Manual update to nvidia_llm

agents/nvidia_llm.py

@@ -4,6 +4,7 @@ import threading
 import queue as _queue_module
 import time
 import requests
+import keyring
 from flask import Flask, request, jsonify
 
 # ── Logging ───────────────────────────────────────────────────────────────────
@@ -19,12 +20,16 @@ logging.getLogger("httpx").setLevel(logging.WARNING)
 # ── Config ────────────────────────────────────────────────────────────────────
 HOST         = os.getenv("NVIDIA_HOST",  "127.0.0.1")
 PORT         = int(os.getenv("NVIDIA_PORT", "8002"))
-NVIDIA_API_KEY = os.getenv("NVIDIA_API_KEY")
+try:
+    NVIDIA_API_KEY = keyring.get_password("system", "NVIDIA_API_KEY") or os.getenv("NVIDIA_API_KEY")
+except Exception as e:
+    NVIDIA_API_KEY = os.getenv("NVIDIA_API_KEY")
 
 _QUEUE_MAX_SIZE    = int(os.getenv("NVIDIA_QUEUE_MAX", "8"))
 _REQUEST_TIMEOUT_S = int(os.getenv("NVIDIA_LLM_TIMEOUT", "600"))
 
-DEFAULT_MODEL = "google/diffusiongemma-26b-a4b-it"
+# DEFAULT_MODEL = "google/diffusiongemma-26b-a4b-it"
+DEFAULT_MODEL = "sarvamai/sarvam-m"
 
 # ── Flask app ─────────────────────────────────────────────────────────────────
 app = Flask(__name__)