From d3ec7c7320151244f07252a73628b1417f694b06 Mon Sep 17 00:00:00 2001
From: Peter Cardenas <16930781+PeterCardenas@users.noreply.github.com>
Date: Sun, 30 Nov 2025 19:07:33 -0800
Subject: [PATCH] fix: gracefully handle utf-8 errors (#2834)

---
 lua/avante/tokenizers.lua | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/lua/avante/tokenizers.lua b/lua/avante/tokenizers.lua
index 9e86f01..36879d8 100644
--- a/lua/avante/tokenizers.lua
+++ b/lua/avante/tokenizers.lua
@@ -52,7 +52,13 @@ function M.encode(prompt)
   if not prompt or prompt == "" then return nil end
   if type(prompt) ~= "string" then error("Prompt is not type string", 2) end
 
-  return tokenizers.encode(prompt)
+  local success, result = pcall(tokenizers.encode, prompt)
+  -- Some output like terminal command output might not be utf-8 encoded, which will cause an error here
+  if not success then
+    Utils.warn("Failed to encode prompt: " .. result)
+    return nil
+  end
+  return result
 end
 
 ---@param prompt string