Support direct cloud LLM endpoint

Author: Asaif Ali

app.py

@@ -229,8 +229,8 @@ def animate_count(placeholder, label: str, icon: str, target: int, duration: flo
     # Local development can keep using the LiteLLM gateway. Cloud deployments
     # can bypass it and call any OpenAI-compatible endpoint directly (for
     # example OpenRouter) by setting OPENAI_BASE_URL / OPENAI_API_KEY.
-    gateway_url = os.getenv("LLM_BASE_URL", "") or os.getenv("LLM_GATEWAY_URL", "")
-    gateway_key = os.getenv("LLM_API_KEY", "") or os.getenv("LLM_GATEWAY_API_KEY", "")
+    gateway_url = os.getenv("LLM_GATEWAY_URL", "")
+    gateway_key = st.session_state.get("llm_gateway_session_token", "") or os.getenv("LLM_GATEWAY_API_KEY", "")
     direct_url = os.getenv("OPENAI_BASE_URL", "")
     direct_key = os.getenv("OPENAI_API_KEY", "")
     gateway_enabled = (
@@ -249,8 +249,8 @@ def animate_count(placeholder, label: str, icon: str, target: int, duration: flo
         llm_base_url = gateway_url or direct_url or "https://openrouter.ai/api/v1"
         llm_api_key = gateway_key or direct_key
         llm_model = (
-            os.getenv("LLM_MODEL", "")
-            or os.getenv("LLM_GATEWAY_MODEL", "")
+            ("" if st.session_state.get("llm_gateway_session_token") else os.getenv("LLM_GATEWAY_MODEL", ""))
+            or os.getenv("LLM_MODEL", "")
             or os.getenv("OPENAI_MODEL", "")
             or os.getenv("OPENAI_MODEL_ID", "")
             or "openrouter/free"

crawler.py

@@ -581,7 +581,7 @@ async def _summarize_via_gateway(self, content: str) -> str:
             )
             prompt = prompt[:4000].rsplit(" ", 1)[0] + "..." if len(prompt) > 4000 else prompt
             payload = {
-                "model": self.llm_model,
+                **({"model": self.llm_model} if self.llm_model else {}),
                 "messages": [
                     {"role": "system", "content": "Summarize only the supplied webpage content."},
                     {"role": "user", "content": prompt},