fixup! feat(rec): Fix prompt storage

app.py

@@ -14,6 +14,7 @@ VLLM_MODEL_NAME = os.getenv("VLLM_MODEL_NAME")
 VLLM_GPU_MEMORY_UTILIZATION = float(os.getenv("VLLM_GPU_MEMORY_UTILIZATION"))
 VLLM_MAX_SEQ_LEN = int(os.getenv("VLLM_MAX_SEQ_LEN"))
 HF_TOKEN = os.getenv("HF_TOKEN")
+VLLM_DTYPE = os.getenv("VLLM_DTYPE")
 
 # -------------------------------- HELPERS -------------------------------------
 def load_prompt(path: Path) -> str:
@@ -40,6 +41,7 @@ local_llm = LLM(
     gpu_memory_utilization=VLLM_GPU_MEMORY_UTILIZATION,
     hf_token=HF_TOKEN,
     enforce_eager=True,
+    dtype=VLLM_DTYPE,
 )
 
 json_schema = StructuredQueryRewriteResponse.model_json_schema()