diff --git a/services/vllm/config_bge.yaml b/services/vllm/config_bge.yaml
new file mode 100644
index 0000000..aba81c0
--- /dev/null
+++ b/services/vllm/config_bge.yaml
@@ -0,0 +1,7 @@
+model: BAAI/bge-large-en-v1.5
+host: "0.0.0.0"
+port: 8002
+uvicorn-log-level: "info"
+gpu-memory-utilization: 0.15
+dtype: float16
+task: embed
diff --git a/services/vllm/config_llama.yaml b/services/vllm/config_llama.yaml
index 280343b..e0331fa 100644
--- a/services/vllm/config_llama.yaml
+++ b/services/vllm/config_llama.yaml
@@ -2,7 +2,9 @@ model: meta-llama/Llama-3.1-8B-Instruct
 host: "0.0.0.0"
 port: 8001
 uvicorn-log-level: "info"
-gpu-memory-utilization: 0.95
-max-model-len: 20480
+gpu-memory-utilization: 0.70
+max-model-len: 16384
 dtype: auto
 enforce-eager: false
+enable-auto-tool-choice: true
+tool-call-parser: "llama3_json"
diff --git a/supervisord.conf b/supervisord.conf
index 767710e..f883457 100644
--- a/supervisord.conf
+++ b/supervisord.conf
@@ -74,6 +74,24 @@ priority=200
 stopwaitsecs=30
 
 
+# vLLM BGE Embedding Server (Port 8002)
+[program:bge]
+command=services/vllm/venv/bin/vllm serve --config services/vllm/config_bge.yaml
+directory=.
+autostart=false
+autorestart=true
+startretries=3
+stderr_logfile=.logs/bge.err.log
+stdout_logfile=.logs/bge.out.log
+stdout_logfile_maxbytes=50MB
+stdout_logfile_backups=10
+stderr_logfile_maxbytes=50MB
+stderr_logfile_backups=10
+environment=HF_HOME="./.cache/vllm",HF_TOKEN="%(ENV_HF_TOKEN)s"
+priority=210
+stopwaitsecs=30
+
+
 # AudioCraft Studio Service
 [program:audiocraft]
 command=services/audiocraft/venv/bin/python services/audiocraft/main.py
@@ -113,7 +131,7 @@ programs=comfyui,webdav-sync
 priority=100
 
 [group:vllm]
-programs=llama
+programs=llama,bge
 priority=200
 
 [group:audiocraft]