Add REASONING env toggle for llama-server reasoning mode

README.md

@@ -30,6 +30,7 @@ Optional tuning variables:
 - `UBATCH_SIZE`: micro-batch size, default `512`
 - `CACHE_TYPE_K`: KV cache type for keys, default `q8_0`
 - `CACHE_TYPE_V`: KV cache type for values, default `q8_0`
+- `REASONING`: reasoning mode, accepts `True`, `False`, or `auto` and maps to llama.cpp `--reasoning`
 - `PARALLEL`: optional override for concurrent prompt decode slots
 - `PORT`: listen port, default `7860`
 

start.sh

@@ -15,6 +15,7 @@ export QUANT_PREFERENCE="${QUANT_PREFERENCE:-q4_k_m,q4_k_s,q4_k,q4_0,q4_1}"
 export PERF_PROFILE="${PERF_PROFILE:-balanced}"
 export CACHE_TYPE_K="${CACHE_TYPE_K:-f16}"
 export CACHE_TYPE_V="${CACHE_TYPE_V:-f16}"
+export REASONING="${REASONING:-auto}"
 
 # Ensure data directories exist and are writable. Run as root in container startup.
 echo "preparing storage: $HF_HOME and $MODEL_DIR"
@@ -115,6 +116,22 @@ server_args=(
   --metrics
 )
 
+case "${REASONING,,}" in
+  true|1|on|yes)
+    server_args+=(--reasoning on)
+    ;;
+  false|0|off|no)
+    server_args+=(--reasoning off)
+    ;;
+  auto|"")
+    server_args+=(--reasoning auto)
+    ;;
+  *)
+    echo "invalid REASONING value: $REASONING (expected True/False/auto)"
+    exit 1
+    ;;
+esac
+
 if [[ -n "${PARALLEL:-}" ]]; then
   server_args+=(--parallel "$PARALLEL")
 fi