fix: adjust VRAM allocation for concurrent Llama+BGE

- Llama: 85% GPU, 8K context (model needs ~15GB base)
- BGE: 10% GPU (1.3GB model)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

services/vllm/config_bge.yaml

@@ -2,6 +2,6 @@ model: BAAI/bge-large-en-v1.5
 host: "0.0.0.0"
 port: 8002
 uvicorn-log-level: "info"
-gpu-memory-utilization: 0.15
+gpu-memory-utilization: 0.10
 dtype: float16
 task: embed

services/vllm/config_llama.yaml

@@ -2,8 +2,8 @@ model: meta-llama/Llama-3.1-8B-Instruct
 host: "0.0.0.0"
 port: 8001
 uvicorn-log-level: "info"
-gpu-memory-utilization: 0.70
-max-model-len: 16384
+gpu-memory-utilization: 0.85
+max-model-len: 8192
 dtype: auto
 enforce-eager: false
 enable-auto-tool-choice: true