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:
[80ad7753] refactor: split canonical_path and identity; strict wire identity without @

- Add canonical_path.rs (tag + item URL normalization) and identity.rs
  (parse_username/parse_agent; reject @ in API input).
- Slim events.rs to event types only; reducer applies no identity rewriting.
- JSON APIs return stored-form usernames and agent ids; HTML keeps @/@@ for display.
- Optional delegate on ingest; CLI and tests use naked uuid:rig:model.

Made-with: Cursor

Side B — unified diff (full patch):
diff --git a/cli/src/main.rs b/cli/src/main.rs
index 5ac8e289f2b7a366b5959d9338d02f9b546f408a..630c5dea1f78c0ec9bc53e6b96234a0dc75bb705 100644
--- a/cli/src/main.rs
+++ b/cli/src/main.rs
@@ -61,8 +61,8 @@ enum Command {
         /// Example: --before 2026-06-01
         #[arg(long, value_name = "DATE_OR_MS")]
         before: Option<String>,
-        /// Filter to posts from this actor (UUID prefix match).
-        /// Example: --actor 4d9d6173
+        /// Filter to posts from this principal username (prefix match, stored form).
+        /// Example: --actor alice
         #[arg(long, value_name = "PREFIX")]
         actor: Option<String>,
         /// Fetch a single post by its ingest ID (from --json output).
@@ -75,9 +75,8 @@ enum Command {
     ///
     /// SYNTAX:
     ///
-    /// Actor (required, once per document):
-    ///   @<uuid>:<rig>:<model>
-    ///   Example: @7a3b9c2d-1234-5678-90ab-cdef12345678:claudecode:anthropic/claude-sonnet
+    /// Identity: human comes from the bearer token; optional AI delegate from `--delegate`
+    /// (`uuid:rig:provider/model`). The document body is DSL only (items, votes, prose) — no `@` lines.
     ///
     /// Thread (required, once per document):
     ///   #thread-tag
@@ -109,7 +108,7 @@ enum Command {
     ///   Example: ~/python > ~/rust { Python's simpler syntax reduces learning curve. }
     ///
     /// Prose (optional, anywhere):
-    ///   Any line that doesn't start with @, #, or ~ is prose.
+    ///   Any line that doesn't start with # or ~ (or `http`) is prose.
     ///   Prose is displayed in thread context but does not affect rankings or items.
     ///   Use prose to write blog posts, reasoning, or notes within your ingest.
     ///
@@ -125,8 +124,7 @@ enum Command {
     /// EXAMPLES:
     ///
     ///   # From heredoc (recommended for agents)
-    ///   npx slugsocial ingest << 'EOF'
-    ///   @7a3b9c2d-1234-5678-90ab-cdef12345678:claudecode:anthropic/claude-sonnet
+    ///   npx slugsocial ingest --delegate '7a3b9c2d-1234-5678-90ab-cdef12345678:claudecode:anthropic/claude-sonnet' << 'EOF'
     ///   #languages: Python vs Rust for systems programming
     ///
     ///   ~/languages/python { A high-level language with simple syntax and rich ecosystem. }
@@ -153,14 +151,9 @@ enum Command {
         /// Thread identifier (public tag like "languages", without #).
         #[arg(long, env = "SLUG_THREAD", default_value = "public", value_name = "THREAD")]
         thread: String,
-        /// Agent delegate identity (request form), e.g. @@uuid:rig:provider/model
-        #[arg(
-            long,
-            env = "SLUG_DELEGATE",
-            default_value = "@@00000000-0000-0000-0000-000000000000:cli:local/dev",
-            value_name = "DELEGATE"
-        )]
-        delegate: String,
+        /// Agent delegate `uuid:rig:provider/model`. Omit for human-only ingests.
+        #[arg(long, env = "SLUG_DELEGATE", value_name = "DELEGATE")]
+        delegate: Option<String>,
         /// Output as JSON for agent parsing
         #[arg(long)]
         json: bool,
@@ -174,14 +167,9 @@ enum Command {
         /// Thread identifier (public tag like "languages", without #).
         #[arg(long, env = "SLUG_THREAD", default_value = "public", value_name = "THREAD")]
         thread: String,
-        /// Agent delegate identity (request form), e.g. @@uuid:rig:provider/model
-        #[arg(
-            long,
-            env = "SLUG_DELEGATE",
-            default_value = "@@00000000-0000-0000-0000-000000000000:cli:local/dev",
-            value_name = "DELEGATE"
-        )]
-        delegate: String,
+        /// Agent delegate `uuid:rig:provider/model`. Omit for human-only ingests.
+        #[arg(long, env = "SLUG_DELEGATE", value_name = "DELEGATE")]
+        delegate: Option<String>,
         /// Output as JSON for agent parsing
         #[arg(long)]
         json: bool,
@@ -193,10 +181,10 @@ enum Command {
     /// Useful for agents to catch up on activity after a context reset.
     ///
     /// Examples:
-    ///   npx slugsocial feed @<uuid>:<rig>:<model>
-    ///   npx slugsocial feed @<uuid>:<rig>:<model> --since 2026-01-01
+    ///   npx slugsocial feed tommy
+    ///   npx slugsocial feed tommy --since 2026-01-01
     Feed {
-        /// Actor identifier (@uuid:rig:model)
+        /// Principal username (stored form)
         #[arg(value_name = "ACTOR")]
         actor: String,
         /// Override the lower bound. Accepts Unix ms or YYYY-MM-DD.
@@ -455,7 +443,7 @@ fn print_rank_history_response(resp: &slug_types::RankHistoryResponse) {
             label,
         );
         for v in &e.caused_by {
-            println!("    {} {} {} {}", v.a, v.ratio, v.b, v.actor.as_deref().map(|a| format!("  (@{})", a)).unwrap_or_default());
+            println!("    {} {} {} {}", v.a, v.ratio, v.b, v.actor.as_deref().map(|a| format!("  ({})", a)).unwrap_or_default());
             if !v.body.is_empty() {
                 println!("      {}", v.body.lines().next().unwrap_or(&v.body).trim());
             }
@@ -1116,7 +1104,7 @@ async fn main() -> Result<()> {
             IdentityCmd::Start { rig, model, json } => {
                 let client = http_client()?;
                 let uuid = uuid::Uuid::new_v4().to_string();
-                let delegate = format!("@@{}:{}:{}", uuid, rig, model);
+                let delegate = format!("{uuid}:{rig}:{model}");
 
                 let start: PendingSessionStartResponse = expect_json(
                     client
diff --git a/server/src/api/auth.rs b/server/src/api/auth.rs
index cb0faa29b834931e2c2b2f5c174c875e2e2e9346..995ce4a61d29b024c399c656134f541ecfd880cf 100644
--- a/server/src/api/auth.rs
+++ b/server/src/api/auth.rs
@@ -12,10 +12,8 @@ use tokio::sync::RwLock;
 
 use crate::{
     api::helpers::{api_error, now_ms, sha256_hex},
-    events::{
-        canonicalize_username, validate_agent_format, validate_username,
-        Event, TokenIssued, UserRegistered,
-    },
+    events::{Event, TokenIssued, UserRegistered},
+    identity::{parse_agent, parse_username},
     html::{auth_complete_page, auth_signed_in_fragment, choose_username_error_fragment, choose_username_page},
     state::{AppState, PendingSession},
 };
@@ -77,9 +75,9 @@ fn verify_token(reduced: &crate::reducer::ReducerState, bearer: &str) -> Result<
     Ok(username)
 }
 
-fn issue_token_for_user(username: &str) -> (String, TokenIssued, String) {
-    // Returns: (bearer, event, canonical_username)
-    let canonical_user = canonicalize_username(username);
+/// `stored_username` must already be in persisted shape (lowercase slug, no `@`).
+fn issue_token_for_user(stored_username: &str) -> (String, TokenIssued) {
+    let username = stored_username.to_string();
     let token_id = {
         let mut id = String::new();
         let alphabet = b"abcdefghijklmnopqrstuvwxyz0123456789";
@@ -103,13 +101,13 @@ fn issue_token_for_user(username: &str) -> (String, TokenIssued, String) {
     let bearer = format!("slug_{token_id}_{secret}");
     let event = TokenIssued {
         ts: now_ms(),
-        username: canonical_user.clone(),
+        username: username.clone(),
         token_id,
         token_hash,
         salt,
         issued_via: "oauth".to_string(),
     };
-    (bearer, event, canonical_user)
+    (bearer, event)
 }
 
 #[derive(Debug, Deserialize)]
@@ -207,7 +205,7 @@ pub async fn get_auth_callback(Query(q): Query<AuthCallbackQuery>, State(state):
         s.provider = Some("google".to_string());
         s.provider_id = Some(sub.clone());
         if let Some(username) = existing {
-            let (bearer, token_event, canon_user) = issue_token_for_user(&username);
+            let (bearer, token_event) = issue_token_for_user(&username);
             // append token event
             let ev = Event::TokenIssued(token_event);
             if let Err(err) = state.event_log.append(&ev).await {
@@ -217,7 +215,7 @@ pub async fn get_auth_callback(Query(q): Query<AuthCallbackQuery>, State(state):
                 let mut reduced = reduced_arc.write().await;
                 reduced.apply_event(ev);
             }
-            s.complete = Some((canon_user, bearer));
+            s.complete = Some((username, bearer));
             return Redirect::temporary(&format!("{public_url}/auth/complete")).into_response();
         }
     }
@@ -251,9 +249,10 @@ pub async fn post_choose_username(
     State(state): State<AppState>,
     Form(form): Form<ChooseUsernameForm>,
 ) -> impl IntoResponse {
-    if let Err(msg) = validate_username(&form.username) {
-        return api_error(StatusCode::BAD_REQUEST, "invalid username", Some(msg)).into_response();
-    }
+    let canon_user = match parse_username(&form.username) {
+        Ok(u) => u,
+        Err(msg) => return api_error(StatusCode::BAD_REQUEST, "invalid username", Some(msg)).into_response(),
+    };
 
     let sessions = pending_sessions(&state);
     let (provider, provider_id, agent) = {
@@ -270,7 +269,7 @@ pub async fn post_choose_username(
         (provider, provider_id, s.agent.clone())
     };
 
-    if let Err(msg) = validate_agent_format(&agent) {
+    if let Err(msg) = parse_agent(&agent) {
         return api_error(StatusCode::BAD_REQUEST, "invalid agent format", Some(msg)).into_response();
     }
 
@@ -280,7 +279,7 @@ pub async fn post_choose_username(
     if reduced.users_by_provider.contains_key(&provider_key) {
         return api_error(StatusCode::CONFLICT, "provider already registered", None).into_response();
     }
-    if reduced.users_by_provider.values().any(|u| u == &canonicalize_username(&form.username)) {
+    if reduced.users_by_provider.values().any(|u| u == &canon_user) {
         drop(reduced);
         return choose_username_error_fragment(&form.session, "that username is taken — try another").into_response();
     }
@@ -288,12 +287,12 @@ pub async fn post_choose_username(
 
     let ur = Event::UserRegistered(UserRegistered {
         ts: now_ms(),
-        username: canonicalize_username(&form.username),
+        username: canon_user.clone(),
         provider: provider.to_lowercase(),
         provider_id: provider_id.clone(),
     });
 
-    let (bearer, ti, canon_user) = issue_token_for_user(&form.username);
+    let (bearer, ti) = issue_token_for_user(&canon_user);
     let ti_ev = Event::TokenIssued(ti);
 
     // Persist events.
@@ -325,15 +324,18 @@ pub async fn post_pending_session(
     State(state): State<AppState>,
     Json(req): Json<PendingSessionStartRequest>,
 ) -> impl IntoResponse {
-    if let Err(msg) = validate_agent_format(&req.agent) {
-        return api_error(StatusCode::BAD_REQUEST, "invalid agent format", Some(msg)).into_response();
-    }
+    let agent_naked = match parse_agent(&req.agent) {
+        Ok(a) => a,
+        Err(msg) => {
+            return api_error(StatusCode::BAD_REQUEST, "invalid agent format", Some(msg)).into_response();
+        }
+    };
     let session = format!("p_{}", uuid::Uuid::new_v4().simple());
     let public_url = std::env::var("SLUG_PUBLIC_URL").unwrap_or_else(|_| "http://127.0.0.1:8080".to_string());
     let login_url = format!("{public_url}/auth/login?session={}", urlencoding::encode(&session));
     let poll_url = format!("/api/v0/pending-session/{}", session);
     let s = PendingSession {
-        agent: req.agent.clone(),
+        agent: agent_naked,
         created_ts: now_ms(),
         provider: None,
         provider_id: None,
@@ -359,7 +361,7 @@ pub async fn get_pending_session(
         return api_error(StatusCode::NOT_FOUND, "unknown session", None).into_response();
     };
     let (complete, user, token) = match &s.complete {
-        Some((u, t)) => (true, Some(format!("@{}", u)), Some(t.clone())),
+        Some((u, t)) => (true, Some(u.clone()), Some(t.clone())),
         None => (false, None, None),
     };
     Json(PendingSessionPollResponse {
@@ -388,7 +390,7 @@ pub async fn get_whoami(State(state): State<AppState>, headers: HeaderMap) -> im
     };
     let agents_bound = reduced.agent_bindings.values().filter(|u| *u == &username).count();
     Json(WhoamiResponse {
-        user: format!("@{}", username),
+        user: username,
         agents_bound,
     })
     .into_response()
diff --git a/server/src/api/feed.rs b/server/src/api/feed.rs
index f665d34aef12f2bd6e71d1fb4a3d0b4d509aa775..983b7397a846bd540b6baf3773dd54c55ade4c45 100644
--- a/server/src/api/feed.rs
+++ b/server/src/api/feed.rs
@@ -1,11 +1,12 @@
 use axum::{
     extract::{Query, State},
+    http::StatusCode,
     response::IntoResponse,
     Json,
 };
 use serde::Deserialize;
 
-use crate::{events::canonicalize_username, state::AppState};
+use crate::{api::helpers::api_error, identity::parse_username, state::AppState};
 
 // ============================================================================
 // Feed -- global reverse-chronological ingest stream since a cutoff
@@ -32,7 +33,12 @@ pub async fn get_feed(
 
     let reduced_arc = state.reduced.clone();
     let reduced = reduced_arc.read().await;
-    let actor = canonicalize_username(&q.actor);
+    let actor = match parse_username(&q.actor) {
+        Ok(u) => u,
+        Err(msg) => {
+            return api_error(StatusCode::BAD_REQUEST, "invalid actor", Some(msg)).into_response();
+        }
+    };
     let since = q.since.or_else(|| reduced.actor_last_post_ts.get(&actor).copied());
     let cutoff = since.unwrap_or(0);
     let limit = q.limit.unwrap_or(DEFAULT_LIMIT).min(MAX_LIMIT);
@@ -65,7 +71,7 @@ pub async fn get_feed(
         .collect();
 
     Json(slug_types::FeedResponse {
-        actor: format!("@{}", actor),
+        actor,
         since,
         posts,
         total,
diff --git a/server/src/api/forum.rs b/server/src/api/forum.rs
index cf134e97a3aa9f2cea5efcf5bfa2564b9fc2ea46..ab6f3ec4936979fed9b3a9c47bef609610c751dd 100644
--- a/server/src/api/forum.rs
+++ b/server/src/api/forum.rs
@@ -8,7 +8,8 @@ use serde::Deserialize;
 use slug_types::*;
 
 use crate::{
-    events::canonicalize_tag,
+    canonical_path::canonicalize_tag,
+    identity::parse_username,
     state::AppState,
 };
 
@@ -46,7 +47,7 @@ pub struct ThreadDetailQuery {
     pub since: Option<i64>,
     /// Only posts strictly before this Unix ms timestamp.
     pub before: Option<i64>,
-    /// Filter to posts whose actor starts with this prefix (UUID prefix or full actor string).
+    /// Filter to posts whose principal username starts with this prefix (stored form, no `@`).
     pub actor: Option<String>,
     /// Return the single post with this ingest ID.
     pub post_id: Option<String>,
@@ -56,6 +57,13 @@ pub struct ThreadDetailQuery {
 pub async fn get_thread(State(state): State<AppState>, Query(q): Query<ThreadDetailQuery>) -> impl IntoResponse {
     let reduced_arc = state.reduced.clone();
     let tag = canonicalize_tag(&q.tag);
+    let actor_prefix = match q.actor.as_deref().map(str::trim) {
+        None | Some("") => String::new(),
+        Some(s) => match parse_username(s) {
+            Ok(u) => u,
+            Err(msg) => return api_error(StatusCode::BAD_REQUEST, "invalid actor filter", Some(msg)).into_response(),
+        },
+    };
     let reduced = reduced_arc.read().await;
 
     // Single post lookup by ingest ID -- return full body untruncated.
@@ -72,7 +80,7 @@ pub async fn get_thread(State(state): State<AppState>, Query(q): Query<ThreadDet
                     id: ing.id.clone(),
                     index: idx,
                     ts: ing.ts,
-                    actor: format!("@{}", ing.principal),
+                    actor: ing.principal.clone(),
                     body: ing.raw.clone(),
                     truncated: false,
                 }],
@@ -92,7 +100,6 @@ pub async fn get_thread(State(state): State<AppState>, Query(q): Query<ThreadDet
         .map(|q| q.iter().rev().cloned().collect())
         .unwrap_or_default();
 
-    let actor_prefix = q.actor.as_deref().unwrap_or("").to_lowercase();
     let filtered: Vec<(usize, _)> = all_ids
         .into_iter()
         .enumerate()
@@ -119,7 +126,7 @@ pub async fn get_thread(State(state): State<AppState>, Query(q): Query<ThreadDet
                 id: ing.id.clone(),
                 index: idx,
                 ts: ing.ts,
-                actor: format!("@{}", ing.principal),
+                actor: ing.principal.clone(),
                 body,
                 truncated,
             }
diff --git a/server/src/api/garden.rs b/server/src/api/garden.rs
index 137f43f26a1983224fd0fa66ae71301626fb9762..ca0e9e735b8e02e31075ffb77d99941129a045f0 100644
--- a/server/src/api/garden.rs
+++ b/server/src/api/garden.rs
@@ -9,7 +9,7 @@ use slug_types::*;
 use std::collections::HashSet;
 
 use crate::{
-    events::canonicalize_item,
+    canonical_path::canonicalize_item,
     path_types::CanonicalItemUrl,
     state::AppState,
 };
@@ -160,7 +160,7 @@ pub async fn get_recent_votes(
             a: v.a.as_str().to_string(),
             b: v.b.as_str().to_string(),
             ratio: format!("{}:{}", v.ratio_left, v.ratio_right),
-            actor: Some(format!("@{}", v.principal)),
+            actor: Some(v.principal.clone()),
             body: v.body.clone(),
             thread: Some(v.thread_id.clone()),
         }).collect();
@@ -207,7 +207,7 @@ pub async fn get_matchup(
                     a: v.a.as_str().to_string(),
                     b: v.b.as_str().to_string(),
                     ratio: format!("{}:{}", v.ratio_left, v.ratio_right),
-                    actor: Some(format!("@{}", v.principal)),
+                    actor: Some(v.principal.clone()),
                     body: v.body.clone(),
                     thread: Some(v.thread_id.clone()),
                 })
diff --git a/server/src/api/helpers.rs b/server/src/api/helpers.rs
index 41b496fcd79948a0e86c286704ed4c062b338e6c..a16c3df6f9d8eaa2e587f5575b78660880062f81 100644
--- a/server/src/api/helpers.rs
+++ b/server/src/api/helpers.rs
@@ -8,7 +8,7 @@ use slug_types::*;
 use std::collections::HashMap;
 
 use crate::{
-    events::canonicalize_item,
+    canonical_path::canonicalize_item,
     path_types::CanonicalItemUrl,
     ranking::connected_components_from_voted_pairs,
     reducer::ReducerState,
diff --git a/server/src/api/ingest.rs b/server/src/api/ingest.rs
index 76ad8efd802b67f3fdf0f3d306830a68736cb384..5a58754bae8ed5f7ae9ce89e659a61de81c1bc01 100644
--- a/server/src/api/ingest.rs
+++ b/server/src/api/ingest.rs
@@ -9,11 +9,10 @@ use slug_types::*;
 use std::collections::{HashMap, HashSet};
 
 use crate::{
+    canonical_path::canonicalize_tag,
     dsl,
-    events::{
-        canonicalize_agent, canonicalize_tag, canonicalize_username, validate_agent_format,
-        AgentBound, Event, Ingest, ThreadCapability, ThreadVisibility,
-    },
+    events::{AgentBound, Event, Ingest, ThreadCapability, ThreadVisibility},
+    identity::parse_agent,
     path_types::CanonicalItemUrl,
     reducer::{ReducerState, ScopeId},
     state::AppState,
@@ -238,11 +237,18 @@ pub async fn post_ingest(
     let event_log = state.event_log.clone();
     let reduced = reduced_arc.read().await;
 
-    if let Err(msg) = validate_agent_format(&req.delegate) {
-        drop(reduced);
-        return api_error(StatusCode::BAD_REQUEST, "invalid delegate format", Some(msg)).into_response();
-    }
-    let delegate = canonicalize_agent(&req.delegate);
+    let delegate: Option<String> = match &req.delegate {
+        None => None,
+        Some(s) if s.trim().is_empty() => None,
+        Some(s) => match parse_agent(s) {
+            Ok(d) => Some(d),
+            Err(msg) => {
+                drop(reduced);
+                return api_error(StatusCode::BAD_REQUEST, "invalid delegate format", Some(msg))
+                    .into_response();
+            }
+        },
+    };
     let thread_id = canonicalize_tag(&req.thread);
 
     let principal = match verify_bearer_principal(&headers, &reduced) {
@@ -290,36 +296,43 @@ pub async fn post_ingest(
         }
     }
 
-    match reduced.agent_bindings.get(&delegate) {
-        Some(u) if u != &principal => {
-            drop(reduced);
-            return api_error(
-                StatusCode::FORBIDDEN,
-                "delegate already bound to another user",
-                None,
-            )
-            .into_response();
+    if let Some(ref d) = delegate {
+        match reduced.agent_bindings.get(d) {
+            Some(u) if u != &principal => {
+                drop(reduced);
+                return api_error(
+                    StatusCode::FORBIDDEN,
+                    "delegate already bound to another user",
+                    None,
+                )
+                .into_response();
+            }
+            _ => {}
         }
-        _ => {}
     }
-    let need_agent_bind = reduced.agent_bindings.get(&delegate).is_none();
+    let need_agent_bind = delegate
+        .as_ref()
+        .map(|d| reduced.agent_bindings.get(d).is_none())
+        .unwrap_or(false);
     drop(reduced);
 
     let mut events_appended: usize = 0;
 
     if need_agent_bind {
-        let ab = Event::AgentBound(AgentBound {
-            ts: now_ms(),
-            agent: delegate.clone(),
-            username: principal.clone(),
-        });
-        if let Err(err) = event_log.append(&ab).await {
-            return api_error(StatusCode::INTERNAL_SERVER_ERROR, format!("{err}"), None);
-        }
-        events_appended += 1;
-        {
-            let mut reduced = reduced_arc.write().await;
-            reduced.apply_event(ab);
+        if let Some(agent_id) = delegate.clone() {
+            let ab = Event::AgentBound(AgentBound {
+                ts: now_ms(),
+                agent: agent_id,
+                username: principal.clone(),
+            });
+            if let Err(err) = event_log.append(&ab).await {
+                return api_error(StatusCode::INTERNAL_SERVER_ERROR, format!("{err}"), None);
+            }
+            events_appended += 1;
+            {
+                let mut reduced = reduced_arc.write().await;
+                reduced.apply_event(ab);
+            }
         }
     }
 
@@ -413,12 +426,18 @@ pub async fn post_check(
     };
     drop(reduced);
 
-    if let Err(msg) = validate_agent_format(&req.delegate) {
-        return api_error(StatusCode::BAD_REQUEST, "invalid delegate format", Some(msg)).into_response();
-    }
-    let delegate = canonicalize_agent(&req.delegate);
+    let delegate: Option<String> = match &req.delegate {
+        None => None,
+        Some(s) if s.trim().is_empty() => None,
+        Some(s) => match parse_agent(s) {
+            Ok(d) => Some(d),
+            Err(msg) => {
+                return api_error(StatusCode::BAD_REQUEST, "invalid delegate format", Some(msg)).into_response();
+            }
+        },
+    };
     let thread_id = canonicalize_tag(&req.thread);
-    let principal = canonicalize_username("placeholder");
+    let principal = "placeholder".to_string();
 
     let event = Event::Ingest(Ingest {
         ts: v.ts,
diff --git a/server/src/api/mod.rs b/server/src/api/mod.rs
index f1d5c139b875c268cc46c3085c332d8de31b092e..f42e907775645166c60aeae2d98c5855223a71be 100644
--- a/server/src/api/mod.rs
+++ b/server/src/api/mod.rs
@@ -65,7 +65,7 @@ mod tests {
             id: format!("test-{ts}"),
             raw: raw.to_string(),
             principal: "test".to_string(),
-            delegate: "@00000000-0000-0000-0000-000000000000:test:local/test".to_string(),
+            delegate: Some("00000000-0000-0000-0000-000000000000:test:local/test".to_string()),
             thread_id: "t".to_string(),
         }));
     }
diff --git a/server/src/api/rank.rs b/server/src/api/rank.rs
index 059095bb512096da5e9699ff9c4b92cb7de7fa1c..f2d348280ec4d70a32044601964cb5dc9382d9d7 100644
--- a/server/src/api/rank.rs
+++ b/server/src/api/rank.rs
@@ -226,7 +226,7 @@ pub async fn get_rank_history(
     let reduced = reduced_arc.read().await;
     let content = reduced.public();
 
-    let item_str = crate::events::canonicalize_item(&q.item);
+    let item_str = crate::canonical_path::canonicalize_item(&q.item);
     let item = CanonicalItemUrl(item_str.clone());
 
     let entries = content.rank_history.get(&item).cloned().unwrap_or_default();
@@ -238,15 +238,15 @@ pub async fn get_rank_history(
             .map(|doc| {
                 doc.statements.into_iter().filter_map(|s| {
                     if let crate::dsl::Stmt::Vote { item1, item2, ratio_left, ratio_right, explanation } = s {
-                        let a = crate::events::canonicalize_item(&item1);
-                        let b = crate::events::canonicalize_item(&item2);
+                        let a = crate::canonical_path::canonicalize_item(&item1);
+                        let b = crate::canonical_path::canonicalize_item(&item2);
                         if a == item_str || b == item_str {
                             Some(VoteRow {
                                 ts: e.ts,
                                 a: item_path_for_api(&a),
                                 b: item_path_for_api(&b),
                                 ratio: format!("{}:{}", ratio_left, ratio_right),
-                                actor: reduced.ingests_by_id.get(&e.post_id).map(|ing| format!("@{}", ing.principal)),
+                                actor: reduced.ingests_by_id.get(&e.post_id).map(|ing| ing.principal.clone()),
                                 body: explanation,
                                 thread: Some(format!("#{}", e.thread)),
                             })
diff --git a/server/src/api/search.rs b/server/src/api/search.rs
index 8028c3117b9b22460fb9677ca969741f6849d2d0..87621e6e4210a28c7802bb941200788f28518940 100644
--- a/server/src/api/search.rs
+++ b/server/src/api/search.rs
@@ -119,7 +119,7 @@ pub async fn get_search(
                 .unwrap_or_else(|| "#unknown".to_string());
             scored_posts.push((score, ingest.ts, slug_types::SearchPostHit {
                 thread,
-                actor: format!("@{}", ingest.principal),
+                actor: ingest.principal.clone(),
                 snippet: snippet_around(&ingest.raw, &words, 160),
                 ts: ingest.ts,
             }));
diff --git a/server/src/api/thread.rs b/server/src/api/thread.rs
index 576b0b5be2767b0c6cba5e0701e4ffc9f215029d..be6ef446e316fef352a7eef5b35b40583dca4442 100644
--- a/server/src/api/thread.rs
+++ b/server/src/api/thread.rs
@@ -8,9 +8,8 @@ use serde::{Deserialize, Serialize};
 
 use crate::{
     api::helpers::{api_error, now_ms},
-    events::{
-        canonicalize_username, Event, GrantAdded, ThreadCapability, ThreadCreated, ThreadVisibility,
-    },
+    events::{Event, GrantAdded, ThreadCapability, ThreadCreated, ThreadVisibility},
+    identity::parse_username,
     state::AppState,
 };
 use super::auth::verify_bearer_principal;
@@ -150,10 +149,10 @@ pub async fn post_thread_grants(
         return api_error(StatusCode::FORBIDDEN, "requires Manage capability", None).into_response();
     }
 
-    let target = canonicalize_username(&req.username);
-    if target.is_empty() {
-        return api_error(StatusCode::BAD_REQUEST, "invalid username", None).into_response();
-    }
+    let target = match parse_username(&req.username) {
+        Ok(u) => u,
+        Err(msg) => return api_error(StatusCode::BAD_REQUEST, "invalid username", Some(msg)).into_response(),
+    };
     if !reduced.users_by_provider.values().any(|u| u == &target) {
         return api_error(StatusCode::NOT_FOUND, format!("user @{target} not found"), None).into_response();
     }
diff --git a/server/src/canonical_path.rs b/server/src/canonical_path.rs
new file mode 100644
index 0000000000000000000000000000000000000000..5c0febe883d8b3978a25896ed0d71df8509a8717
--- /dev/null
+++ b/server/src/canonical_path.rs
@@ -0,0 +1,92 @@
+//! Normalization for thread tags and ontology item URLs (DSL ↔ stored canonical form).
+//! Not event types — see `events` and `path_types`.
+
+/// Thread / public tag: stored without leading `#`, lowercase.
+pub fn canonicalize_tag(input: &str) -> String {
+    input.trim().trim_start_matches('#').to_lowercase()
+}
+
+/// 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("https://") {
+        let (host, tail) = rest.split_once('/').map_or((rest, ""), |(h, t)| (h, t));
+        let host = host.trim().to_lowercase();
+        if tail.is_empty() {
+            return format!("https://{}", host);
+        } else {
+            return 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();
+        if tail.is_empty() {
+            return format!("http://{}", host);
+        } else {
+            return 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 {
+        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("/"))
+}
diff --git a/server/src/events.rs b/server/src/events.rs
index 7775e59974a521f5fea9361a600802898d05c6ab..26edac7505ce4e67a042a5511113efba5fea4950 100644
--- a/server/src/events.rs
+++ b/server/src/events.rs
@@ -1,150 +1,5 @@
 use serde::{Deserialize, Serialize};
 
-/// Canonical identifiers stored without sigils.
-/// - tags are stored without leading '#'
-/// - items are stored without leading '/'
-pub fn canonicalize_tag(input: &str) -> String {
-    input.trim().trim_start_matches('#').to_lowercase()
-}
-
-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("https://") {
-        let (host, tail) = rest.split_once('/').map_or((rest, ""), |(h, t)| (h, t));
-        let host = host.trim().to_lowercase();
-        if tail.is_empty() {
-            return format!("https://{}", host);
-        } else {
-            return 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();
-        if tail.is_empty() {
-            return format!("http://{}", host);
-        } else {
-            return 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 {
-        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;
-    }
-
-    // Should be unreachable since all canonical items are now URLs
-    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("/"))
-}
-
-/// Canonical username stored without leading '@'.
-pub fn canonicalize_username(input: &str) -> String {
-    input.trim().trim_start_matches('@').to_lowercase()
-}
-
-/// Validate username: lowercase alphanumeric + '-' '_' only; length 1-32.
-pub fn validate_username(username: &str) -> Result<(), String> {
-    let u = canonicalize_username(username);
-    if u.is_empty() || u.len() > 32 {
-        return Err("username must be 1-32 characters".to_string());
-    }
-    if !u
-        .chars()
-        .all(|c| c.is_ascii_lowercase() || c.is_ascii_digit() || c == '-' || c == '_')
-    {
-        return Err("username must be lowercase alphanumeric with '-' or '_' only".to_string());
-    }
-    Ok(())
-}
-
-/// Canonical agent identity stored with single leading '@'.
-///
-/// Request/display form is `@@uuid:rig:provider/model` but we store `@uuid:rig:provider/model`.
-pub fn canonicalize_agent(input: &str) -> String {
-    let s = input.trim();
-    let s = s.strip_prefix("@@").or_else(|| s.strip_prefix('@')).unwrap_or(s);
-    format!("@{}", s.to_lowercase())
-}
-
-/// Validate agent format: @@<uuid-v4>:<rig>:<provider/model>
-pub fn validate_agent_format(agent: &str) -> Result<(), String> {
-    let a = agent.trim();
-    if !a.starts_with("@@") && !a.starts_with('@') {
-        return Err("agent must start with @@".to_string());
-    }
-    let a = a.strip_prefix("@@").or_else(|| a.strip_prefix('@')).unwrap_or(a);
-    let parts: Vec<&str> = a.split(':').collect();
-    if parts.len() != 3 {
-        return Err("agent must be @@<uuid>:<rig>:<provider/model>".to_string());
-    }
-    let (uuid_part, rig_part, model_part) = (parts[0], parts[1], parts[2]);
-    if uuid::Uuid::parse_str(uuid_part).is_err() {
-        return Err("agent uuid must be a valid UUID v4".to_string());
-    }
-    if rig_part.trim().is_empty() {
-        return Err("agent rig must be non-empty".to_string());
-    }
-    if !model_part.contains('/') {
-        return Err("agent model must be <provider/model>".to_string());
-    }
-    Ok(())
-}
-
 #[derive(Debug, Clone, Copy, Serialize, Deserialize, PartialEq, Eq)]
 #[serde(rename_all = "snake_case")]
 pub enum ThreadVisibility {
@@ -236,10 +91,11 @@ pub struct Ingest {
     pub id: String,
     /// Raw DSL+prose body only (no identity/routing metadata).
     pub raw: String,
-    /// Human principal username (no leading '@').
+    /// Human principal username (wire and storage: no `@`).
     pub principal: String,
-    /// Delegate agent identity (canonical stored with single leading '@').
-    pub delegate: String,
+    /// AI delegate id `uuid:rig:model` (wire and storage: no `@`). Omitted when absent.
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub delegate: Option<String>,
     /// Thread identifier: public tag (e.g. "languages") or private id/slug (e.g. "a7f2k9x/project-review").
     pub thread_id: String,
 }
@@ -247,5 +103,3 @@ pub struct Ingest {
 fn generate_id() -> String {
     uuid::Uuid::new_v4().to_string()
 }
-
-
diff --git a/server/src/html/editor.rs b/server/src/html/editor.rs
index 1d2c4be7eef4cc437cacddd0949307895f3ef018..4f206a0a895b47592df7669cd8d2fb4d83286fe7 100644
--- a/server/src/html/editor.rs
+++ b/server/src/html/editor.rs
@@ -33,7 +33,7 @@ pub async fn editor_page() -> impl IntoResponse {
             p class="muted" { "write DSL, see what happens. nothing is saved." }
             div class="editor-container" {
                 textarea id="editor-input" rows="12" cols="80"
-                    placeholder="@your-uuid:rig:provider/model\n#your-thread\n\n~/path/item-a { description }\n~/path/item-b { description }\n\n~/path/item-a 3:1 ~/path/item-b { reasoning }"
+                    placeholder="your-uuid:rig:provider/model\n#your-thread\n\n~/path/item-a { description }\n~/path/item-b { description }\n\n~/path/item-a 3:1 ~/path/item-b { reasoning }"
                     autocomplete="off" autofocus {}
                 div id="editor-status" class="muted" { "type to check…" }
                 div id="editor-results" {}
@@ -110,7 +110,7 @@ pub async fn editor_check(
                 id: uuid::Uuid::new_v4().to_string(),
                 raw: form.text.clone(),
                 principal: String::new(),
-                delegate: String::new(),
+                delegate: None,
                 thread_id: String::new(),
             });
             let mut simulated = { reduced_arc.read().await.clone() };
diff --git a/server/src/html/forum.rs b/server/src/html/forum.rs
index d40cc6398aaf3b2348d91498b5b3a80ed593e55c..a4e17ddf0064b08dfa221c964ccaa0eb99948b5b 100644
--- a/server/src/html/forum.rs
+++ b/server/src/html/forum.rs
@@ -7,14 +7,14 @@ use serde::Deserialize;
 use maud::{html, Markup};
 
 use crate::{
-    events::canonicalize_tag,
+    canonical_path::canonicalize_tag,
     reducer::ReducerState,
     state::AppState,
     timeago,
 };
 
 use super::{
-    actor_label, bc_threads, cli_panel, layout, now_ms,
+    authorship_address, bc_threads, cli_panel, layout, now_ms,
     recency_class, render_linkified_with_embeds,
 };
 
@@ -260,7 +260,7 @@ pub async fn thread_post_view(
                 @let ago = timeago::timeago(now, ing.ts);
                 div class="ingest-entry" data-ingest-id=(ing.id) {
                     div class="ingest-meta muted" title=(hover) {
-                        span class="address" { "@" (actor_label(&ing.delegate)) }
+                        span class="address" { (authorship_address(&ing.principal, &ing.delegate)) }
                         " · "
                         (ago)
                     }
diff --git a/server/src/html/garden.rs b/server/src/html/garden.rs
index bddd90aa6eb288c0c97bb08f472e2934fd2fe67a..dcd7f0d4a2d7b602026b643c564d212cb084ff9f 100644
--- a/server/src/html/garden.rs
+++ b/server/src/html/garden.rs
@@ -5,7 +5,7 @@ use axum::{
 use maud::html;
 
 use crate::{
-    events::canonicalize_item,
+    canonical_path::canonicalize_item,
     path_types::CanonicalItemUrl,
     ranking::{connected_components_from_voted_pairs, ranked_items_subset},
     scope_rank::{build_children_rankings, ChildrenRankings},
@@ -14,7 +14,7 @@ use crate::{
 };
 
 use super::{
-    actor_label, bc_path, cli_panel, layout, now_ms, ratio_pct, render_linkified_with_embeds,
+    authorship_address, bc_path, cli_panel, layout, now_ms, ratio_pct, render_linkified_with_embeds,
     breadcrumb_path::OntologyPath,
 };
 
@@ -204,8 +204,8 @@ fn build_rank_history(
             .map(|doc| {
                 doc.statements.into_iter().filter_map(|s| {
                     if let crate::dsl::Stmt::Vote { item1, item2, ratio_left, ratio_right, explanation } = s {
-                        let a_str = crate::events::canonicalize_item(&item1);
-                        let b_str = crate::events::canonicalize_item(&item2);
+                        let a_str = crate::canonical_path::canonicalize_item(&item1);
+                        let b_str = crate::canonical_path::canonicalize_item(&item2);
                         if a_str == item || b_str == item {
                             Some(crate::reducer::VoteData {
                                 ts: e.ts,
@@ -216,9 +216,7 @@ fn build_rank_history(
                                 principal: reduced.ingests_by_id.get(&e.post_id)
                                     .map(|ing| ing.principal.clone())
                                     .unwrap_or_default(),
-                                delegate: reduced.ingests_by_id.get(&e.post_id)
-                                    .map(|ing| ing.delegate.clone())
-                                    .unwrap_or_default(),
+                                delegate: reduced.ingests_by_id.get(&e.post_id).and_then(|ing| ing.delegate.clone()),
                                 thread_id: e.thread.clone(),
                             })
                         } else { None }
@@ -331,7 +329,7 @@ async fn render_scope_view(state: AppState, path: OntologyPath) -> axum::respons
                             @let right_class = if v.b.as_str() == model.item { "ratio-right current" } else { "ratio-right" };
                             div class="ont-vote-entry" {
                                 div class="ont-vote-meta" title=(hover) {
-                                    span class="address" { "@" (actor_label(&v.delegate)) }
+                                    span class="address" { (authorship_address(&v.principal, &v.delegate)) }
                                     " · "
                                     (ago)
                                 }
@@ -482,7 +480,7 @@ mod tests {
             id: format!("ing-{ts}"),
             raw: raw.to_string(),
             principal: "testuser".to_string(),
-            delegate: "@00000000-0000-0000-0000-000000000000:test:local/test".to_string(),
+            delegate: Some("00000000-0000-0000-0000-000000000000:test:local/test".to_string()),
             thread_id: String::new(),
         }));
     }
diff --git a/server/src/html/mod.rs b/server/src/html/mod.rs
index 8b48a25cce79f0eefc7e29849e1a667cae4c7986..82c5993fabfeaf2ea8b9034bccaef37924100264 100644
--- a/server/src/html/mod.rs
+++ b/server/src/html/mod.rs
@@ -238,9 +238,9 @@ pub(super) fn bc_threads(thread_tag: Option<&str>) -> Markup {
     }
 }
 
-/// Input is canonicalized without leading '@' (usually uuid:rig:provider/model).
-pub(super) fn actor_label(actor: &str) -> String {
-    let a = actor.trim_start_matches('@').trim();
+/// Short display label for a stored agent id (`uuid:rig:model`, no `@`).
+pub(super) fn actor_label(agent_naked: &str) -> String {
+    let a = agent_naked.trim();
     let parts: Vec<&str> = a.split(':').collect();
     if parts.len() >= 3 {
         let rig = parts[1].trim();
@@ -258,6 +258,13 @@ pub(super) fn actor_label(actor: &str) -> String {
     a.to_string()
 }
 
+/// HTML attribution only: human `@name`, or agent `@@uuid8:rig:model` when a delegate is present.
+pub(super) fn authorship_address(principal: &str, delegate: &Option<String>) -> String {
+    match delegate {
+        Some(d) => format!("@@{}", actor_label(d)),
+        None => format!("@{}", principal),
+    }
+}
 
 /// Escape HTML special chars for safe injection.
 fn escape_html(s: &str) -> String {
diff --git a/server/src/html/search.rs b/server/src/html/search.rs
index f56f31546b4870d41e594d0e3ac2a4b50ba831b8..df5cb0a59ff72956f4bc94e53b86ab8337452817 100644
--- a/server/src/html/search.rs
+++ b/server/src/html/search.rs
@@ -11,7 +11,7 @@ use crate::{
     timeago,
 };
 
-use super::{actor_label, bc_segment, cli_panel, layout, now_ms};
+use super::{authorship_address, bc_segment, cli_panel, layout, now_ms};
 
 /// Escape HTML special chars for safe injection.
 fn escape_html(s: &str) -> String {
@@ -44,7 +44,8 @@ struct ThreadRow {
 
 struct PostRow {
     thread: String,
-    actor: String,
+    /// Pre-formatted attribution string for display (includes `@` / `@@` from `authorship_address`).
+    actor_display: String,
     text: String,
     ts: i64,
 }
@@ -151,7 +152,7 @@ fn search(state: &ReducerState, q: &str, limit: usize) -> SearchResults {
                 .unwrap_or_else(|| "unknown".to_string());
             scored_posts.push((score, PostRow {
                 thread,
-                actor: ingest.principal.clone(),
+                actor_display: authorship_address(&ingest.principal, &ingest.delegate),
                 text: ingest.raw.clone(),
                 ts: ingest.ts,
             }));
@@ -321,7 +322,7 @@ fn render_search_results(results: &SearchResults, query: &str) -> Markup {
                             li {
                                 div class="search-post-meta muted" {
                                     a href=(format!("/t/{}", r.thread)) { "#" (r.thread) }
-                                    " · " (actor_label(&r.actor))
+                                    " · " (r.actor_display)
                                     " · " (timeago::timeago(now, r.ts))
                                 }
                                 div class="search-snippet" {
diff --git a/server/src/html/tree.rs b/server/src/html/tree.rs
index 568a08edbb548291ab6f03b46c79cdcb4b432468..339f69af3b3279918a97b1dced62e4e3e0b528e3 100644
--- a/server/src/html/tree.rs
+++ b/server/src/html/tree.rs
@@ -13,7 +13,7 @@ use serde::{Deserialize, Serialize};
 use std::collections::BTreeSet;
 
 use crate::{
-    events::canonicalize_item,
+    canonical_path::canonicalize_item,
     path_types::{CanonicalItemUrl, RelativePath},
     scope_rank::ChildrenRankings,
     state::AppState,
@@ -702,7 +702,8 @@ pub async fn tree_select(
 mod tests {
     use super::*;
 
-    use crate::events::{canonicalize_item, Event, Ingest};
+    use crate::canonical_path::canonicalize_item;
+    use crate::events::{Event, Ingest};
     use crate::reducer::ReducerState;
 
     fn ingest(raw: &str) -> Event {
@@ -711,7 +712,7 @@ mod tests {
             id: "test-ingest".to_string(),
             raw: raw.to_string(),
             principal: "tester".to_string(),
-            delegate: String::new(),
+            delegate: None,
             thread_id: String::new(),
         })
     }
@@ -789,9 +790,9 @@ mod tests {
     fn reducer_parent_key_for_tilde_items_is_without_trailing_slash() {
         // This is the reducer invariant that the tree view must match.
         let item = canonicalize_item("~/alphabet/a");
-        assert_eq!(crate::events::item_parent_path(&item).unwrap(), "https://slug.social/~/alphabet");
+        assert_eq!(crate::canonical_path::item_parent_path(&item).unwrap(), "https://slug.social/~/alphabet");
         let item2 = canonicalize_item("~/a");
-        assert_eq!(crate::events::item_parent_path(&item2).unwrap(), "https://slug.social/~");
+        assert_eq!(crate::canonical_path::item_parent_path(&item2).unwrap(), "https://slug.social/~");
     }
 
     #[test]
diff --git a/server/src/identity.rs b/server/src/identity.rs
new file mode 100644
index 0000000000000000000000000000000000000000..ad654ae813f8f6fe6323746e2544250687d3d47f
--- /dev/null
+++ b/server/src/identity.rs
@@ -0,0 +1,66 @@
+//! Usernames and agent delegate ids. Wire JSON and query params use **stored form only** (no `@` / `@@`).
+//! The HTTP layer validates here; the reducer does not rewrite identity. For humans, `@name` / `@@agent`
+//! appear only in HTML (see `html::authorship_address`).
+
+/// Parse username from query/body: trim, lowercase. `@` is not allowed (use `tommy`, not `@tommy`).
+pub fn parse_username(input: &str) -> Result<String, String> {
+    let s = input.trim();
+    if s.is_empty() {
+        return Err("username must not be empty".to_string());
+    }
+    if s.contains('@') {
+        return Err(
+            "username must not contain '@' — use stored form (e.g. `tommy`)".to_string(),
+        );
+    }
+    let u = s.to_lowercase();
+    validate_username_naked(&u)?;
+    Ok(u)
+}
+
+fn validate_username_naked(u: &str) -> Result<(), String> {
+    if u.len() > 32 {
+        return Err("username must be 1-32 characters".to_string());
+    }
+    if !u
+        .chars()
+        .all(|c| c.is_ascii_lowercase() || c.is_ascii_digit() || c == '-' || c == '_')
+    {
+        return Err("username must be lowercase alphanumeric with '-' or '_' only".to_string());
+    }
+    Ok(())
+}
+
+/// Parse agent id: trim, lowercase. Must be `uuid:rig:provider/model` with no `@`.
+pub fn parse_agent(input: &str) -> Result<String, String> {
+    let s = input.trim();
+    if s.is_empty() {
+        return Err("agent id must not be empty".to_string());
+    }
+    if s.contains('@') {
+        return Err(
+            "agent id must not contain '@' — use `uuid:rig:provider/model`".to_string(),
+        );
+    }
+    let a = s.to_lowercase();
+    validate_agent_naked(&a)?;
+    Ok(a)
+}
+
+fn validate_agent_naked(a: &str) -> Result<(), String> {
+    let parts: Vec<&str> = a.split(':').collect();
+    if parts.len() != 3 {
+        return Err("agent must be <uuid>:<rig>:<provider/model>".to_string());
+    }
+    let (uuid_part, rig_part, model_part) = (parts[0], parts[1], parts[2]);
+    if uuid::Uuid::parse_str(uuid_part).is_err() {
+        return Err("agent uuid must be a valid UUID v4".to_string());
+    }
+    if rig_part.trim().is_empty() {
+        return Err("agent rig must be non-empty".to_string());
+    }
+    if !model_part.contains('/') {
+        return Err("agent model must be <provider/model>".to_string());
+    }
+    Ok(())
+}
diff --git a/server/src/lib.rs b/server/src/lib.rs
index 1820b0dd9cd2d0b6ce05789e3225e0d86a3d357a..173fbdd6b3f23c062a200b268d04491d5e030b3f 100644
--- a/server/src/lib.rs
+++ b/server/src/lib.rs
@@ -1,10 +1,12 @@
 #[macro_use]
 pub mod paths;
 pub mod api;
+pub mod canonical_path;
 pub mod dsl;
 pub mod html;
 pub mod event_log;
 pub mod events;
+pub mod identity;
 pub mod middleware;
 pub mod path_types;
 pub mod ranking;
diff --git a/server/src/path_types.rs b/server/src/path_types.rs
index 5a903de4f8309b6c3874e2cd70f2eb5650fd50a8..997597f6659c1ba2092bc5348e393bf7a8fa7ae0 100644
--- a/server/src/path_types.rs
+++ b/server/src/path_types.rs
@@ -14,9 +14,9 @@ use std::fmt;
 
 use serde::{Deserialize, Serialize};
 
-use crate::events::canonicalize_item;
+use crate::canonical_path::canonicalize_item;
 
-/// Canonical item identifier as produced by `events::canonicalize_item`.
+/// Canonical item identifier as produced by `canonical_path::canonicalize_item`.
 ///
 /// In practice this is usually:
 /// - `https://slug.social/~/...` for ontology items, or
diff --git a/server/src/ranking.rs b/server/src/ranking.rs
index 70119611473a430113429eafb6d92b2bf96a3eb5..f70da0d076da2ba8f0abb5a4415babae1c2305f9 100644
--- a/server/src/ranking.rs
+++ b/server/src/ranking.rs
@@ -265,7 +265,7 @@ mod tests {
             ratio_right: r,
             body: "because".to_string(),
             principal: "test".to_string(),
-            delegate: "@00000000-0000-0000-0000-000000000000:test:local/test".to_string(),
+            delegate: Some("00000000-0000-0000-0000-000000000000:test:local/test".to_string()),
             thread_id: "untagged".to_string(),
         }
     }
diff --git a/server/src/reducer.rs b/server/src/reducer.rs
index 9ab063a340e228edaa8967ec35762809fccd5d3a..550bcda1a1068df13908d19484d86ad46d65ddf1 100644
--- a/server/src/reducer.rs
+++ b/server/src/reducer.rs
@@ -2,7 +2,8 @@ use std::collections::{HashMap, HashSet, VecDeque};
 
 use serde::{Deserialize, Serialize};
 
-use crate::events::{canonicalize_agent, canonicalize_tag, canonicalize_username, Event, Ingest, ThreadCapability};
+use crate::canonical_path::canonicalize_tag;
+use crate::events::{Event, Ingest, ThreadCapability};
 use crate::path_types::CanonicalItemUrl;
 
 #[derive(Debug, Clone, Hash, PartialEq, Eq, PartialOrd, Ord)]
@@ -21,10 +22,10 @@ pub struct VoteData {
     pub ratio_left: i32,
     pub ratio_right: i32,
     pub body: String,
-    /// Human principal username (no leading '@').
+    /// Human principal username (no `@` in stored events).
     pub principal: String,
-    /// Agent delegate identity (stored with single leading '@').
-    pub delegate: String,
+    /// AI delegate id, if any (no `@` in stored events).
+    pub delegate: Option<String>,
     /// Thread id where this vote was cast (public tag or private id/slug).
     pub thread_id: String,
 }
@@ -99,8 +100,6 @@ 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.principal = canonicalize_username(&vote.principal);
-        vote.delegate = canonicalize_agent(&vote.delegate);
         vote.thread_id = canonicalize_tag(&vote.thread_id);
         if vote.ratio_left < 0 {
             vote.ratio_left = 0;
@@ -189,7 +188,7 @@ pub struct ReducerState {
     pub users_by_provider: HashMap<(String, String), String>,
     /// token_id -> (username, salt, token_hash)
     pub tokens_by_id: HashMap<String, (String, String, String)>,
-    /// agent delegate (canonical '@...') -> username
+    /// agent id (naked `uuid:rig:model`) -> username
     pub agent_bindings: HashMap<String, String>,
 
     pub ingests_by_id: HashMap<String, Ingest>,
@@ -330,26 +329,27 @@ impl ReducerState {
     pub fn apply_event(&mut self, event: Event) {
         match event {
             Event::UserRegistered(ur) => {
-                let username = canonicalize_username(&ur.username);
-                self.users_by_provider
-                    .insert((ur.provider.to_lowercase(), ur.provider_id.clone()), username);
+                self.users_by_provider.insert(
+                    (ur.provider.to_lowercase(), ur.provider_id.clone()),
+                    ur.username,
+                );
             }
             Event::TokenIssued(ti) => {
-                let username = canonicalize_username(&ti.username);
-                self.tokens_by_id
-                    .insert(ti.token_id.clone(), (username, ti.salt.clone(), ti.token_hash.clone()));
+                self.tokens_by_id.insert(
+                    ti.token_id.clone(),
+                    (ti.username, ti.salt.clone(), ti.token_hash.clone()),
+                );
             }
             Event::AgentBound(ab) => {
-                let username = canonicalize_username(&ab.username);
-                let agent = canonicalize_agent(&ab.agent);
-                self.agent_bindings.insert(agent, username);
+                if ab.agent.is_empty() {
+                    return;
+                }
+                self.agent_bindings.insert(ab.agent, ab.username);
             }
             Event::ThreadCreated(tc) => {
                 self.threads.entry(tc.thread_id.clone()).or_default().visibility = tc.visibility;
             }
             Event::Ingest(mut ing) => {
-                ing.principal = canonicalize_username(&ing.principal);
-                ing.delegate = canonicalize_agent(&ing.delegate);
                 ing.thread_id = canonicalize_tag(&ing.thread_id);
 
                 self.ingests_by_id.insert(ing.id.clone(), ing.clone());
@@ -528,7 +528,7 @@ impl ReducerState {
                 let caps = self.grants
                     .entry(ga.thread_id)
                     .or_default()
-                    .entry(canonicalize_username(&ga.username))
+                    .entry(ga.username)
                     .or_default();
                 for cap in ga.capabilities {
                     caps.insert(cap);
@@ -536,7 +536,7 @@ impl ReducerState {
             }
             Event::GrantRevoked(gr) => {
                 if let Some(thread_grants) = self.grants.get_mut(&gr.thread_id) {
-                    let username = canonicalize_username(&gr.username);
+                    let username = gr.username;
                     if let Some(caps) = thread_grants.get_mut(&username) {
                         for cap in &gr.capabilities {
                             caps.remove(cap);
diff --git a/server/tests/basic.rs b/server/tests/basic.rs
index 8d7dd87d327c969ad953ecaa6b021a03e4d1842e..21cacd3049c9f5a307e75eefd5ebfc7bf0097b5d 100644
--- a/server/tests/basic.rs
+++ b/server/tests/basic.rs
@@ -1,6 +1,7 @@
 use slugsocial_server::{
     event_log::EventLog,
-    events::{canonicalize_item, canonicalize_tag, Event, Ingest},
+    canonical_path::{canonicalize_item, canonicalize_tag},
+    events::{Event, Ingest},
     ranking::ranked_items,
     reducer::{GroupState, ReducerState},
 };
@@ -15,7 +16,7 @@ fn ingest_event(ts: i64, raw: &str) -> Event {
         id: format!("test-{ts}"),
         raw: raw.to_string(),
         principal: "test".to_string(),
-        delegate: "@@00000000-0000-0000-0000-000000000000:test:local/test".to_string(),
+        delegate: Some("00000000-0000-0000-0000-000000000000:test:local/test".to_string()),
         thread_id: "t".to_string(),
     })
 }
@@ -491,7 +492,7 @@ fn reducer_negative_ratio_clamped_to_zero() {
         ratio_right: -3,
         body: "negative".to_string(),
         principal: "test".to_string(),
-        delegate: "@@00000000-0000-0000-0000-000000000000:test:local/test".to_string(),
+        delegate: Some("00000000-0000-0000-0000-000000000000:test:local/test".to_string()),
         thread_id: "t".to_string(),
     });
     assert_eq!(group.idx_to_item.len(), 2);
diff --git a/server/tests/integration.rs b/server/tests/integration.rs
index 24ef464631cd269889b8bb751ef0e90289443e97..eaa25f2c5cc21df3096d9b8f54c060c83207bfd6 100644
--- a/server/tests/integration.rs
+++ b/server/tests/integration.rs
@@ -87,7 +87,7 @@ async fn test_ingest_actor_with_colons_is_detected_and_validated() {
     // Old archive style: agent includes colons but UUID is only a prefix (invalid).
     // We should detect the agent line, then fail with "invalid agent format".
     let ingest_payload = serde_json::json!({
-        "delegate": "@@aec1e31c:claudecode:anthropic/claude-sonnet-4.5",
+        "delegate": "aec1e31c:claudecode:anthropic/claude-sonnet-4.5",
         "thread": "t",
         "text": "~/x {x}\n",
     });
@@ -108,6 +108,26 @@ async fn test_ingest_actor_with_colons_is_detected_and_validated() {
         hint.to_lowercase().contains("uuid"),
         "hint should mention uuid, got: {hint}"
     );
+
+    let at_payload = serde_json::json!({
+        "delegate": "@00000000-0000-0000-0000-000000000000:test:local/test",
+        "thread": "t",
+        "text": "~/x {x}\n",
+    });
+    let at_resp = client
+        .post(&format!("http://{}/api/v0/ingest", addr))
+        .header("Authorization", format!("Bearer {}", test_bearer()))
+        .json(&at_payload)
+        .send()
+        .await
+        .unwrap();
+    assert_eq!(at_resp.status(), reqwest::StatusCode::BAD_REQUEST);
+    let at_body: serde_json::Value = at_resp.json().await.unwrap();
+    let at_hint = at_body["hint"].as_str().unwrap_or_default();
+    assert!(
+        at_hint.contains('@'),
+        "hint should reject '@' in delegate, got: {at_hint}"
+    );
 }
 
 #[tokio::test]
@@ -117,7 +137,7 @@ async fn test_vote_endpoint() {
 
     // /api/v0/vote was removed; all votes are submitted via ingest.
     let ingest_payload = serde_json::json!({
-        "delegate": "@@00000000-0000-0000-0000-000000000000:test:local/test",
+        "delegate": "00000000-0000-0000-0000-000000000000:test:local/test",
         "thread": "cli",
         "text": "~/clap {cli parser}\n~/argh {cli parser}\n~/clap 3:1 ~/argh {because clap is more full-featured}\n",
     });
@@ -143,7 +163,7 @@ async fn test_rank_endpoint() {
 
     // Ingest items + vote (vote endpoint removed).
     let ingest_payload = serde_json::json!({
-        "delegate": "@@00000000-0000-0000-0000-000000000000:test:local/test",
+        "delegate": "00000000-0000-0000-0000-000000000000:test:local/test",
         "thread": "langs",
         "text": "~/rust {systems}\n~/go {concurrency}\n~/rust 3:1 ~/go {because i prefer rust for systems work}\n",
     });
@@ -180,7 +200,7 @@ async fn test_check_endpoint_does_not_commit() {
     let client = reqwest::Client::new();
 
     let check_payload = serde_json::json!({
-        "delegate": "@@00000000-0000-0000-0000-000000000000:test:local/test",
+        "delegate": "00000000-0000-0000-0000-000000000000:test:local/test",
         "thread": "t",
         "text": "~/a {x}\n~/b {y}\n~/a 2:1 ~/b {because}\n",
     });
@@ -220,7 +240,7 @@ async fn test_garden_item_pair_matchup_include_threads() {
 
     // Ingest with thread_id metadata so item_threads and vote.thread_id are populated.
     let ingest_payload = serde_json::json!({
-        "delegate": "@@00000000-0000-0000-0000-000000000000:test:local/test",
+        "delegate": "00000000-0000-0000-0000-000000000000:test:local/test",
         "thread": "sorting-hat",
         "text": "~/sorts/insertion { O(n^2) }\n~/sorts/mergesort { O(n log n) }\n~/sorts/insertion 3:1 ~/sorts/mergesort { simpler for small n }\n",
     });
@@ -324,7 +344,7 @@ async fn test_rank_history() {
     // First ingest: rust vs python — two votes on rust in one doc (the multi-vote case).
     ingest(
         serde_json::json!({
-            "delegate": "@@00000000-0000-0000-0000-000000000001:rig:test/model",
+            "delegate": "00000000-0000-0000-0000-000000000001:rig:test/model",
             "thread": "hist-test",
             "text": "~/hist/rust { systems }\n~/hist/python { scripting }\n~/hist/go { concurrency }\n~/hist/rust 3:1 ~/hist/python { ownership over gc }\n~/hist/rust 2:1 ~/hist/go { performance over simplicity }\n",
         }),
@@ -354,7 +374,7 @@ async fn test_rank_history() {
     // Second ingest: python beats go — rust not directly touched, so python gets a new entry.
     ingest(
         serde_json::json!({
-            "delegate": "@@00000000-0000-0000-0000-000000000002:rig:test/model",
+            "delegate": "00000000-0000-0000-0000-000000000002:rig:test/model",
             "thread": "hist-test",
             "text": "~/hist/python 3:1 ~/hist/go { dynamic typing is worth it }\n",
         }),
@@ -404,7 +424,7 @@ async fn pair_returns_connectivity_stats() {
 
     // Ingest 4 items with 1 vote (a vs b), leaving c and d as isolates.
     let doc = serde_json::json!({
-        "delegate": "@@00000000-0000-0000-0000-000000000001:testrig:test/model",
+        "delegate": "00000000-0000-0000-0000-000000000001:testrig:test/model",
         "thread": "connectivity-test",
         "text": "~/conn/a { item a }\n~/conn/b { item b }\n~/conn/c { item c }\n~/conn/d { item d }\n~/conn/a 3:1 ~/conn/b { a is better }\n",
     });
@@ -438,7 +458,7 @@ async fn pair_returns_connectivity_stats() {
 
     // Add a vote connecting c to a — should reduce components.
     let doc2 = serde_json::json!({
-        "delegate": "@@00000000-0000-0000-0000-000000000001:testrig:test/model",
+        "delegate": "00000000-0000-0000-0000-000000000001:testrig:test/model",
         "thread": "connectivity-test",
         "text": "~/conn/c 2:1 ~/conn/a { c beats a }\n",
     });
diff --git a/test/auth.bb b/test/auth.bb
index b54ea509aafd2d5ee85877665982d7fc00cd8b20..611e04f1806ef81d678a91f499597fe55f691dd7 100644
--- a/test/auth.bb
+++ b/test/auth.bb
@@ -155,7 +155,7 @@
 
      (println "\nstarting pending session…")
      (let [start-resp (http-post-json (str base-url "/api/v0/pending-session")
-                                      {:agent "@@00000000-0000-0000-0000-000000000000:bb:local/dev"})
+                                      {:agent "00000000-0000-0000-0000-000000000000:bb:local/dev"})
            _ (assert! (= 200 (:status start-resp)) "pending-session start returns 200")
            start-json (json/parse-string (:body start-resp) true)]
        (assert! (clojure.string/starts-with? (:session start-json) "p_") "session id has p_ prefix")
diff --git a/test/grants.bb b/test/grants.bb
index 962ff01cdc25d6d6977cc9a810c404918c20856d..f72799ae6fe099f48bdf316deee07f15882e9d45 100644
--- a/test/grants.bb
+++ b/test/grants.bb
@@ -187,11 +187,11 @@
      ;; Register two users. The mock google cycles through google-user-alice then google-user-bob.
      (println "\nregistering alice…")
      (let [alice-token (register-user base-url
-                                      "@@00000000-0000-0000-0000-000000000001:test:local/dev"
+                                      "00000000-0000-0000-0000-000000000001:test:local/dev"
                                       "alice")
            _ (println "registering bob…")
            bob-token   (register-user base-url
-                                      "@@00000000-0000-0000-0000-000000000002:test:local/dev"
+                                      "00000000-0000-0000-0000-000000000002:test:local/dev"
                                       "bob")
 
            ;; Alice creates a private thread.
@@ -206,14 +206,14 @@
        ;; Alice (owner) can post prose to her own private thread.
        (println "\nalice posts prose to her private thread…")
        (assert! (= 200 (:status (ingest! base-url alice-token thread-id
-                                         "@@00000000-0000-0000-0000-000000000001:test:local/dev"
+                                         "00000000-0000-0000-0000-000000000001:test:local/dev"
                                          "Hello from alice.")))
                 "alice prose post succeeds")
 
        ;; Bob has no grants at all — should get 403.
        (println "\nbob (no grants) tries to post prose…")
        (assert! (= 403 (:status (ingest! base-url bob-token thread-id
-                                         "@@00000000-0000-0000-0000-000000000002:test:local/dev"
+                                         "00000000-0000-0000-0000-000000000002:test:local/dev"
                                          "Hello from bob, unauthorized.")))
                 "bob without grants gets 403")
 
@@ -226,7 +226,7 @@
 
        (println "\nbob (View only) tries to post prose…")
        (assert! (= 403 (:status (ingest! base-url bob-token thread-id
-                                         "@@00000000-0000-0000-0000-000000000002:test:local/dev"
+                                         "00000000-0000-0000-0000-000000000002:test:local/dev"
                                          "Hello from bob, view only.")))
                 "bob with View but no Post gets 403")
 
@@ -239,21 +239,21 @@
 
        (println "\nbob (View + Post) posts prose…")
        (assert! (= 200 (:status (ingest! base-url bob-token thread-id
-                                         "@@00000000-0000-0000-0000-000000000002:test:local/dev"
+                                         "00000000-0000-0000-0000-000000000002:test:local/dev"
                                          "Hello from bob, now authorised.")))
                 "bob with View + Post succeeds for prose")
 
        ;; Alice defines two items and votes on them in the private thread.
        (println "\nalice posts items + vote to private thread…")
        (assert! (= 200 (:status (ingest! base-url alice-token thread-id
-                                         "@@00000000-0000-0000-0000-000000000001:test:local/dev"
+                                         "00000000-0000-0000-0000-000000000001:test:local/dev"
                                          "~/fruits/apple { A crisp red apple. }\n~/fruits/banana { A yellow banana. }\n~/fruits/apple > ~/fruits/banana { apples are better }")))
                 "alice vote in private thread succeeds")
 
        ;; Bob (View + Post, no Vote) tries to vote — should be 403.
        (println "\nbob (no Vote) tries to vote…")
        (assert! (= 403 (:status (ingest! base-url bob-token thread-id
-                                         "@@00000000-0000-0000-0000-000000000002:test:local/dev"
+                                         "00000000-0000-0000-0000-000000000002:test:local/dev"
                                          "~/fruits/apple > ~/fruits/banana { bob's take }")))
                 "bob without Vote gets 403")
 
@@ -266,7 +266,7 @@
 
        (println "\nbob (View + Post + Vote) votes…")
        (assert! (= 200 (:status (ingest! base-url bob-token thread-id
-                                         "@@00000000-0000-0000-0000-000000000002:test:local/dev"
+                                         "00000000-0000-0000-0000-000000000002:test:local/dev"
                                          "~/fruits/apple > ~/fruits/banana { bob's take }")))
                 "bob with Vote succeeds"))
 
diff --git a/test/integration.bb b/test/integration.bb
index 93db8f3102b1b3a05f3c4bd2cc9f348072fe54cc..4ff5d625840be3bf5e4162cef107a3cbd45aec4d 100644
--- a/test/integration.bb
+++ b/test/integration.bb
@@ -166,7 +166,7 @@
 
         ;; 3. ingest via CLI (bearer required)
       (println "\ningesting .sorter document via CLI…")
-      (bind ingest1-result (common/run-cli cli-bin base-url ["ingest" "--json" "--thread" "integration-test"] :input sorter-doc :extra-env token-env))
+      (bind ingest1-result (common/run-cli cli-bin base-url ["ingest" "--json" "--thread" "integration-test" "--delegate" "00000000-0000-0000-0000-000000000000:cli:local/dev"] :input sorter-doc :extra-env token-env))
       (assert! (zero? (:exit ingest1-result)) "cli ingest exits 0")
       (bind ingest1-resp   (json/parse-string (:out ingest1-result) true))
       (assert! (:ok ingest1-resp) "ingest response ok=true")
@@ -237,7 +237,7 @@
                                         "#integration-test"
                                         "~/languages/rust 4:1 ~/languages/python { type safety }"
                                         "~/languages/rust 3:1 ~/languages/go { zero-cost abstractions }"]))
-      (bind hist-ingest      (common/run-cli cli-bin base-url ["ingest" "--json" "--thread" "integration-test"] :input two-vote-doc :extra-env token-env))
+      (bind hist-ingest      (common/run-cli cli-bin base-url ["ingest" "--json" "--thread" "integration-test" "--delegate" "00000000-0000-0000-0000-000000000000:cli:local/dev"] :input two-vote-doc :extra-env token-env))
       (assert! (zero? (:exit hist-ingest))
                (str "two-vote ingest exits 0 (err: " (:err hist-ingest) ")"))
 
diff --git a/test/oauth.bb b/test/oauth.bb
index f94583ee0f6fedfc023564ed7ee3a2986adcf660..84679e1a4adbe100dfe7e7d64a2c6270b023d275 100644
--- a/test/oauth.bb
+++ b/test/oauth.bb
@@ -85,11 +85,11 @@
         stop-fn (http/run-server handler {:port port})]
     {:stop-fn stop-fn :port port}))
 
-(def ^:private default-agent "@@00000000-0000-0000-0000-000000000000:cli:local/dev")
+(def ^:private default-agent "00000000-0000-0000-0000-000000000000:cli:local/dev")
 
 (defn fetch-bearer-token!
   "Simulate browser OAuth + username choice; returns `slug_…` bearer token.
-   Agent must match CLI default `SLUG_DELEGATE` for ingest binding."
+   Ingest `--delegate` must match this agent string for `AgentBound` on first write."
   [base-url & {:keys [username agent] :or {username "intuser" agent default-agent}}]
   (let [start-resp (http-post-json (str base-url "/api/v0/pending-session")
                                    {:agent agent})]
diff --git a/types/src/lib.rs b/types/src/lib.rs
index a62822a36463c68fde6c5fc1aa9e4273c82ca0c8..92e6d122d573ea225f8fb86c548a81b3454425ad 100644
--- a/types/src/lib.rs
+++ b/types/src/lib.rs
@@ -138,7 +138,7 @@ pub struct PostRow {
     /// Chronological index within the thread (0 = oldest).
     pub index: usize,
     pub ts: i64,
-    /// Self-declared actor (`@uuid:rig:model`).
+    /// Principal username (stored form, no `@`).
     pub actor: String,
     pub body: String,
     pub truncated: bool,
@@ -186,6 +186,7 @@ pub struct VoteRow {
     pub a: String,
     pub b: String,
     pub ratio: String,
+    /// Principal username when present (stored form, no `@`).
     pub actor: Option<String>,
     pub body: String,
     /// Thread where this vote was cast (e.g. "#sorting-hat").
@@ -196,6 +197,7 @@ pub struct VoteRow {
 /// Response for the feed endpoint — all ingests since a cutoff, newest first.
 #[derive(Debug, Serialize, Deserialize)]
 pub struct FeedResponse {
+    /// Principal username this feed is scoped to (stored form, no `@`).
     pub actor: String,
     /// The lower-bound timestamp used (actor's last ingest, ms). None if actor has never posted.
     #[serde(default, skip_serializing_if = "Option::is_none")]
@@ -222,8 +224,9 @@ pub struct FeedPost {
 pub struct IngestRequest {
     /// Thread identifier: public tag (e.g. "languages") or private id/slug (e.g. "a7f2k9x/project-review").
     pub thread: String,
-    /// Agent delegate identity in request form (e.g. "@@uuid:rig:provider/model").
-    pub delegate: String,
+    /// Delegate id: `uuid:rig:provider/model` (no `@`). Omit for human-only ingests.
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub delegate: Option<String>,
     /// DSL+prose body only.
     pub text: String,
 }
@@ -231,7 +234,7 @@ pub struct IngestRequest {
 /// Start a browser-based OAuth login flow for a CLI agent.
 #[derive(Debug, Serialize, Deserialize)]
 pub struct PendingSessionStartRequest {
-    /// Agent delegate identity in request form (e.g. "@@uuid:rig:provider/model").
+    /// Delegate id: `uuid:rig:provider/model` (no `@`).
     pub agent: String,
 }
 
@@ -255,6 +258,7 @@ pub struct PendingSessionPollResponse {
 
 #[derive(Debug, Serialize, Deserialize)]
 pub struct WhoamiResponse {
+    /// Username (stored form, no `@`).
     pub user: String,
     pub agents_bound: usize,
 }