Refactor call_api function to call_nvidia_api

app.py

@@ -27,23 +27,9 @@ def user(message, history):
     history.append({"role": "user", "content": message})
     return history
 
-def call_api(history, max_tokens, temperature, top_p, system_message="", seed=42):
-    print("Calling API...")
-    messages = []
-
-    # Adicionando a mensagem de sistema, se presente
-    if system_message:
-        messages.append({"role": "system", "content": system_message})
-
-    # Adicionando mensagens de usuário e assistente da conversa
-    for msg in history:
-        messages.append({"role": msg["role"], "content": msg["content"]})
-
-    # Verificando se a última mensagem é do usuário, se não, ajuste conforme necessário
-    if not messages or messages[-1]["role"] != "user":
-        print("A última mensagem deve ser do usuário com conteúdo preenchido.")
-        return ""
-
+def call_nvidia_api(history, max_tokens, temperature, top_p, seed=42):
+    # Preparar o payload com o histórico de chat formatado
+    messages = [{"role": "user" if i % 2 == 0 else "assistant", "content": msg} for i, msg in enumerate(history)]
     payload = {
         "messages": messages,
         "temperature": temperature,
@@ -52,6 +38,7 @@ def call_api(history, max_tokens, temperature, top_p, system_message="", seed=42
         "seed": seed,
         "stream": True
     }
+    
     response = requests.post(INVOKE_URL, headers=headers, json=payload, stream=True)
     full_response = ""
     for line in response.iter_lines():
@@ -60,27 +47,24 @@ def call_api(history, max_tokens, temperature, top_p, system_message="", seed=42
             if decoded_line.startswith("data:"):
                 try:
                     json_data = json.loads(decoded_line[5:])
+                    # Processar a resposta da API aqui
+                    # Supondo que a resposta da API seja diretamente o texto a ser adicionado ao chat
+                    full_response += json_data.get("content", "")
                 except json.JSONDecodeError:
                     print(f"Invalid JSON: {decoded_line[5:]}")
-                    json_data = {}
-                if "choices" in json_data and len(json_data["choices"]) > 0:
-                    deltas = json_data["choices"][0].get("delta", {})
-                    if "content" in deltas:
-                        full_response += deltas["content"]
-    print(f"Full API response: {full_response}")
     return full_response
 
 def chat(history, system_message, max_tokens, temperature, top_p, top_k, repetition_penalty):
     print("Starting chat...")
-    system_message_to_use = system_message if system_message.strip() else BASE_SYSTEM_MESSAGE
-    if history and history[-1]["role"] == "user":
-        history = user(history[-1]["content"], history)
+    # Chamar a API da NVIDIA aqui com o histórico formatado
+    assistant_response = call_nvidia_api(history, max_tokens, temperature, top_p)
+    # Atualizar o histórico com a resposta do assistente
+    if history:
+        history[-1][1] += assistant_response
     else:
-        history.append({"role": "system", "content": system_message_to_use})
-    assistant_response = call_api(history, max_tokens, temperature, top_p)
-    if assistant_response:
-        history.append({"role": "assistant", "content": assistant_response})
-    return history, "", ""
+        history.append(["", assistant_response])
+    return history, history, ""
+
 
 # Gradio interface setup
 with gr.Blocks() as demo: