diff --git a/gpt4all-chat/chatllm.cpp b/gpt4all-chat/chatllm.cpp
index 073a1d7a..b54a8696 100644
--- a/gpt4all-chat/chatllm.cpp
+++ b/gpt4all-chat/chatllm.cpp
@@ -770,14 +770,14 @@ bool ChatLLM::prompt(const QList<QString> &collectionList, const QString &prompt
 
 bool ChatLLM::promptInternal(const QList<QString> &collectionList, const QString &prompt, const QString &promptTemplate,
         int32_t n_predict, int32_t top_k, float top_p, float min_p, float temp, int32_t n_batch, float repeat_penalty,
-        int32_t repeat_penalty_tokens)
+        int32_t repeat_penalty_tokens, bool isToolCallResponse)
 {
     if (!isModelLoaded())
         return false;
 
     QList<SourceExcerpt> databaseResults;
     const int retrievalSize = MySettings::globalInstance()->localDocsRetrievalSize();
-    if (!collectionList.isEmpty()) {
+    if (!collectionList.isEmpty() && !isToolCallResponse) {
         emit requestRetrieveFromDB(collectionList, prompt, retrievalSize, &databaseResults); // blocks
         emit sourceExcerptsChanged(databaseResults);
     }
@@ -847,7 +847,7 @@ bool ChatLLM::promptInternal(const QList<QString> &collectionList, const QString
         }
 
         SuggestionMode mode = MySettings::globalInstance()->suggestionMode();
-        if (mode == SuggestionMode::On || (!databaseResults.isEmpty() && mode == SuggestionMode::LocalDocsOnly))
+        if (mode == SuggestionMode::On || (mode == SuggestionMode::SourceExcerptsOnly && (!databaseResults.isEmpty() || isToolCallResponse)))
             generateQuestions(elapsed);
         else
             emit responseStopped(elapsed);
@@ -876,7 +876,7 @@ bool ChatLLM::toolCallInternal(const QString &toolCall, int32_t n_predict, int32
     } else {
         qWarning() << "WARNING: Could not find the tool for " << toolCall;
         return promptInternal(QList<QString>()/*collectionList*/, QString() /*prompt*/, promptTemplate,
-            n_predict, top_k, top_p, min_p, temp, n_batch, repeat_penalty, repeat_penalty_tokens);
+            n_predict, top_k, top_p, min_p, temp, n_batch, repeat_penalty, repeat_penalty_tokens, true /*isToolCallResponse*/);
     }
 
     const QString apiKey = MySettings::globalInstance()->braveSearchAPIKey();
@@ -888,7 +888,7 @@ bool ChatLLM::toolCallInternal(const QString &toolCall, int32_t n_predict, int32
     emit sourceExcerptsChanged(braveResponse.second);
 
     return promptInternal(QList<QString>()/*collectionList*/, braveResponse.first, promptTemplate,
-        n_predict, top_k, top_p, min_p, temp, n_batch, repeat_penalty, repeat_penalty_tokens);
+        n_predict, top_k, top_p, min_p, temp, n_batch, repeat_penalty, repeat_penalty_tokens, true /*isToolCallResponse*/);
 }
 
 void ChatLLM::setShouldBeLoaded(bool b)
diff --git a/gpt4all-chat/chatllm.h b/gpt4all-chat/chatllm.h
index 31e99c71..b4874b70 100644
--- a/gpt4all-chat/chatllm.h
+++ b/gpt4all-chat/chatllm.h
@@ -199,7 +199,7 @@ Q_SIGNALS:
 protected:
     bool promptInternal(const QList<QString> &collectionList, const QString &prompt, const QString &promptTemplate,
         int32_t n_predict, int32_t top_k, float top_p, float min_p, float temp, int32_t n_batch, float repeat_penalty,
-        int32_t repeat_penalty_tokens);
+        int32_t repeat_penalty_tokens, bool isToolCallResponse = false);
     bool toolCallInternal(const QString &toolcall, int32_t n_predict, int32_t top_k, float top_p, float min_p, float temp, int32_t n_batch, float repeat_penalty,
         int32_t repeat_penalty_tokens);
     bool handlePrompt(int32_t token);
diff --git a/gpt4all-chat/mysettings.cpp b/gpt4all-chat/mysettings.cpp
index 57f94ca2..e1530a85 100644
--- a/gpt4all-chat/mysettings.cpp
+++ b/gpt4all-chat/mysettings.cpp
@@ -51,7 +51,7 @@ static const QVariantMap basicDefaults {
     { "saveChatsContext",         false },
     { "serverChat",               false },
     { "userDefaultModel",         "Application default" },
-    { "suggestionMode",           QVariant::fromValue(SuggestionMode::LocalDocsOnly) },
+    { "suggestionMode",           QVariant::fromValue(SuggestionMode::SourceExcerptsOnly) },
     { "localdocs/chunkSize",      512 },
     { "localdocs/retrievalSize",  3 },
     { "localdocs/showReferences", true },
diff --git a/gpt4all-chat/mysettings.h b/gpt4all-chat/mysettings.h
index d4aa4c5a..59301cf2 100644
--- a/gpt4all-chat/mysettings.h
+++ b/gpt4all-chat/mysettings.h
@@ -21,9 +21,9 @@ namespace MySettingsEnums {
      *       ApplicationSettings.qml, as well as the corresponding name lists in mysettings.cpp */
 
     enum class SuggestionMode {
-        LocalDocsOnly = 0,
-        On            = 1,
-        Off           = 2,
+        SourceExcerptsOnly = 0,
+        On                 = 1,
+        Off                = 2,
     };
     Q_ENUM_NS(SuggestionMode)
 
diff --git a/gpt4all-chat/qml/ApplicationSettings.qml b/gpt4all-chat/qml/ApplicationSettings.qml
index 1e459e99..5fe4bea5 100644
--- a/gpt4all-chat/qml/ApplicationSettings.qml
+++ b/gpt4all-chat/qml/ApplicationSettings.qml
@@ -350,7 +350,7 @@ MySettingsTab {
             Layout.alignment: Qt.AlignRight
             // NOTE: indices match values of SuggestionMode enum, keep them in sync
             model: ListModel {
-                ListElement { name: qsTr("When chatting with LocalDocs") }
+                ListElement { name: qsTr("When source excerpts are cited") }
                 ListElement { name: qsTr("Whenever possible") }
                 ListElement { name: qsTr("Never") }
             }