From e54440b7d19f88302f36f81acde546df9ebdfcf1 Mon Sep 17 00:00:00 2001
From: Danny Avila <danny@librechat.ai>
Date: Sat, 18 Apr 2026 12:07:22 -0400
Subject: [PATCH] =?UTF-8?q?=F0=9F=94=81=20fix:=20don't=20seed=20initialize?=
 =?UTF-8?q?.js=20agentConfigs=20from=20the=20pre-pruning=20callback?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Addresses the fifth P1 codex finding on #12740: `onAgentInitialized`
fires during BFS, BEFORE the helper prunes agents that become
disconnected once `filterOrphanedEdges` runs. Writing the sub-agent
straight into the outer `agentConfigs` there and then only additively
merging the pruned `discoveredConfigs` left stranded entries in the
outer map, and `AgentClient` would still hand them to `createRun` as
extra parallel start nodes (the exact failure mode the pass-4 prune
was meant to eliminate for the API controllers).

Drop the `agentConfigs.set` from the callback and replace the additive
merge with a direct copy from `discoveredConfigs`, which is now the
single authoritative source of what the run should see. The
per-agent tool context map is still populated during BFS — stale
entries there are harmless because they're only read by closure inside
`ON_TOOL_EXECUTE` and are unreachable once the agent is not in
`agentConfigs`.
---
 .../services/Endpoints/agents/initialize.js     | 17 +++++++++++------
 1 file changed, 11 insertions(+), 6 deletions(-)

diff --git a/api/server/services/Endpoints/agents/initialize.js b/api/server/services/Endpoints/agents/initialize.js
index 0a329dea0e..c8515bd0fa 100644
--- a/api/server/services/Endpoints/agents/initialize.js
+++ b/api/server/services/Endpoints/agents/initialize.js
@@ -260,8 +260,13 @@ const initializeClient = async ({ req, res, signal, endpointOption }) => {
         getCodeGeneratedFiles: db.getCodeGeneratedFiles,
         filterFilesByAgentAccess,
       },
+      // The callback fires during BFS, before the helper prunes agents
+      // whose edges end up filtered. Don't populate `agentConfigs` here —
+      // `discoveredConfigs` (returned below) is the authoritative pruned
+      // set. The per-agent tool context map is OK to keep populated even
+      // for pruned ids: it's only read by closure in ON_TOOL_EXECUTE,
+      // stale entries are unreachable at runtime.
       onAgentInitialized: (agentId, agent, config) => {
-        agentConfigs.set(agentId, config);
         agentToolContexts.set(agentId, {
           agent,
           toolRegistry: config.toolRegistry,
@@ -277,12 +282,12 @@ const initializeClient = async ({ req, res, signal, endpointOption }) => {
     },
   );
 
-  // Merge any entries the onAgentInitialized callback did not receive
-  // (defensive: discoverConnectedAgents already returns a canonical map)
+  // Copy the pruned discovery result into the outer map. Anything the
+  // helper dropped (skipped or unreachable after edge filtering) is
+  // intentionally absent. `processAddedConvo` below may still add more
+  // entries for parallel multi-convo execution.
   for (const [agentId, config] of discoveredConfigs) {
-    if (!agentConfigs.has(agentId)) {
-      agentConfigs.set(agentId, config);
-    }
+    agentConfigs.set(agentId, config);
   }
 
   let userMCPAuthMap = discoveredMCPAuthMap;