world: pre-seed repo tree on clone (V3)

Repo focus mode used to open into an empty tree — the file/dir nodes
only synthesized as agents touched files via tool calls. Cold-start
users saw a lonely repo:<id> orb with nothing under it. Now the tree
pre-seeds from the actual cloned repo the moment a client subscribes.

Backend:
- active_research_topics also returns repo_workspace_path.
- preseed_repo_paths(clone_path) runs `git ls-files` (bounded to top 200)
  against the on-disk clone. Silently returns empty on any failure so a
  missing clone / git-off-PATH / empty repo just degrades to the pre-V3
  behavior (tree still builds on touch).
- SSE loop, on first sight of a topic per client, emits one
  node.activity per pre-seeded path (label = leaf name, heat 0) so the
  tree is quiet-solid at rest.

Frontend:
- engine.onNodeActivity now synthesizes the dir:<partial> chain for
  file: nodeIds the same way onTouch does — otherwise the pre-seed
  would render as flat leaves under ROOT.
- Same 5-line synthesis extracted from onTouch; both paths now agree on
  the layout.

Cap of 200 keeps the SSE payload bounded on huge repos; the tail fills
in as agents actually touch files. When we later add per-file heat map
(V4), the 200 already-known files get first-class treatment out of the
gate.

crates/cm-api/src/routes/world.rs

@@ -62,16 +62,21 @@ async fn active_runs(pool: &PgPool, ws: WorkspaceId) -> Vec<(String, String)> {
 }
 
 /// Active research topics + their assigned agents. Each returned row is a
-/// `(topic_id, title, agent_id)` — one row per (topic, agent) pair. Emitted
-/// from the SSE loop as `repo:<topic_id>` project orbs so the World shows a
-/// clickable, labeled landmark for every in-flight R&D initiative — no need
-/// for a file touch to land first.
+/// `(topic_id, title, agent_id, repo_workspace_path)` — one row per
+/// (topic, agent) pair. Emitted from the SSE loop as `repo:<topic_id>`
+/// project orbs so the World shows a clickable, labeled landmark for
+/// every in-flight R&D initiative — no need for a file touch to land
+/// first. `repo_workspace_path` (when non-null) is the on-disk clone
+/// location; the SSE loop uses it to pre-seed the repo tree.
 async fn active_research_topics(
     pool: &PgPool,
     ws: WorkspaceId,
-) -> Vec<(String, String, String)> {
+) -> Vec<(String, String, String, Option<String>)> {
     let rows = sqlx::query(
-        "SELECT t.id::text AS topic_id, t.title AS title, ra.agent_id::text AS agent_id
+        "SELECT t.id::text AS topic_id,
+                t.title AS title,
+                t.repo_workspace_path AS repo_path,
+                ra.agent_id::text AS agent_id
            FROM research_topics t
            JOIN research_topic_agents ra ON ra.topic_id = t.id
           WHERE t.workspace_id = $1
@@ -87,11 +92,46 @@ async fn active_research_topics(
                 r.get::<String, _>("topic_id"),
                 r.get::<String, _>("title"),
                 r.get::<String, _>("agent_id"),
+                r.try_get::<Option<String>, _>("repo_path").unwrap_or(None),
             )
         })
         .collect()
 }
 
+/// Cap on pre-seeded file entries per repo. Large repos surface only the
+/// top N so the SSE payload stays bounded — a subsequent tool call
+/// exercising a specific path will fill in additional nodes on demand.
+const REPO_PRESEED_CAP: usize = 200;
+
+/// Read the top-level file list of a topic's cloned repo via `git ls-files`
+/// so the SSE loop can pre-seed dir:/file: nodes in the client engine.
+/// Bounded by `REPO_PRESEED_CAP`. Returns an empty vec on any failure
+/// (missing clone, git not on PATH, empty repo) — a missing pre-seed
+/// degrades gracefully to the pre-V3 behavior (tree builds as agents
+/// touch files).
+async fn preseed_repo_paths(clone_path: &str) -> Vec<String> {
+    let path = std::path::Path::new(clone_path);
+    if !path.join(".git").exists() {
+        return Vec::new();
+    }
+    let out = tokio::process::Command::new("git")
+        .arg("-C")
+        .arg(path)
+        .arg("ls-files")
+        .output()
+        .await;
+    let Ok(out) = out else { return Vec::new() };
+    if !out.status.success() {
+        return Vec::new();
+    }
+    String::from_utf8_lossy(&out.stdout)
+        .lines()
+        .filter(|l| !l.trim().is_empty())
+        .take(REPO_PRESEED_CAP)
+        .map(|s| s.to_string())
+        .collect()
+}
+
 /// Enabled scheduled loops + their assigned agents. Same shape as
 /// `active_research_topics` — `(loop_id, title, agent_id)` per (loop, agent).
 /// Emitted as `loop:<loop_id>` landmark orbs so recurring/scheduled work is
@@ -428,13 +468,36 @@ pub async fn world_live(State(state): State<AppState>, Authed(user): Authed) ->
             // so the affinity is visible even in idle windows.
             let research = active_research_topics(&pool, ws).await;
             let mut seen_topics = std::collections::HashSet::new();
-            for (topic_id, title, agent_id) in &research {
+            for (topic_id, title, agent_id, repo_path) in &research {
                 let node_id = format!("repo:{topic_id}");
                 if seen_topics.insert(topic_id.clone()) {
                     yield sse(
                         "node.activity",
                         json!({ "nodeId": node_id, "label": title, "kind": "service", "heat": 0.0 }),
                     );
+                    // Pre-seed the repo tree (V3). One-shot on first sight
+                    // of the topic per SSE client. Each file emits with
+                    // heat=0 so the tree is quiet-solid at rest — activity
+                    // still hot-swaps as agents touch files. Bounded to
+                    // REPO_PRESEED_CAP so payload stays reasonable.
+                    if let Some(clone_path) = repo_path {
+                        for p in preseed_repo_paths(clone_path).await {
+                            let leaf = std::path::Path::new(&p)
+                                .file_name()
+                                .and_then(|s| s.to_str())
+                                .unwrap_or(&p)
+                                .to_string();
+                            yield sse(
+                                "node.activity",
+                                json!({
+                                    "nodeId": format!("file:{p}"),
+                                    "label": leaf,
+                                    "kind": "service",
+                                    "heat": 0.0,
+                                }),
+                            );
+                        }
+                    }
                 }
                 yield sse(
                     "world.touch",