frank-default-small: enable tool calling (--enable-auto-tool-choice --tool-call-parser qwen3_xml)

Clients send tool_choice=auto; without these flags vLLM rejects with
HTTP 400. The Avesed chat template renders the Qwen3 XML tool format,
so qwen3_xml is the matching parser.

scripts/frank-default-small.env

@@ -39,6 +39,13 @@ NGRAM_LOOKUP_MAX=5
 NGRAM_LOOKUP_MIN=1
 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.
+ENABLE_AUTO_TOOL_CHOICE=1
+TOOL_CALL_PARSER=qwen3_xml
+
 # RDNA flash-attention for the vision encoder (and general FA paths).
 # Without this, the mm encoder falls back to torch SDPA and the warmup
 # OOMs (256 GiB score matrix at max_model_len; measured 2026-08-17).