Pass API key via query string, fix cardiac scenario

src/inference/qwen.rs

@@ -1,7 +1,7 @@
 use anyhow::Result;
 use serde_json::Value;
 
-/// Call the real vLLM endpoint on AMD MI300X.
+/// Call the real vLLM endpoint on AMD MI300X via ngrok tunnel.
 /// Falls back to mock if the GPU is unreachable.
 pub async fn generate(redacted_prompt: &str) -> Result<(String, String, String)> {
     let vllm_url = std::env::var("VLLM_URL")
@@ -9,11 +9,13 @@ pub async fn generate(redacted_prompt: &str) -> Result<(String, String, String)>
     let api_key = std::env::var("VLLM_API_KEY")
         .unwrap_or_else(|_| "abc-123".to_string());
 
+    // Append API key as query parameter (ngrok-friendly)
+    let url_with_key = format!("{}?api_key={}", vllm_url, api_key);
+
     let client = reqwest::Client::new();
     let resp = client
-        .post(&vllm_url)
+        .post(&url_with_key)
         .header("ngrok-skip-browser-warning", "true")
-        .header("Authorization", format!("Bearer {}", api_key))
         .json(&serde_json::json!({
             "model": "Qwen/Qwen2.5-7B-Instruct",
             "prompt": redacted_prompt,

static/index.html

@@ -90,7 +90,7 @@
 
     <script>
         const scenarios = {
-            cardiac: "Patient John Morrison, DOB 05/12/1957, MRN 847291-A, SSN 123-45-6789, Insurance Aetna #Z829-443. Presents with acute substernal chest pain radiating to left arm, diaphoresis, nausea. History of HTN, DM2. ECG shows ST elevation in leads II, III, aVF. Troponin pending.",
+            cardiac: "Patient John Morrison, 67 yo, MRN 847291A, SSN 123-45-6789. Presents with acute substernal chest pain radiating to left arm. History of HTN, DM2. ECG shows ST elevation in leads II, III, aVF. Troponin pending.",
             pediatric: "Patient Jane Doe, 7yo female, MRN 293-B, Parents: Michael & Sarah Doe, Phone 555-123-4567. Fell from tree, complaining of left upper quadrant pain. CT shows grade III splenic laceration. Vitals: HR 120, BP 90/60."
         };