feat: mlx-proxy 서버 + n8n 워크플로우 LLM/임베딩 URL 분리

mlx-vlm 기반 ollama 호환 프록시 서버 추가 (port 11435).
n8n GEN 노드 6개에 callLLM 래퍼 주입 (health check + ollama fallback).
임베딩/리랭커는 ollama(LOCAL_EMBED_URL)로 분리.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Hyungi Ahn
2026-03-19 10:00:00 +09:00
parent a050f2e7d5
commit 1137754964
5 changed files with 162 additions and 14 deletions

View File

@@ -28,6 +28,12 @@ LOCAL_OLLAMA_URL=http://host.docker.internal:11434
# Ollama (GPU 서버 — RTX 4070Ti Super, 기본 모델: id-9b:latest)
GPU_OLLAMA_URL=http://192.168.1.186:11434
# mlx-proxy (맥미니 — LLM 생성용, ollama 호환, 기본 모델: qwen3.5:27b)
LOCAL_LLM_URL=http://host.docker.internal:11435
# 임베딩 전용 (ollama — bge-m3, bge-reranker)
LOCAL_EMBED_URL=http://host.docker.internal:11434
# Qdrant (Docker 내부에서 접근)
QDRANT_URL=http://host.docker.internal:6333
@@ -63,3 +69,4 @@ CHAT_BRIDGE_URL=http://host.docker.internal:8091
CALDAV_BRIDGE_URL=http://host.docker.internal:8092
DEVONTHINK_BRIDGE_URL=http://host.docker.internal:8093
MAIL_BRIDGE_URL=http://host.docker.internal:8094
KB_WRITER_URL=http://host.docker.internal:8095