Upload ailib.py

ailib.py

@@ -0,0 +1,55 @@
+import openai
+import time
+
+OPENAI_TRANSCRIPTION_MODEL='whisper-1'
+OPENAI_COMPLETION_MODEL = "gpt-4o-mini"
+TOKEN_LIMIT = 4096 
+
+openai_client = openai.OpenAI()
+
+def openai_query(prompt):
+    response = openai_client.chat.completions.create(
+            model = OPENAI_COMPLETION_MODEL,
+            messages = [{'role': 'user',
+                         'content': prompt}
+                       ],
+            temperature = 0, max_tokens = TOKEN_LIMIT,
+            top_p=1.0, frequency_penalty=0.0, 
+            presence_penalty=0.0)
+    resp_text = response.choices[0].message.content      
+    return resp_text
+
+def nueva_ficha(resumen_historial, resumen_hoy):
+    
+    new_prompt = f'''Escribe una nueva ficha médica, considerando el historial médico: "{resumen_historial}",
+                    y también lo que ocurrió hoy: "{resumen_hoy}"
+                    '''
+    nueva = openai_query(new_prompt)
+    
+    return nueva
+    
+
+def whisper_transcribe(fn, temperature=0):
+    print('WT:', type(fn))
+    t0 = time.time()
+    
+    if isinstance(fn, str):
+        audio_file = open(fn,'rb')  # redundant?
+    else:
+        open('temp.wav','wb').write(fn)
+        audio_file = 'temp.wav'
+        
+    whisper = openai_client.audio.transcriptions.create
+    transcript = whisper(model=OPENAI_TRANSCRIPTION_MODEL,
+                         file = audio_file, language='es',
+                         temperature=0.0)        # to do, explore temperature
+    dt = round(time.time()-t0,2)
+
+    transcript = transcript.text
+    print(f'Whisper transcribe [dt={dt} secs]')
+    return transcript
+
+def summarize(text):
+    prompt = f"Resume todos los aspectos médicos de esta consulta: '{text}'"
+    resp_text = openai_query(prompt)
+    return resp_text