diff --git a/scripts/frank-default-small.env b/scripts/frank-default-small.env
index 5ad2769..fc4dd93 100644
--- a/scripts/frank-default-small.env
+++ b/scripts/frank-default-small.env
@@ -42,9 +42,12 @@ NUM_SPEC_TOKENS=5
 # Tool calling. Clients (pi, OpenWebUI) send tool_choice=auto; without
 # --enable-auto-tool-choice vLLM rejects it with HTTP 400. The Avesed
 # chat template renders tools in the Qwen3 XML format, so the qwen3_xml
-# parser is the matching one.
+# parser is the matching one. CHAT_TEMPLATE must point at the repo's
+# chat_template.jinja (the tokenizer_config.json ships none; vLLM's
+# built-in template does not render the XML tool instructions).
 ENABLE_AUTO_TOOL_CHOICE=1
 TOOL_CALL_PARSER=qwen3_xml
+CHAT_TEMPLATE=/home/dank/models/avesed-qwen36-chat_template.jinja
 
 # RDNA flash-attention for the vision encoder (and general FA paths).
 # Without this, the mm encoder falls back to torch SDPA and the warmup
diff --git a/scripts/serve-frank-small.sh b/scripts/serve-frank-small.sh
index 150e72f..a5a188a 100755
--- a/scripts/serve-frank-small.sh
+++ b/scripts/serve-frank-small.sh
@@ -35,6 +35,10 @@ fi
 : "${NUM_SPEC_TOKENS:=5}"
 : "${ENABLE_AUTO_TOOL_CHOICE:=1}"
 : "${TOOL_CALL_PARSER:=qwen3_xml}"
+# The Avesed tokenizer_config.json has no chat_template; vLLM would use its
+# built-in template which does not render the model's XML tool format. Use
+# the repo's chat_template.jinja (ships the <tool_call> instructions).
+: "${CHAT_TEMPLATE:=/home/dank/models/avesed-qwen36-chat_template.jinja}"
 
 if [[ ! -x "$VLLM_BIN" ]]; then
     echo "missing vllm binary: $VLLM_BIN" >&2
@@ -59,4 +63,5 @@ exec "$VLLM_BIN" serve "$MODEL" \
     --speculative-config "{\"method\":\"ngram\",\"prompt_lookup_max\":$NGRAM_LOOKUP_MAX,\"prompt_lookup_min\":$NGRAM_LOOKUP_MIN,\"num_speculative_tokens\":$NUM_SPEC_TOKENS}" \
     --reasoning-parser qwen3 \
     --trust-remote-code \
+    --chat-template "$CHAT_TEMPLATE" \
     "${TOOL_ARGS[@]}"