@yeaft/webchat-agent 0.1.918 → 0.1.920

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@yeaft/webchat-agent",
-  "version": "0.1.918",
+  "version": "0.1.920",
   "description": "Remote agent for Yeaft WebChat — connects worker machines to the central server",
   "main": "index.js",
   "type": "module",

package/yeaft/snapshot-filter.js

@@ -16,9 +16,9 @@
  *
  * Why this exists separately from the disk replay path:
  *   `web-bridge.js` builds an in-memory `baseSnapshot` for every running
- *   VP turn. Before this filter the snapshot was only `threadId`-scoped
- *   and leaked other VPs' tool calls + thinking blocks into the next
- *   turn's messages, producing orphan `tool_use` ids and Anthropic 422s.
+ *   VP turn. Before this filter the snapshot was unfiltered and leaked
+ *   other VPs' tool calls + thinking blocks into the next turn's
+ *   messages, producing orphan `tool_use` ids and Anthropic 422s.
  *   The disk path uses `ConversationStore.loadSessionHistoryForVp` —
  *   the same rules implemented here, but reading from `messages/*.md`
  *   instead of the in-memory tape. We intentionally duplicate the rule
@@ -27,8 +27,16 @@
  *
  * Pure function; does not mutate inputs.
  *
- * @param {object[]} snapshot — entries from getOrCreateSessionHistory(sessionId)
- *                              (already threadId-filtered by the caller)
+ * @param {object[]} snapshot — entries from getOrCreateSessionHistory(sessionId).
+ *                              The caller MUST NOT pre-filter by `threadId`:
+ *                              threads represent intra-VP concurrent tasks,
+ *                              not isolated conversations, and the LLM needs
+ *                              every prior thread's text for continuity.
+ *                              (See the rationale block in
+ *                              `web-bridge.js#ensureDriverRunning`; pre-fix,
+ *                              a `m.threadId === current` filter dropped
+ *                              every prior thread and the LLM saw a 1-message
+ *                              context for every turn after the first.)
  * @param {string} vpId — the VP we're about to send a turn for
  * @returns {object[]}
  */

package/yeaft/web-bridge.js

@@ -1053,12 +1053,14 @@ function ensureDriverRunning(sessionId, vpId, threadId = 'main') {
       vpAborts.set(key, vpAbort);
       turnAbortCtrls.set(turnId, vpAbort);
       turnAbortMeta.set(turnId, { sessionId, vpId, threadId: thread.threadId, key });
+      // History snapshot covers EVERY prior thread of the session.
+      // Threads represent intra-VP concurrent tasks, not isolated
+      // conversations, so the LLM needs cross-thread continuity.
+      // VP isolation lives in `filterSnapshotForVp`; tool_use/result
+      // pairing lives in `pairSanitize`. Mirrors disk-side
+      // `loadSessionHistoryForVp`, which applies no threadId filter.
       const baseSnapshot = pairSanitize(
-        filterSnapshotForVp(
-          getOrCreateSessionHistory(sessionId)
-            .filter((m) => !m.threadId || m.threadId === 'main' || m.threadId === thread.threadId),
-          vpId,
-        ),
+        filterSnapshotForVp(getOrCreateSessionHistory(sessionId), vpId),
       );
       const trigger = envelope?.trigger || 'fallback';
       const { text, prompt, promptParts } = buildVpPromptPayload(vpId, envelope);