Update app.py

app.py

@@ -10,6 +10,17 @@ headers = {
     "Content-Type": "application/json"
 }
 
+
+import { HfInference } from '@huggingface/inference'
+
+hf = new HfInference(API_URL, API_KEY)
+prompt = 'Write instructions to teach anyone to write a discharge plan. List the entities, features and relationships to CCDA and FHIR objects in boldface.'
+stream = hf.textGenerationStream({ inputs: prompt })
+for await (const r of stream) { 
+    st.markdown(r.token.text)
+}
+
+
 def query(payload):
     response = requests.post(API_URL, headers=headers, json=payload)
     st.markdown(response.json())
@@ -19,7 +30,7 @@ def get_output(prompt):
     return query({"inputs": prompt})
 
 def main():
-    st.title("Medical Transcription Summarizer")
+    st.title("Medical Llama Test Bench with Inference Endpoints Llama 7B")
     example_input = st.text_input("Enter your example text:")
 
     if st.button("Summarize with Variation 1"):