🩹 fix: Codex review on context projection (G1 guard, IDOR, recount, summary)

- Guard `currentActive` against a stale window: a model/window switch on the
  current branch left the live snapshot outranking the projection (G1 didn't
  fire). Now defers to the projection unless streaming or the window matches.
- Scope branch lookups to the authenticated user (`getMessages` filter +
  injected `userId`) — was loading any conversation by id (IDOR).
- Recount messages with no stored `tokenCount` via the tokenizer instead of
  charging 0, so snapshot-less/imported histories don't under-report.
- Fall back (null) for already-summarized branches rather than projecting from
  the full raw parent chain (the next call would send summary + tail); the
  client's summary-baseline-aware estimate handles them until a follow-up
  replays the summary boundary.

api/server/controllers/ContextProjectionController.js

@@ -18,7 +18,7 @@ async function contextProjectionController(req, res) {
       return;
     }
     const projection = await resolveContextProjection(
-      { getMessages: db.getMessages, getAgent: db.getAgent },
+      { userId: req.user?.id, getMessages: db.getMessages, getAgent: db.getAgent },
       params,
     );
     res.json(projection ?? null);

client/src/hooks/Chat/useTokenUsage.ts

@@ -237,7 +237,13 @@ export default function useTokenUsage({
      *  one branch's breakdown onto its siblings. */
     const currentActive =
       snapshot != null &&
-      (isSubmitting || (snapshot.anchorMessageId != null && branchTotals.containsAnchor));
+      (isSubmitting ||
+        (snapshot.anchorMessageId != null &&
+          branchTotals.containsAnchor &&
+          /** G1: once streaming ends, a model/window switch leaves the live
+           *  snapshot's baked window stale — defer to the projection instead of
+           *  showing the old window/prune boundary on the current branch. */
+          (resolvedMax == null || snapshot.breakdown.maxContextTokens === resolvedMax)));
 
     /** Precedence: live/active snapshot → fresh persisted branch snapshot →
      *  server projection (covers G1 stale-window + G2 snapshot-less branches) →
@@ -258,8 +264,7 @@ export default function useTokenUsage({
     if (effective != null) {
       const breakdown = effective.breakdown;
       const maxTokens = effective.contextBudget ?? breakdown.maxContextTokens;
-      const instructionTokens =
-        effective.effectiveInstructionTokens ?? breakdown.instructionTokens;
+      const instructionTokens = effective.effectiveInstructionTokens ?? breakdown.instructionTokens;
       const baseUsed =
         effective.remainingContextTokens != null
           ? maxTokens - effective.remainingContextTokens
@@ -324,5 +329,6 @@ export default function useTokenUsage({
     branchSnapshot,
     branchSnapshotFresh,
     projection,
+    resolvedMax,
   ]);
 }

packages/api/src/endpoints/projection.ts

@@ -7,6 +7,7 @@ interface ProjectionMessage {
   messageId: string;
   parentMessageId?: string | null;
   tokenCount?: number;
+  summaryTokenCount?: number;
   isCreatedByUser?: boolean;
   text?: string;
 }
@@ -19,8 +20,10 @@ interface ProjectionAgent {
 }
 
 export interface ContextProjectionDeps {
+  /** Authenticated requester — branch lookups are scoped to this user. */
+  userId?: string;
   getMessages: (
-    filter: { conversationId: string },
+    filter: { conversationId: string; user?: string },
     select?: string,
   ) => Promise<ProjectionMessage[]>;
   getAgent: (filter: { id: string }) => Promise<ProjectionAgent | null>;
@@ -87,14 +90,23 @@ export async function resolveContextProjection(
   params: TContextProjectionRequest,
 ): Promise<TContextUsageEvent | null> {
   const stored = await deps.getMessages(
-    { conversationId: params.conversationId },
-    'messageId parentMessageId tokenCount isCreatedByUser text',
+    { conversationId: params.conversationId, user: deps.userId },
+    'messageId parentMessageId tokenCount summaryTokenCount isCreatedByUser text',
   );
   const branch = resolveBranch(stored, params.messageId);
   if (branch.length === 0) {
     return null;
   }
 
+  /** A summarized/compacted branch's next call sends the saved summary + the
+   *  post-summary tail, NOT this raw parent chain — projecting from the full
+   *  history would prune/count the wrong context and omit the summary. Until the
+   *  follow-up replays the summary boundary, fall back (null) so the client's
+   *  summary-baseline-aware estimate handles these branches. */
+  if (branch.some((message) => (message.summaryTokenCount ?? 0) > 0)) {
+    return null;
+  }
+
   let instructions: string | undefined;
   let providerValue: string | undefined = params.endpoint;
   let model = params.model;
@@ -112,18 +124,26 @@ export async function resolveContextProjection(
     return null;
   }
 
+  const encoding = (model ?? '').toLowerCase().includes('claude') ? 'claude' : 'o200k_base';
+  const tokenCounter = await createTokenCounter(encoding);
+
   const messages: BaseMessage[] = [];
   const indexTokenCountMap: Record<string, number> = {};
   for (let i = 0; i < branch.length; i++) {
     const message = branch[i];
     const text = message.text ?? '';
-    messages.push(message.isCreatedByUser === true ? new HumanMessage(text) : new AIMessage(text));
-    indexTokenCountMap[String(i)] = message.tokenCount ?? 0;
+    const lcMessage =
+      message.isCreatedByUser === true ? new HumanMessage(text) : new AIMessage(text);
+    messages.push(lcMessage);
+    /** Recount messages with no stored count (imported / pre-feature) rather
+     *  than charging 0 — a real 0 and "unknown" must not collapse, or the
+     *  snapshot-less histories this endpoint targets would under-report. */
+    indexTokenCountMap[String(i)] =
+      message.tokenCount != null && message.tokenCount > 0
+        ? message.tokenCount
+        : tokenCounter(lcMessage);
   }
 
-  const encoding = (model ?? '').toLowerCase().includes('claude') ? 'claude' : 'o200k_base';
-  const tokenCounter = await createTokenCounter(encoding);
-
   return projectAgentContextUsage({
     agent: {
       agentId: params.agentId ?? 'projection',