Atualizar app/app.py

app/app.py

@@ -418,6 +418,51 @@ def admin_conversations(email):
 #       API
 # ================
 
+@app.route("/api/chat", methods=["POST"])
+@login_required
+@rate_limited
+def chat():
+    data = request.json
+    model = data.get("model", "llama3.2")
+    messages = data.get("messages", [])
+
+    url = f"{app.config['OLLAMA_BASE_URL']}/api/chat"
+    payload = {"model": model, "messages": messages}
+
+    r = requests.post(url, json=payload)
+    r.raise_for_status()
+
+    return jsonify(r.json())
+
+@app.route("/api/chat/stream")
+@login_required
+@rate_limited
+def chat_stream():
+    model = request.args.get("model", "llama3.2")
+    cid = request.args.get("conversation_id")
+
+    db = SessionLocal()
+    msgs = db.query(Message).filter_by(conversation_id=cid).order_by(Message.created_at).all()
+    db.close()
+
+    chatml = [{"role": m.role, "content": m.content} for m in msgs]
+
+    def generate():
+        url = f"{app.config['OLLAMA_BASE_URL']}/api/chat"
+        payload = {"model": model, "messages": chatml, "stream": True}
+
+        with requests.post(url, json=payload, stream=True) as r:
+            for line in r.iter_lines():
+                if not line:
+                    continue
+                data = json.loads(line.decode())
+                token = data.get("message", {}).get("content", "")
+                yield f"data: {json.dumps({'token': token})}\n\n"
+
+    return Response(generate(), mimetype="text/event-stream")
+
+
+
 
 @app.route("/api/new_conversation", methods=["POST"])
 @login_required