@gotcos/glasses-server 6.36.2 → 6.36.4

package/CHANGELOG.md

@@ -2219,6 +2219,51 @@ unsaved capture, and makes batch status stop lying about finished work.
 
 # Changelog
 
+## [6.36.4] - 2026-08-17
+
+### A cloned Codex conversation gets its own identity
+
+Miles forked "Markt POS 2.0 build" into "POS Nation 3.0 build" and the fork never
+appeared in the sessions list. It had not failed to index -- it was indexed AS ITS
+PARENT. The list showed one `Markt POS 2.0 build` row whose modified time was the
+fork's activity: two conversations, one identity, and the newer one invisible.
+
+Cloning copies the parent's `session_meta` record wholesale, so the new rollout carries
+the PARENT's id under its own filename. `listCodexSessions` keyed on `meta.id` and
+folded them together.
+
+The filename is unique per rollout by construction; the meta record is content and can
+be a copy. **When they disagree, the filename now wins.** An ordinary session, where
+they agree, is untouched.
+
+Verified against the real machine before and after: the fork's `session_meta.id` reads
+`019e0943…` (the parent) while its filename reads `01a0119c…`.
+
+### Also worth recording, not fixed here
+
+That fork's rollout is **771,177,205 bytes** -- 735 MB, against other rollouts small
+enough not to register. The clone duplicated the entire parent transcript and Codex
+rewrites the whole file, which is past Node's 512 MB string limit; COS survives only
+because it reads bounded windows. Giving the fork its own identity does not shrink it,
+and a clone of a clone doubles again.
+
+## [6.36.3] - 2026-08-17
+
+### The seeded query was being crowded out by the steps
+
+Caught by probing the live stream after shipping 6.36.2, before Miles tested it: a
+real session seeded 8 events -- 6 tool calls, one prose, one status -- and **no
+prompt**. The seed took "the last 7 events of any kind", and in a busy run the
+user's question is twenty or thirty steps back, so the activity you opened the page
+to watch is exactly what pushed the query off it. The one case the feature exists
+for was the one case it failed.
+
+The newest prompt in the read window is now emitted FIRST and unconditionally,
+outside the step budget. Measured on this Mac's largest transcript (87.2 MB): the
+last 256 KiB holds 133 records including 3 user turns, so the window reaches a query
+comfortably. The client pins rather than lists it, so it costs nothing in the
+scrolling window.
+
 ## [6.36.2] - 2026-08-17
 
 ### The live view stops being a blank slate

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@gotcos/glasses-server",
-  "version": "6.36.2",
+  "version": "6.36.4",
   "description": "COS Glasses \u2014 self-hosted AI heads-up-display server for Even G2 smart glasses, powered by Claude Code, Codex, or Cursor Agent CLI",
   "type": "module",
   "bin": {

package/server/lib/agent-session-store.ts

@@ -899,7 +899,26 @@ export async function listCodexSessions(
       if (rows.length >= Math.max(0, limit)) break
       const meta = await peekCodexMeta(candidate.file)
       if (!meta || meta.subagent) continue
-      const native = meta.id || candidate.name.slice(0, -6)
+      // THE FILENAME WINS WHEN THEY DISAGREE.
+      //
+      // A rollout's `session_meta.id` is normally its own id, and normally the filename
+      // agrees. A CLONED conversation breaks that: the clone copies the parent's
+      // `session_meta` record wholesale, so the new rollout carries the PARENT's id
+      // while its filename carries its own.
+      //
+      // Miles forked "Markt POS 2.0 build" into "POS Nation 3.0 build" and the fork
+      // never appeared in the session list. It had not failed to index -- it was indexed
+      // AS its parent, so the list showed one `Markt POS 2.0 build` row whose modified
+      // time was the fork's activity. Two conversations, one identity, and the newer one
+      // invisible.
+      //
+      // The filename is unique per rollout by construction; the meta record is content
+      // and can be a copy. So when they disagree, trust the filename.
+      const fileId = idFromCodexFilename(candidate.name)
+      const metaId = (meta.id || '').trim().toLowerCase()
+      const native = fileId && metaId && fileId !== metaId
+        ? fileId
+        : (meta.id || candidate.name.slice(0, -6))
       const title = names.get(native) || meta.title || 'Codex session'
       if (isKeepWarmSessionTitle(title)) continue
       const created = meta.created || createdFromCodexFilename(candidate.name) || isoFromMtime(candidate.birthtimeMs)

package/server/routes/agent-session-stream.ts

@@ -264,7 +264,21 @@ agentSessionStreamRouter.get('/agent-sessions/:provider/:sessionId/stream', asyn
       // Status drafts are dropped from the seed: they describe the state at some past
       // moment and the opening status above is the CURRENT one. Replaying an old
       // `done` after it would tell the client the live session had finished.
-      const steps = drafts.filter(d => d.kind === 'tool' || d.kind === 'prose' || d.kind === 'prompt')
+      // THE QUERY IS SEEDED SEPARATELY, AND ALWAYS.
+      //
+      // Taking "the last 7 events" and hoping the prompt is among them does not work,
+      // and a live probe against a real session proved it: 8 seeded events, 6 tools and
+      // one prose, and NO prompt -- because in a busy run the user's question is twenty
+      // or thirty steps back and gets pushed out of the window by the very activity you
+      // opened the page to watch. That is the exact case the query exists for.
+      //
+      // So the newest prompt in the whole read window is emitted FIRST, unconditionally,
+      // and the step budget is spent entirely on steps. The client pins it rather than
+      // listing it, so it costs nothing in the scrolling window.
+      const lastPrompt = [...drafts].reverse().find(d => d.kind === 'prompt')
+      if (lastPrompt) write({ ...lastPrompt, at: Date.now() })
+
+      const steps = drafts.filter(d => d.kind === 'tool' || d.kind === 'prose')
       for (const draft of steps.slice(-SEED_EVENTS)) {
         // NOT tagged as seeded. A replayed step is a step that really happened, and a
         // second rendering style for it would be a distinction without a use. The one