diff --git a/cmd/sidekick-tui/main.go b/cmd/sidekick-tui/main.go
index aaa26b9..9c97fc7 100644
--- a/cmd/sidekick-tui/main.go
+++ b/cmd/sidekick-tui/main.go
@@ -90,6 +90,22 @@ func (m model) userMessages() []string {
   return msgs
 }
 
+func filterHistory(h []sidekick.Message) []sidekick.Message {
+  var history []sidekick.Message
+  for _, msg := range h {
+    if msg.Role == sidekick.MessageRoleSystem {
+      // Filter out TUI-specific system messages that are for display only
+      if strings.HasPrefix(msg.Content, banner) ||
+        msg.Content == "Request cancelled." ||
+        strings.HasPrefix(msg.Content, "Error: ") {
+        continue
+      }
+    }
+    history = append(history, msg)
+  }
+  return history
+}
+
 type agentResponseMsg struct {
   response string
   err      error
@@ -493,7 +509,7 @@ func (m *model) runAgent() tea.Cmd {
   var ctx context.Context
   ctx, m.cancel = context.WithCancel(context.Background())
   return func() tea.Msg {
-    res, err := m.agent.Run(ctx, m.history, m.pool)
+    res, err := m.agent.Run(ctx, filterHistory(m.history), m.pool)
     return agentResponseMsg{response: res, err: err}
   }
 }
diff --git a/config.toml b/config.toml
index 91f1b13..d10ae4e 100644
--- a/config.toml
+++ b/config.toml
@@ -23,7 +23,8 @@ port = 8080
 
   [models.ollama]
   # model_id = "Jackrong/Qwen3.5-2B-Claude-4.6-Opus-Reasoning-Distilled-GGUF:Q4_K_M"
-  model_id = "unsloth/Qwen3.5-0.8B-GGUF:Q8_0"
+  # model_id = "nvidia/NVIDIA-Nemotron-3-Nano-4B-GGUF:Q4_K_M"
+  model_id = "Intel/Qwen3-Coder-30B-A3B-Instruct-gguf-q2ks-mixed-AutoRound"
   # endpoint = "http://127.0.0.1:11434/v1"
   endpoint = "http://127.0.0.1:8080/v1"
   api_key = ""