diff --git a/router/agy_proxy.py b/router/agy_proxy.py
index 16adb8e9..05368826 100644
--- a/router/agy_proxy.py
+++ b/router/agy_proxy.py
@@ -43,6 +43,8 @@ async def save(self) -> None:
logger = logging.getLogger("agy-proxy")
+AGY_DAEMON_URL = (os.getenv("AGY_DAEMON_URL") or "http://127.0.0.1:5005").rstrip("/")
+
# In container: mounted from host /home/gpav/.local/bin/agy
AGY_BINARY = os.environ.get("AGY_BINARY_PATH", "/usr/local/bin/agy")
if not os.path.exists(AGY_BINARY):
@@ -91,7 +93,7 @@ async def _run_agy_print(client: httpx.AsyncClient, prompt: str, model_override:
"""
Forward the agy execution request to the host-side agy daemon.
"""
- url = "http://127.0.0.1:5005/run"
+ url = f"{AGY_DAEMON_URL}/run"
payload = {
"prompt": prompt,
"model_override": model_override,
@@ -299,7 +301,7 @@ async def try_agy_proxy(prompt: str, messages: list = None,
tier_timeout = min(AGY_TIMEOUT_SECS, remaining)
if stream:
- url = "http://127.0.0.1:5005/run"
+ url = f"{AGY_DAEMON_URL}/run"
payload = {
"prompt": proxy_prompt,
"model_override": tier["env_override"],
diff --git a/router/main.py b/router/main.py
index 90cd9a4c..23a465e4 100644
--- a/router/main.py
+++ b/router/main.py
@@ -39,6 +39,37 @@
except ImportError:
try_agy_proxy = None
+from urllib.parse import urlparse
+
+# Global Configuration from Environment
+LITELLM_URL = (os.getenv("LITELLM_ADMIN_URL") or "http://127.0.0.1:4000").rstrip("/")
+LLAMA_SERVER_URL = (os.getenv("LLAMA_SERVER_URL") or "http://127.0.0.1:8080").rstrip("/")
+LANGFUSE_URL = (os.getenv("LANGFUSE_URL") or "http://127.0.0.1:3001").rstrip("/")
+VALKEY_HOST = os.getenv("VALKEY_HOST") or "127.0.0.1"
+
+def _get_valkey_port() -> int:
+ val = os.getenv("VALKEY_PORT")
+ if not val:
+ return 6379
+ try:
+ return int(val)
+ except ValueError:
+ return 6379
+
+VALKEY_PORT = _get_valkey_port()
+
+def _get_port_suffix(url: str) -> str:
+ try:
+ port = urlparse(url).port
+ return f":{port}" if port else ""
+ except Exception:
+ return ""
+
+LITELLM_PORT = _get_port_suffix(LITELLM_URL)
+LLAMA_SERVER_PORT = _get_port_suffix(LLAMA_SERVER_URL)
+LANGFUSE_PORT = _get_port_suffix(LANGFUSE_URL)
+
+
_redis_client = None
_redis_last_init_attempt = 0.0
_REDIS_RETRY_INTERVAL_SECONDS = 5.0
@@ -53,8 +84,8 @@ def get_redis():
return None
_redis_last_init_attempt = now
try:
- host = os.getenv("VALKEY_HOST", "127.0.0.1")
- port = int(os.getenv("VALKEY_PORT", "6379"))
+ host = VALKEY_HOST
+ port = VALKEY_PORT
_redis_client = aioredis.Redis(host=host, port=port, decode_responses=True, socket_timeout=1.0)
logger.info(f"Valkey client initialized at {host}:{port}")
except Exception as e:
@@ -177,7 +208,7 @@ def get_langfuse():
_langfuse_client = langfuse.Langfuse(
public_key=os.getenv("LANGFUSE_PUBLIC_KEY", ""),
secret_key=os.getenv("LANGFUSE_SECRET_KEY", ""),
- host=os.getenv("LANGFUSE_HOST", "http://127.0.0.1:3001"),
+ host=LANGFUSE_URL,
release="llm-triage-router-v1",
)
logger.info("Langfuse client initialized")
@@ -242,7 +273,7 @@ async def push_aggregate_scores():
port = config.get("server", {}).get("port", 5000)
router_model_conf = config.get("router", {}).get("router_model", {})
-router_api_base = router_model_conf.get("api_base", "http://127.0.0.1:8080/v1")
+router_api_base = router_model_conf.get("api_base", f"{LLAMA_SERVER_URL}/v1")
router_api_key = router_model_conf.get("api_key", "local-token")
router_model_name = router_model_conf.get("model", "qwen-0.8b-routing")
@@ -410,7 +441,7 @@ async def sync_adaptive_router_roster(master_key: str):
logger.warning("No LITELLM_MASTER_KEY — skipping roster sync")
return
headers = {"Authorization": f"Bearer {master_key}", "Content-Type": "application/json"}
- admin_url = "http://127.0.0.1:4000"
+ admin_url = LITELLM_URL
try:
async with httpx.AsyncClient(timeout=5.0) as client:
r = await client.get("https://openrouter.ai/api/v1/models")
@@ -567,7 +598,7 @@ async def _register_ollama_models_in_db(master_key: str):
logger.warning("No LiteLLM master key provided — skipping Ollama DB registration")
return
- admin_url = os.getenv("LITELLM_ADMIN_URL", "http://127.0.0.1:4000")
+ admin_url = LITELLM_URL
headers = {"Authorization": f"Bearer {master_key}", "Content-Type": "application/json"}
ollama_models = []
@@ -680,10 +711,10 @@ async def lifespan(app: FastAPI):
get_http_client()
await sync_cooldowns_from_valkey()
- litellm_ready_url = "http://127.0.0.1:4000/health/readiness"
+ litellm_ready_url = f"{LITELLM_URL}/health/readiness"
litellm_master_key = os.getenv("LITELLM_MASTER_KEY", "")
max_wait = 180
- logger.info(f"⏳ Waiting for LiteLLM on :4000 (max {max_wait}s)...")
+ logger.info(f"⏳ Waiting for LiteLLM on {LITELLM_URL} (max {max_wait}s)...")
for i in range(max_wait):
try:
async with httpx.AsyncClient(timeout=2.0) as client:
@@ -1131,7 +1162,7 @@ async def get_llamacpp_metrics() -> dict:
try:
async with httpx.AsyncClient(timeout=3.0) as client:
# Fetch model list
- r = await client.get("http://127.0.0.1:8080/v1/models")
+ r = await client.get(f"{LLAMA_SERVER_URL}/v1/models")
if r.status_code == 200:
data = r.json()
for m in data.get("data", []):
@@ -1146,7 +1177,7 @@ async def get_llamacpp_metrics() -> dict:
"n_embd": meta.get("n_embd"),
})
# Fetch props for build info
- r2 = await client.get("http://127.0.0.1:8080/props")
+ r2 = await client.get(f"{LLAMA_SERVER_URL}/props")
if r2.status_code == 200:
props = r2.json()
result["build"] = props.get("build_info", "unknown")
@@ -1154,7 +1185,7 @@ async def get_llamacpp_metrics() -> dict:
loaded = [m["id"] for m in result["models"] if m["status"] == "loaded"]
slot_model = loaded[0] if loaded else (result["models"][0]["id"] if result["models"] else None)
if slot_model:
- r3 = await client.get(f"http://127.0.0.1:8080/slots?model={slot_model}")
+ r3 = await client.get(f"{LLAMA_SERVER_URL}/slots?model={slot_model}")
if r3.status_code == 200:
slots_data = r3.json()
for s in slots_data:
@@ -1330,8 +1361,8 @@ def get_pie_chart_gradient() -> str:
@app.api_route("/v1/memory{path:path}", methods=["GET", "POST", "DELETE", "PUT"])
async def proxy_memory(request: Request, path: str = ""):
- """Proxies memory API calls to the LiteLLM gateway on port 4000."""
- litellm_base = "http://127.0.0.1:4000/v1/memory"
+ """Proxies memory API calls to the LiteLLM gateway."""
+ litellm_base = f"{LITELLM_URL}/v1/memory"
# Resolve the destination URL
url = f"{litellm_base}{path}"
@@ -1384,7 +1415,7 @@ async def proxy_models():
async with httpx.AsyncClient(timeout=10.0) as client:
auth_header = "Bearer " + (litellm_key or "")
r = await client.get(
- "http://127.0.0.1:4000/v1/models",
+ f"{LITELLM_URL}/v1/models",
headers={"Authorization": auth_header}
)
data = r.json()
@@ -2131,10 +2162,10 @@ async def get_dashboard_data():
"""Fetch all metrics and pre-compute HTML snippets for the dashboard."""
await sync_cooldowns_from_valkey()
# 1. Run live health checks
- valkey_status = await check_tcp_port("127.0.0.1", 6379)
- litellm_status = await check_http_endpoint("http://127.0.0.1:4000/")
- llama_server_status = await check_http_endpoint("http://127.0.0.1:8080/health")
- langfuse_status = await check_http_endpoint("http://127.0.0.1:3001")
+ valkey_status = await check_tcp_port(VALKEY_HOST, VALKEY_PORT)
+ litellm_status = await check_http_endpoint(f"{LITELLM_URL}/")
+ llama_server_status = await check_http_endpoint(f"{LLAMA_SERVER_URL}/health")
+ langfuse_status = await check_http_endpoint(LANGFUSE_URL)
# 1c. Check Gemini OAuth token status
oauth_status = await asyncio.to_thread(get_gemini_oauth_status)
@@ -2456,6 +2487,7 @@ async def get_dashboard_data():
"t_tokens": t_tokens,
"llamacpp_models_html": llamacpp_models_html,
"llamacpp_slots_html": llamacpp_slots_html,
+ "llamacpp": llamacpp,
"avg_triage_latency_ms": stats["avg_triage_latency_ms"],
"avg_proxy_latency_ms": stats["avg_proxy_latency_ms"],
"cache_hits": stats["cache_hits"],
@@ -2493,6 +2525,7 @@ async def get_dashboard():
t_tokens = data["t_tokens"]
llamacpp_models_html = data["llamacpp_models_html"]
llamacpp_slots_html = data["llamacpp_slots_html"]
+ llamacpp = data["llamacpp"]
avg_triage_latency_ms = data["avg_triage_latency_ms"]
avg_proxy_latency_ms = data["avg_proxy_latency_ms"]
cache_hits = data["cache_hits"]
@@ -3012,7 +3045,7 @@ async def get_dashboard():
@@ -3077,7 +3110,7 @@ async def get_dashboard():
LiteLLM Proxy
- :4000
+ {LITELLM_PORT}
{'Online' if litellm_status else 'Offline'}
@@ -3087,7 +3120,7 @@ async def get_dashboard():
Valkey Cache
- :6379
+ :{VALKEY_PORT}
{'Online' if valkey_status else 'Offline'}
@@ -3097,7 +3130,7 @@ async def get_dashboard():
Llama-Server
- :8080
+ {LLAMA_SERVER_PORT}
{'Online' if llama_server_status else 'Offline'}
@@ -3107,7 +3140,7 @@ async def get_dashboard():
Langfuse Traces
- :3001
+ {LANGFUSE_PORT}
{'Online' if langfuse_status else 'Offline'}
@@ -3144,15 +3177,15 @@ async def get_dashboard():