemId) -> String {
+    tree.get(id)
+        .and_then(|n| n.data.as_ref())
+        .map(|d| d.title.clone())
+        .unwrap_or_else(|| display_label(id))
+}
+
+fn ratio_pct(ratio_left: i32, ratio_right: i32) -> f64 {
+    let l = ratio_left.max(0) as f64;
+    let r = ratio_right.max(0) as f64;
+    let sum = l + r;
+    if sum <= 0.0 {
+        50.0
+    } else {
+        (l / sum) * 100.0
+    }
+}
+
+fn ratios_for_page(v: &VoteData, page_left: &ItemId, page_right: &ItemId) -> (i32, i32) {
+    match (v.a.as_str(), v.b.as_str()) {
+        (a, b) if a == page_left.as_str() && b == page_right.as_str() => {
+            (v.ratio_left, v.ratio_right)
+        }
+        (a, b) if a == page_right.as_str() && b == page_left.as_str() => {
+            (v.ratio_right, v.ratio_left)
+        }
+        _ => (v.ratio_left, v.ratio_right),
+    }
+}
+
+fn edge_votes(group: &GroupState, left: &ItemId, right: &ItemId) -> Vec<VoteData> {
+    group
+        .recent_votes
+        .iter()
+        .filter(|v| {
+            (v.a.as_str() == left.as_str() && v.b.as_str() == right.as_str())
+                || (v.a.as_str() == right.as_str() && v.b.as_str() == left.as_str())
+        })
+        .cloned()
+        .collect()
+}
+
+fn vote_edge_history(tree: &GlobalTree, group: &GroupState, left: &ItemId, right: &ItemId) -> Markup {
+    let mut votes = edge_votes(group, left, right);
+    votes.sort_by(|a, b| b.ts.cmp(&a.ts));
+    let legend_left = child_title(tree, left);
+    let legend_right = child_title(tree, right);
+    html! {
+        @if votes.is_empty() {
+            p class="muted vote-edge-empty" { "no votes on this pair yet" }
+        } @else {
+            h3 class="vote-edge-history-title" {
+                "votes on this pair"
+                span class="vote-edge-history-axis muted" { " · " (legend_left) " : " (legend_right) }
+            }
+            ul class="vote-edge-history" {
+                @for v in &votes {
+                    @let (r_left, r_right) = ratios_for_page(v, left, right);
+                    @let pct = ratio_pct(r_left, r_right);
+                    li class="vote-edge-history-row" {
+                        div class="vote-edge-meta" {
+                            span class="vote-edge-ratio" { (format!("{}:{}", r_left, r_right)) }
+                        }
+                        div class="ratio-bar vote-edge-bar" aria-hidden="true" {
+                            div class="ratio-left" style={(format!("width: {:.3}%;", pct))} {}
+                            div class="ratio-right" style={(format!("width: {:.3}%;", 100.0 - pct))} {}
+                        }
+                    }
+                }
+            }
+        }
+    }
+}
+
+fn vote_back_nav(parent: &ItemId) -> Markup {
+    html! {
+        div class="vote-compare-nav" {
+            a class="vote-compare-back muted" href=(item_href(parent)) { "← back to " (display_label(parent)) }
+        }
+    }
+}
+
+fn vote_compare_actions(parent: &ItemId, next: Option<&(ItemId, ItemId)>) -> Markup {
+    let next_href = next.map(|(l, r)| vote_compare_href(parent, l, r));
+    html! {
+        div id="vote-compare-actions" class="vote-compare-actions" {
+            button type="submit" class="btn-primary" data-testid="vote-post" { "post vote" }
+            @if let Some(href) = &next_href {
+                a class="btn-secondary vote-compare-next" data-testid="vote-next-pair" href=(href) { "next pair" }
+            } @else {
+                span class="btn-secondary vote-compare-next is-disabled" { "no next pair" }
+            }
+        }
+    }
+}
+
+/// After recording a vote on the compare page: refresh edge history and next-pair link.
+pub(crate) fn vote_recorded_morph(
+    tree: &GlobalTree,
+    parent: &ItemId,
+    left: &ItemId,
+    right: &ItemId,
+) -> JsBuilder {
+    let pool = children_of(tree, parent);
+    let empty = NodeState::default();
+    let group = tree
+        .get(parent)
+        .unwrap_or(&empty)
+        .local_ranking
+        .clone();
+    let edge_history = vote_edge_history(tree, &group, left, right);
+    let next_pair = suggest_next(&group, left, right, &pool);
+    let actions = vote_compare_actions(parent, next_pair.as_ref());
+    JsBuilder::new()
+        .morph_inner_selector("#vote-edge-history-region", edge_history)
+        .morph_selector("#vote-compare-actions", actions)
+}
+
+fn vote_compare_item_card(tree: &GlobalTree, item: &ItemId, side_class: &str) -> Markup {
+    let href = item_href(item);
+    let title = child_title(tree, item);
+    html! {
+        div class=(format!("vote-compare-side {side_class}")) {
+            a class=(format!("vote-compare-item {side_class}")) href=(href) {
+                @if let Some(row) = crate::render::reddit::child_row_markup(tree, item, &href) {
+                    (row)
+                } @else {
+                    strong { (title) }
+                }
+            }
+            @if let Some(node) = tree.get(item) {
+                @if crate::render::reddit::is_reddit_post(item) {
+                    @if let Some(data) = &node.data {
+                        @if let Some(src) = data.image_url.as_ref().or(data.thumb_url.as_ref()) {
+                            figure class="vote-compare-figure" {
+                                img class="vote-compare-image" src=(src) alt="" loading="lazy";
+                            }
+                        }
+                        @if let Some(author) = &data.author {
+                            p class="muted small" { "by " (author) }
+                        }
+                    }
+                } @else if let Some(data) = &node.data {
+                    @if let Some(body) = &data.body_html {
+                        div class="vote-compare-item-body" {
+                            (maud::PreEscaped(body))
+                        }
+                    }
+                }
+            }
+        }
+    }
+}
+
+
+fn suggest_next(group: &GroupState, left: &ItemId, right: &ItemId, pool: &[ItemId]) -> Option<(ItemId, ItemId)> {
+    suggest_next_pair_in_pool(group, pool, Some((left, right)))
+}
+
+pub async fn vote_page(
+    State(state): State<AppState>,
+    Query(q): Query<VoteQuery>,
+) -> impl IntoResponse {
+    let parent = parse_item_param(&q.parent);
+    let left_param = q.left.as_deref().map(parse_item_param);
+    let right_param = q.right.as_deref().map(parse_item_param);
+
+    let tree = state.tree.read().await;
+    let empty = NodeState::default();
+    let parent_node = tree.get(&parent).unwrap_or(&empty);
+
+    let (left, right) = match resolve_pair(
+        &tree,
+        &parent,
+        left_param.as_ref(),
+        right_param.as_ref(),
+    ) {
+        Ok(p) => p,
+        Err(e) => {
+            let (msg, status) = e.status_message();
+            return (status, msg).into_response();
+        }
+    };
+
+    let pool = children_of(&tree, &parent);
+    let group = &parent_node.local_ranking;
+    let next_pair = suggest_next(group, &left, &right, &pool);
+    let edge_history = vote_edge_history(&tree, group, &left, &right);
+
+    let rpc_json = template_json_compact(&serde_json::json!({
+        "action": "record_vote",
+        "a": left.as_str(),
+        "b": right.as_str(),
+        "ratio_left": {"$form:i32": "ratio_left"},
+        "ratio_right": {"$form:i32": "ratio_right"},
+        "scope": parent.as_str(),
+        "vote_compare": true,
+    }))
+    .expect("vote rpc json");
+
+    let title = format!(
+        "vote — {} vs {}",
+        child_title(&tree, &left),
+        child_title(&tree, &right)
+    );
+
+    let body = html! {
+        section class="vote-compare-shell" {
+            h1 { "compare" }
+            (breadcrumb_path(&parent))
+            p class="muted vote-compare-scope" {
+                "ranking children of "
+                a href=(item_href(&parent)) { (child_title(&tree, &parent)) }
+            }
+            div class="vote-compare-pair" {
+                (vote_compare_item_card(&tree, &left, "vote-compare-left"))
+                span class="vote-compare-vs" { "vs" }
+                (vote_compare_item_card(&tree, &right, "vote-compare-right"))
+            }
+            (vote_back_nav(&parent))
+            div id="vote-edge-history-region" {
+                (edge_history)
+            }
+            form id="vote-compare-form" method="POST" action="/ui" {
+                input type="hidden" name=(UI_RPC_FIELD) value=(rpc_json);
+                input type="hidden" name="ratio_left" id="vote-ratio-left" value="50";
+                input type="hidden" name="ratio_right" id="vote-ratio-right" value="50";
+                label class="vote-compare-slider-label" {
+                    span id="vote-slider-left-label" { (child_title(&tree, &left)) }
+                    input type="range" id="vote-preference-slider" min="0" max="100" value="50"
+                        aria-valuemin="0" aria-valuemax="100";
+                    span id="vote-slider-right-label" { (child_title(&tree, &right)) }
+                }
+                (vote_compare_actions(&parent, next_pair.as_ref()))
+            }
+        }
+    };
+
+    drop(tree);
+
+    let path = format!("/vote?parent={}", urlencoding::encode(parent.as_str()));
+    state.views.increment(path.clone());
+    let views = state.views.get_views(&path);
+
+    Html(
+        layout(
+            &title,
+            body,
+            views,
+        )
+        .into_string(),
+    )
+    .into_response()
+}
diff --git a/server/src/pair.rs b/server/src/pair.rs
new file mode 100644
index 0000000000000000000000000000000000000000..606ffa51038a57ffacf335aa48bdb1185f8483fb
--- /dev/null
+++ b/server/src/pair.rs
@@ -0,0 +1,340 @@
+//! Pick two children of a parent scope for pairwise voting.
+//!
+//! Pair selection prefers **bridge** votes — comparisons between items in
+//! different connected components of the voted-pairs graph — so the pool
+//! merges into one ranking group before refining within it.
+
+use rand::seq::SliceRandom;
+use std::collections::{HashMap, HashSet};
+
+use crate::{
+    path_types::ItemId,
+    ranking::connected_components_from_voted_pairs,
+    reducer::{GlobalTree, GroupState},
+};
+
+fn pairs_match(a: &ItemId, b: &ItemId, x: &ItemId, y: &ItemId) -> bool {
+    (a == x && b == y) || (a == y && b == x)
+}
+
+fn pair_is_voted(group: &GroupState, a: &ItemId, b: &ItemId) -> bool {
+    let Some(&ai) = group.item_to_idx.get(a) else {
+        return false;
+    };
+    let Some(&bi) = group.item_to_idx.get(b) else {
+        return false;
+    };
+    let (i, j) = if ai < bi { (ai, bi) } else { (bi, ai) };
+    group.voted_pairs.contains(&(i, j))
+}
+
+/// Component id per pool item: voted-pairs graph components plus one id per
+/// never-voted child.
+fn component_ids(group: &GroupState, pool: &[ItemId]) -> HashMap<ItemId, usize> {
+    let n = group.idx_to_item.len();
+    let (comps, isolates) =
+        connected_components_from_voted_pairs(n, group.voted_pairs.iter().copied());
+
+    let mut out: HashMap<ItemId, usize> = HashMap::new();
+    for (comp_idx, comp) in comps.iter().enumerate() {
+        for &idx in comp {
+            if idx < n {
+                out.insert(group.idx_to_item[idx].clone(), comp_idx);
+            }
+        }
+    }
+    let mut next = comps.len();
+    for &idx in &isolates {
+        if idx < n {
+            out.insert(group.idx_to_item[idx].clone(), next);
+            next += 1;
+        }
+    }
+    for item in pool {
+        out.entry(item.clone()).or_insert_with(|| {
+            let id = next;
+            next += 1;
+            id
+        });
+    }
+    out
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)]
+enum PairPriority {
+    /// Unvoted edge between two components — grows the ranking group.
+    BridgeUnvoted = 0,
+    /// Unvoted edge inside one component — refines order.
+    WithinUnvoted = 1,
+    /// Re-vote across components (rare once merged).
+    BridgeVoted = 2,
+    /// Re-vote within a component.
+    WithinVoted = 3,
+}
+
+fn pair_priority(
+    group: &GroupState,
+    components: &HashMap<ItemId, usize>,
+    a: &ItemId,
+    b: &ItemId,
+) -> PairPriority {
+    let voted = pair_is_voted(group, a, b);
+    let bridge = components.get(a) != components.get(b);
+    match (bridge, voted) {
+        (true, false) => PairPriority::BridgeUnvoted,
+        (false, false) => PairPriority::WithinUnvoted,
+        (true, true) => PairPriority::BridgeVoted,
+        (false, true) => PairPriority::WithinVoted,
+    }
+}
+
+/// All unordered pairs from `pool`, optionally skipping `exclude`.
+fn candidate_pairs(
+    pool: &[ItemId],
+    exclude: Option<(&ItemId, &ItemId)>,
+) -> Vec<(ItemId, ItemId)> {
+    let mut out = Vec::new();
+    for i in 0..pool.len() {
+        for j in (i + 1)..pool.len() {
+            let a = &pool[i];
+            let b = &pool[j];
+            if a == b {
+                continue;
+            }
+            if exclude.is_some_and(|(x, y)| pairs_match(a, b, x, y)) {
+                continue;
+            }
+            out.push((a.clone(), b.clone()));
+        }
+    }
+    out
+}
+
+/// Pick the next pair to vote on within `pool`.
+///
+/// 1. Prefer unvoted **bridge** pairs (connect separate ranking components).
+/// 2. Then unvoted within-component pairs (refinement).
+/// 3. Then already-voted pairs (re-compare).
+pub fn suggest_next_pair_in_pool(
+    group: &GroupState,
+    pool: &[ItemId],
+    exclude: Option<(&ItemId, &ItemId)>,
+) -> Option<(ItemId, ItemId)> {
+    let candidates = candidate_pairs(pool, exclude);
+    if candidates.is_empty() {
+        return None;
+    }
+    let components = component_ids(group, pool);
+    let best = candidates
+        .iter()
+        .map(|(a, b)| (pair_priority(group, &components, a, b), (a, b)))
+        .min_by_key(|(p, _)| *p)?
+        .0;
+    let best_pairs: Vec<(ItemId, ItemId)> = candidates
+        .into_iter()
+        .filter(|(a, b)| pair_priority(group, &components, a, b) == best)
+        .collect();
+    best_pairs.choose(&mut rand::thread_rng()).cloned()
+}
+
+/// Random distinct pair from `children` (legacy pair.rs behavior).
+pub fn random_pair(children: &[ItemId]) -> Option<(ItemId, ItemId)> {
+    if children.len() < 2 {
+        return None;
+    }
+    let left = children.choose(&mut rand::thread_rng())?;
+    let mut right = children.choose(&mut rand::thread_rng())?;
+    let mut guard = 0;
+    while left == right && guard < 32 {
+        right = children.choose(&mut rand::thread_rng())?;
+        guard += 1;
+    }
+    if left == right {
+        return None;
+    }
+    Some((left.clone(), right.clone()))
+}
+
+/// Sorted children of `parent` from the global tree.
+pub fn children_of(tree: &GlobalTree, parent: &ItemId) -> Vec<ItemId> {
+    let Some(node) = tree.get(parent) else {
+        return Vec::new();
+    };
+    let mut children: Vec<ItemId> = node.children.iter().cloned().collect();
+    children.sort_by(|a, b| a.as_str().cmp(b.as_str()));
+    children
+}
+
+/// Resolve a pair to compare under `parent`.
+pub fn resolve_pair(
+    tree: &GlobalTree,
+    parent: &ItemId,
+    left: Option<&ItemId>,
+    right: Option<&ItemId>,
+) -> Result<(ItemId, ItemId), PairError> {
+    let children = children_of(tree, parent);
+    if children.len() < 2 {
+        return Err(PairError::TooFewChildren);
+    }
+    let child_set: HashSet<_> = children.iter().collect();
+
+    match (left, right) {
+        (Some(l), Some(r)) => {
+            if l == r {
+                return Err(PairError::SameItem);
+            }
+            if !child_set.contains(l) || !child_set.contains(r) {
+                return Err(PairError::NotChild);
+            }
+            Ok((l.clone(), r.clone()))
+        }
+        (None, None) => {
+            let group = tree
+                .get(parent)
+                .map(|n| &n.local_ranking)
+                .cloned()
+                .unwrap_or_default();
+            suggest_next_pair_in_pool(&group, &children, None).ok_or(PairError::NoPair)
+        }
+        _ => Err(PairError::IncompletePair),
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum PairError {
+    TooFewChildren,
+    SameItem,
+    NotChild,
+    IncompletePair,
+    NoPair,
+}
+
+impl PairError {
+    pub fn status_message(&self) -> (&'static str, axum::http::StatusCode) {
+        match self {
+            Self::TooFewChildren => (
+                "parent needs at least 2 children to vote",
+                axum::http::StatusCode::BAD_REQUEST,
+            ),
+            Self::SameItem => (
+                "left and right must differ",
+                axum::http::StatusCode::BAD_REQUEST,
+            ),
+            Self::NotChild => (
+                "left and right must be children of parent",
+                axum::http::StatusCode::BAD_REQUEST,
+            ),
+            Self::IncompletePair => (
+                "provide both left and right, or neither",
+                axum::http::StatusCode::BAD_REQUEST,
+            ),
+            Self::NoPair => (
+                "no pair available",
+                axum::http::StatusCode::BAD_REQUEST,
+            ),
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::reducer::{GlobalTree, VoteData};
+
+    fn seed_children(parent: &ItemId, ids: &[&str]) -> GlobalTree {
+        let mut tree = GlobalTree::new();
+        tree.ensure_path(parent);
+        for id in ids {
+            let child = ItemId::parse(id).unwrap();
+            tree.ensure_path(&child);
+            if let Some(p) = tree.nodes.get_mut(parent) {
+                p.children.insert(child);
+            }
+        }
+        tree
+    }
+
+    fn pair_set(pair: &(ItemId, ItemId)) -> HashSet<&str> {
+        [pair.0.as_str(), pair.1.as_str()].into_iter().collect()
+    }
+
+    #[test]
+    fn suggest_prefers_unvoted_pair() {
+        let parent = ItemId::parse("reddit.com/r/rust").unwrap();
+        let mut tree = seed_children(
+            &parent,
+            &[
+                "reddit.com/r/rust/a",
+                "reddit.com/r/rust/b",
+                "reddit.com/r/rust/c",
+            ],
+        );
+        let vote =
+            VoteData::from_recorded(1, "reddit.com/r/rust/a", "reddit.com/r/rust/b", 2, 1).unwrap();
+        tree.apply_vote(&parent, vote);
+        let group = tree.get(&parent).unwrap().local_ranking.clone();
+        let pool = children_of(&tree, &parent);
+        let (l, r) = suggest_next_pair_in_pool(&group, &pool, None).unwrap();
+        let voted_ab = (l.as_str() == "reddit.com/r/rust/a" && r.as_str() == "reddit.com/r/rust/b")
+            || (l.as_str() == "reddit.com/r/rust/b" && r.as_str() == "reddit.com/r/rust/a");
+        assert!(!voted_ab);
+    }
+
+    #[test]
+    fn suggest_bridges_separate_components() {
+        let parent = ItemId::parse("reddit.com/r/rust").unwrap();
+        let mut tree = seed_children(
+            &parent,
+            &[
+                "reddit.com/r/rust/a",
+                "reddit.com/r/rust/b",
+                "reddit.com/r/rust/c",
+                "reddit.com/r/rust/d",
+            ],
+        );
+        let ab = VoteData::from_recorded(1, "reddit.com/r/rust/a", "reddit.com/r/rust/b", 2, 1).unwrap();
+        let cd = VoteData::from_recorded(2, "reddit.com/r/rust/c", "reddit.com/r/rust/d", 2, 1).unwrap();
+        tree.apply_vote(&parent, ab);
+        tree.apply_vote(&parent, cd);
+        let group = tree.get(&parent).unwrap().local_ranking.clone();
+        let pool = children_of(&tree, &parent);
+        let pair = suggest_next_pair_in_pool(&group, &pool, None).unwrap();
+        let chosen = pair_set(&pair);
+        let from_ab = chosen.contains("reddit.com/r/rust/a") || chosen.contains("reddit.com/r/rust/b");
+        let from_cd = chosen.contains("reddit.com/r/rust/c") || chosen.contains("reddit.com/r/rust/d");
+        assert!(from_ab && from_cd, "expected bridge pair, got {:?}", chosen);
+    }
+
+    #[test]
+    fn suggest_connects_isolate_to_existing_component() {
+        let parent = ItemId::parse("reddit.com/r/rust").unwrap();
+        let mut tree = seed_children(
+            &parent,
+            &[
+                "reddit.com/r/rust/a",
+                "reddit.com/r/rust/b",
+                "reddit.com/r/rust/c",
+            ],
+        );
+        let ab = VoteData::from_recorded(1, "reddit.com/r/rust/a", "reddit.com/r/rust/b", 2, 1).unwrap();
+        tree.apply_vote(&parent, ab);
+        let group = tree.get(&parent).unwrap().local_ranking.clone();
+        let pool = children_of(&tree, &parent);
+        let pair = suggest_next_pair_in_pool(&group, &pool, None).unwrap();
+        let chosen = pair_set(&pair);
+        assert!(chosen.contains("reddit.com/r/rust/c"));
+        assert!(chosen.contains("reddit.com/r/rust/a") || chosen.contains("reddit.com/r/rust/b"));
+    }
+
+    #[test]
+    fn resolve_pair_picks_from_pool() {
+        let parent = ItemId::parse("reddit.com/r/rust").unwrap();
+        let tree = seed_children(&parent, &["reddit.com/r/rust/a", "reddit.com/r/rust/b"]);
+        let pair = resolve_pair(&tree, &parent, None, None).unwrap();
+        let pool: HashSet<_> = ["reddit.com/r/rust/a", "reddit.com/r/rust/b"]
+            .into_iter()
+            .collect();
+        assert!(pool.contains(pair.0.as_str()));
+        assert!(pool.contains(pair.1.as_str()));
+    }
+}
diff --git a/server/src/path_types.rs b/server/src/path_types.rs
index 12dce9888f5cd4e1a0974d12d6468368d0f775b9..a0d1a028b8bd3c5d1c714dea3ec597d1ac39e1bc 100644
--- a/server/src/path_types.rs
+++ b/server/src/path_types.rs
@@ -40,6 +40,22 @@ impl ItemId {
         Self::canonicalize(raw_url).map(Self)
     }
 
+    /// Normalize strings from forms, events, and Reddit imports into the same
+    /// stored id shape (e.g. drop post title slug after comment id).
+    pub fn from_storage(s: &str) -> Option<Self> {
+        let t = s.trim();
+        if t.is_empty() {
+            return None;
+        }
+        if t.contains("://") || t.starts_with("r/") {
+            return Self::from_url(t).or_else(|| Self::parse(t));
+        }
+        if t.starts_with("reddit.com/") && t.contains("/comments/") {
+            return Self::from_url(t).or_else(|| Self::parse(t));
+        }
+        Self::parse(t).or_else(|| Self::from_url(t))
+    }
+
     /// Map legacy scope keys (`""`, `"rust"`) to fractal parent nodes.
     pub fn from_legacy_scope(raw: &str) -> Self {
         let s = raw.trim();
@@ -324,6 +340,12 @@ mod tests {
         assert_eq!(id.as_str(), "reddit.com/r/amitheasshole");
     }
 
+    #[test]
+    fn from_storage_strips_post_title_slug() {
+        let id = ItemId::from_storage("reddit.com/r/rust/comments/aaa/announcing_rust_199").unwrap();
+        assert_eq!(id.as_str(), "reddit.com/r/rust/comments/aaa");
+    }
+
     #[test]
     fn from_browse_uri_strips_prefix() {
         let id = ItemId::from_browse_uri("/~/https://reddit.com/r/rust").unwrap();
diff --git a/server/src/reddit.rs b/server/src/reddit.rs
index 1168ec2afc77c092514eec91b513bac301cbe225..626454e5a2f638734193b5190a2beea286af85b6 100644
--- a/server/src/reddit.rs
+++ b/server/src/reddit.rs
@@ -567,7 +567,7 @@ fn parse_children(_parent: &ItemId, payload: &Value) -> Vec<(ItemId, Value)> {
             _ => continue,
         };
         let path = format!("reddit.com{}", permalink.trim_end_matches('/'));
-        if let Some(id) = ItemId::parse(&path) {
+        if let Some(id) = ItemId::from_storage(&path) {
             out.push((id, child.clone()));
         }
     }
diff --git a/server/src/reducer.rs b/server/src/reducer.rs
index 336e78b77d3ab59361b89af5c9868e13da4ac962..fc23f41137d7df33997afe19c533505c250cc305 100644
--- a/server/src/reducer.rs
+++ b/server/src/reducer.rs
@@ -28,8 +28,8 @@ impl VoteData {
         ratio_left: i32,
         ratio_right: i32,
     ) -> Option<Self> {
-        let a = ItemId::parse(a)?;
-        let b = ItemId::parse(b)?;
+        let a = ItemId::from_storage(a)?;
+        let b = ItemId::from_storage(b)?;
         if a == b {
             return None;
         }
@@ -85,8 +85,8 @@ impl GroupState {
     }
 
     pub fn apply_vote(&mut self, mut vote: VoteData) {
-        vote.a = ItemId::parse(vote.a.as_str()).unwrap_or_else(|| vote.a.clone());
-        vote.b = ItemId::parse(vote.b.as_str()).unwrap_or_else(|| vote.b.clone());
+        vote.a = ItemId::from_storage(vote.a.as_str()).unwrap_or(vote.a.clone());
+        vote.b = ItemId::from_storage(vote.b.as_str()).unwrap_or(vote.b.clone());
         if vote.ratio_left < 0 {
             vote.ratio_left = 0;
         }
diff --git a/server/src/state.rs b/server/src/state.rs
index d238701208a1c708b94a778a6a2e1891a678ecbc..4c3008e73cc74d8483a2064a0a280e73d82a01ec 100644
--- a/server/src/state.rs
+++ b/server/src/state.rs
@@ -18,7 +18,7 @@ pub fn parse_item_param(raw: &str) -> ItemId {
     if s.is_empty() {
         return ItemId::root();
     }
-    ItemId::from_url(s).or_else(|| ItemId::parse(s)).unwrap_or_else(|| ItemId::opaque(s))
+    ItemId::from_storage(s).unwrap_or_else(|| ItemId::opaque(s))
 }
 
 /// Legacy: normalize raw ranking subject into a scope key for old event replay.
diff --git a/server/src/ui_action.rs b/server/src/ui_action.rs
index 53581e1362bfcc5dfb4ae3069c41ea6f7be41437..237411a689fb10ad1b7022ea66aee1969c7507ec 100644
--- a/server/src/ui_action.rs
+++ b/server/src/ui_action.rs
@@ -31,9 +31,9 @@ pub enum HtmlUiAction {
         /// Parent node [`ItemId`] string; empty = tree root.
         #[serde(default)]
         scope: String,
-        /// After vote, navigate here (vote compare page).
+        /// Posted from `/vote` compare UI — morph edge history in place.
         #[serde(default)]
-        next: String,
+        vote_compare: bool,
     },
     /// Parse pasted Reddit URL/path; redirect to subreddit ranking on success.
     ParseQuery {
@@ -98,7 +98,7 @@ mod tests {
                 ratio_left: 60,
                 ratio_right: 40,
                 scope: "parent".into(),
-                next: String::new(),
+                vote_compare: false,
             }
         );
     }
@@ -129,7 +129,7 @@ mod tests {
                 ratio_left: 2,
                 ratio_right: 1,
                 scope: "amitheasshole".into(),
-                next: String::new(),
+                vote_compare: false,
             }
         );
     }
@@ -156,7 +156,7 @@ mod tests {
                 ratio_left: 2,
                 ratio_right: 1,
                 scope: String::new(),
-                next: String::new(),
+                vote_compare: false,
             }
         );
     }
diff --git a/server/static/sorter.css b/server/static/sorter.css
index 3c2cec5cb5094013387147c621d75bd341bc06cc..b95718ce463b200a5667d3014dac2119836a0bef 100644
--- a/server/static/sorter.css
+++ b/server/static/sorter.css
@@ -289,11 +289,27 @@ h1 {
 }
 
 .vote-compare-nav {
+  margin: 1rem 0 0.5rem;
+}
+
+.vote-compare-actions {
   display: flex;
-  justify-content: space-between;
+  flex-wrap: wrap;
   align-items: center;
-  gap: 1rem;
-  margin: 1rem 0;
+  gap: 0.75rem;
+  margin-top: 0.5rem;
+}
+
+.vote-compare-actions .btn-secondary {
+  margin-top: 0;
+  display: inline-block;
+  text-decoration: none;
+  line-height: 1.4;
+}
+
+.vote-compare-actions .vote-compare-next.is-disabled {
+  opacity: 0.6;
+  cursor: default;
 }
 
 .vote-compare-next {
diff --git a/server/static/sorter_ui.js b/server/static/sorter_ui.js
index c8d7c3ef2413399ced7898a7edb3e7a2c8e16163..ab15d27f508cc6f5fbca9d8926fd0ca26d64857b 100644
--- a/server/static/sorter_ui.js
+++ b/server/static/sorter_ui.js
@@ -104,11 +104,10 @@
       if (f.getAttribute('data-navigate') === 'full') return;
       e.preventDefault();
       await postUiForm(f);
-      if (f.id === 'vote-form' || f.id === 'vote-compare-form') {
+      if (f.id === 'vote-form') {
         f.reset();
-        var slider = f.querySelector('#vote-preference-slider');
-        if (slider) slider.value = '50';
-        initVoteSlider();
+        var firstField = f.querySelector('input[type="text"]');
+        if (firstField) firstField.focus();
       }
     });
   }
diff --git a/server/tests/integration_ui.rs b/server/tests/integration_ui.rs
index df7d9ab357531c5146d0b326c149fdaa6b531a4a..56d2db313b313ffce07eff38a8ccdaa2fbb9498f 100644
--- a/server/tests/integration_ui.rs
+++ b/server/tests/integration_ui.rs
@@ -25,6 +25,53 @@ async fn start_test_server() -> (SocketAddr, TempDir) {
     (addr, tmp)
 }
 
+#[tokio::test]
+async fn post_ui_vote_compare_morphs_edge_history() {
+    let (addr, _tmp) = start_test_server().await;
+    let parent = "reddit.com/r/rust";
+    let a = "reddit.com/r/rust/comments/aaa/announcing_rust_199";
+    let b = "reddit.com/r/rust/comments/bbb/what_are_you_working_on";
+
+    let rpc = serde_json::json!({
+        "action": "record_vote",
+        "a": a,
+        "b": b,
+        "ratio_left": {"$form:i32": "ratio_left"},
+        "ratio_right": {"$form:i32": "ratio_right"},
+        "scope": parent,
+        "vote_compare": true,
+    })
+    .to_string();
+    let mut form = HashMap::new();
+    form.insert(UI_RPC_FIELD.to_string(), rpc);
+    form.insert("ratio_left".into(), "70".into());
+    form.insert("ratio_right".into(), "30".into());
+
+    let client = reqwest::Client::new();
+    let body = client
+        .post(format!("http://{addr}/ui"))
+        .form(&form)
+        .send()
+        .await
+        .unwrap()
+        .text()
+        .await
+        .unwrap();
+
+    assert!(
+        body.contains("vote-edge-history"),
+        "expected edge history morph, got: {body}"
+    );
+    assert!(
+        body.contains("70:30"),
+        "expected recorded ratio in morph, got: {body}"
+    );
+    assert!(
+        !body.contains("no votes on this pair yet"),
+        "should not show empty edge history after vote, got: {body}"
+    );
+}
+
 #[tokio::test]
 async fn post_ui_record_vote_morphs_ranking_and_persists() {
     let (addr, tmp) = start_test_server().await;
diff --git a/test/reddit_import.clj b/test/reddit_import.clj
index 45a2a19f20799d77e84d8aa64735ab5e7e45f97c..b476488526252c13fd73bdda76e5201678e4a714 100644
--- a/test/reddit_import.clj
+++ b/test/reddit_import.clj
@@ -2,9 +2,8 @@
   (:require [babashka.process :as process]
             [clojure.java.io :as io]
             [clojure.string :as str]
-            [clojure.test :refer [deftest is testing]])
-  (:import [com.sun.net.httpserver HttpServer HttpHandler HttpExchange]
-           [java.net InetSocketAddress]))
+            [clojure.test :refer [deftest is testing]]
+            [test.support.mock-reddit :as mock-reddit]))
 
 (defn- repo-root []
   (.getCanonicalPath (io/file (System/getProperty "user.dir"))))
@@ -13,29 +12,6 @@
   (with-open [s (java.net.ServerSocket. 0)]
     (.getLocalPort s)))
 
-(defn- start-mock-reddit [port fixtures-dir]
-  (let [about (.getBytes (slurp (io/file fixtures-dir "r_rust_about.json")) "UTF-8")
-        listing (.getBytes (slurp (io/file fixtures-dir "r_rust_listing.json")) "UTF-8")
-        server (HttpServer/create (InetSocketAddress. "127.0.0.1" port) 0)
-        handler
-        (proxy [HttpHandler] []
-          (handle [^HttpExchange exchange]
-            ;; Route by path: `/r/<sub>/about.json` is the subreddit entity,
-            ;; `/r/<sub>.json` is the children listing.
-            (let [path (.getPath (.getRequestURI exchange))
-                  body (if (str/includes? path "/about")
-                         about
-                         listing)]
-              (.sendResponseHeaders exchange 200 (alength body))
-              (let [out (.getResponseBody exchange)]
-                (.write out body)
-                (.close out)))))]
-    (.createContext server "/" handler)
-    (.setExecutor server nil)
-    (.start server)
-    (fn stop []
-      (.stop server 0))))
-
 (defn- wait-health [base-url ms]
   (let [deadline (+ (System/currentTimeMillis) ms)
         url (str base-url "/healthz")]
@@ -100,7 +76,7 @@
 (deftest reddit-fetch-via-mock-api
   (testing "Fetch more queues import; event log stores full payload; page shows title"
     (let [root (repo-root)
-          fixtures (str root "/test/fixtures/reddit")
+          fixtures (mock-reddit/fixtures-dir root)
           data-dir (.getAbsolutePath
                     (doto (io/file (System/getProperty "java.io.tmpdir")
                                    (str "sorter2-reddit-" (System/currentTimeMillis)))
@@ -110,7 +86,7 @@
           reddit-base (str "http://127.0.0.1:" reddit-port)
           app-base (str "http://127.0.0.1:" app-port)
           bin (str root "/target/release/sorter2-server")
-          stop-mock (start-mock-reddit reddit-port fixtures)]
+          stop-mock (mock-reddit/start-mock-reddit reddit-port fixtures)]
       (try
         (is (zero? (:exit (process/shell {:dir root}
                                          "cargo" "build" "--release" "--package" "sorter2-server")))
diff --git a/test/support/mock_reddit.clj b/test/support/mock_reddit.clj
new file mode 100644
index 0000000000000000000000000000000000000000..5efa92db3e1f79b8f423a2f1adcbda959123c1ad
--- /dev/null
+++ b/test/support/mock_reddit.clj
@@ -0,0 +1,35 @@
+(ns test.support.mock-reddit
+  "In-process HTTP stub for Reddit API fixtures (`test/fixtures/reddit/`)."
+  (:require [clojure.java.io :as io]
+            [clojure.string :as str])
+  (:import [com.sun.net.httpserver HttpServer HttpHandler HttpExchange]
+           [java.net InetSocketAddress]))
+
+(defn fixtures-dir
+  ([] (fixtures-dir (System/getProperty "user.dir")))
+  ([root] (str root "/test/fixtures/reddit")))
+
+(defn start-mock-reddit
+  "Start a mock Reddit API on `port`. Returns a zero-arg `stop` function."
+  ([port] (start-mock-reddit port (fixtures-dir)))
+  ([port dir]
+   (let [about (.getBytes (slurp (io/file dir "r_rust_about.json")) "UTF-8")
+         listing (.getBytes (slurp (io/file dir "r_rust_listing.json")) "UTF-8")
+         server (HttpServer/create (InetSocketAddress. "127.0.0.1" port) 0)
+         handler
+         (proxy [HttpHandler] []
+           (handle [^HttpExchange exchange]
+             ;; `/r/<sub>/about.json` → subreddit entity; `/r/<sub>.json` → listing.
+             (let [path (.getPath (.getRequestURI exchange))
+                   body (if (str/includes? path "/about")
+                          about
+                          listing)]
+               (.sendResponseHeaders exchange 200 (alength body))
+               (let [out (.getResponseBody exchange)]
+                 (.write out body)
+                 (.close out)))))]
+     (.createContext server "/" handler)
+     (.setExecutor server nil)
+     (.start server)
+     (fn stop []
+       (.stop server 0)))))
diff --git a/test/vote_compare.clj b/test/vote_compare.clj
new file mode 100644
index 0000000000000000000000000000000000000000..fbc00281e466a1ba0833193c7b96fa30aeb45a5c
--- /dev/null
+++ b/test/vote_compare.clj
@@ -0,0 +1,101 @@
+(ns test.vote-compare
+  (:require [babashka.process :as process]
+            [clojure.java.io :as io]
+            [clojure.string :as str]
+            [clojure.test :refer [deftest is testing]]
+            [com.blockether.spel.core :as core]
+            [com.blockether.spel.locator :as loc]
+            [com.blockether.spel.page :as page]
+            [test.support.mock-reddit :as mock-reddit])
+  (:import [java.net URLEncoder]))
+
+(defn- repo-root []
+  (.getCanonicalPath (io/file (System/getProperty "user.dir"))))
+
+(defn- pick-port []
+  (with-open [s (java.net.ServerSocket. 0)]
+    (.getLocalPort s)))
+
+(defn- wait-health [base-url ms]
+  (let [deadline (+ (System/currentTimeMillis) ms)
+        url (str base-url "/healthz")]
+    (loop []
+      (let [resp (try
+                   (process/shell {:out :string :err :string}
+                                  "curl" "-sf" url)
+                   (catch Exception _ nil))]
+        (if (and resp (zero? (:exit resp)) (= "ok" (str/trim (:out resp ""))))
+          true
+          (if (< (System/currentTimeMillis) deadline)
+            (do (Thread/sleep 200) (recur))
+            false))))))
+
+(defn- curl-fetch-children [base item]
+  (process/shell {:out :string :err :string}
+                 "curl" "-sfN" "--max-time" "20"
+                 "-X" "POST" (str base "/ui")
+                 "--data-urlencode"
+                 (str "__rpc__={\"action\":\"fetch_entity\",\"item\":\"" item
+                      "\",\"kind\":\"children\"}")))
+
+(defn- vote-page-url [base parent]
+  (str base "/vote?parent="
+       (URLEncoder/encode parent "UTF-8")))
+
+(deftest vote-compare-shows-recorded-vote-after-post
+  (testing "post vote on /vote morphs edge history (mock Reddit children seeded)"
+    (let [root (repo-root)
+          fixtures (mock-reddit/fixtures-dir root)
+          data-dir (.getAbsolutePath
+                    (doto (io/file (System/getProperty "java.io.tmpdir")
+                                   (str "sorter2-vote-" (System/currentTimeMillis)))
+                      (.mkdirs)))
+          reddit-port (pick-port)
+          app-port (pick-port)
+          reddit-base (str "http://127.0.0.1:" reddit-port)
+          app-base (str "http://127.0.0.1:" app-port)
+          bin (str root "/target/release/sorter2-server")
+          stop-mock (mock-reddit/start-mock-reddit reddit-port fixtures)]
+      (try
+        (is (zero? (:exit (process/shell {:dir root}
+                                         "cargo" "build" "--release" "--package" "sorter2-server")))
+            "release build succeeds")
+        (let [proc (process/process {:dir root
+                                     :env (into (into {} (System/getenv))
+                                                {"SORTER2_SKIP_DOTENV" "1"
+                                                 "SORTER2_DATA_DIR" data-dir
+                                                 "SORTER2_EVENT_LOG" (str data-dir "/events.jsonl")
+                                                 "PORT" (str app-port)
+                                                 "REDDIT_API_BASE" reddit-base
+                                                 "REDDIT_OAUTH_BASE" reddit-base
+                                                 "REDDIT_CLIENT_ID" ""
+                                                 "REDDIT_CLIENT_SECRET" ""
+                                                 "REDDIT_APP_ID" ""
+                                                 "REDDIT_APP_SECRET" ""})
+                                     :out :string
+                                     :err :string}
+                                    bin)]
+          (try
+            (is (wait-health app-base 20000) "app healthz")
+            (let [fetch (curl-fetch-children app-base "reddit.com/r/rust")]
+              (is (zero? (:exit fetch)) "fetch posts via mock Reddit")
+              (is (str/includes? (:out fetch) "Idiomorph.morph")))
+            (core/with-testing-page [pg]
+              (page/navigate pg (vote-page-url app-base "reddit.com/r/rust"))
+              (page/wait-for-selector pg "#vote-compare-form")
+              (let [before (loc/text-content (page/locator pg "#vote-edge-history-region"))]
+                (is (str/includes? before "no votes on this pair yet")
+                    "empty edge history before first vote"))
+              (loc/click (page/get-by-test-id pg "vote-post"))
+              (page/wait-for-selector pg ".vote-edge-history-title")
+              (let [after (loc/text-content (page/locator pg "#vote-edge-history-region"))]
+                (is (str/includes? after "votes on this pair")
+                    "shows edge history title after vote")
+                (is (str/includes? after "50:50")
+                    "shows submitted ratio after vote")
+                (is (not (str/includes? after "no votes on this pair yet"))
+                    "does not revert to empty edge history")))
+            (finally
+              (process/destroy proc))))
+        (finally
+          (stop-mock))))))