multi models and multi agent support for rasa in single deployment

[pvagare]

As we know in the api.ai we have different agent and different training modules for each agent.
So is there such mechanism available in rasa.ai ?

Please let me know early as possible. Thanks !

[pvagare]

Sorry for this issue. I haven't gone through all documentation.
I got the solution : https://rasa-nlu.readthedocs.io/en/stable/http.html#section-http-config
I am closing issue.
Thank you.

[tmbo]

No worries 👍

[devashishmamgain]

@tmbo whats the latest docs link to achieve multiple agents support in the same machine?

[wrathagom]

Assuming by latest you mean what is on Master: https://nlu.rasa.com/master/http.html#serving-multiple-apps

[devashishmamgain]

thanks @wrathagom is there anything equivalent for rasa core?

[wrathagom]

Not to my knowledge, but I have a lot less experience with Core than I do NLU. Might be worth asking on the Rasa Core Gitter.

[devashishmamgain]

tried gitter sometime back, didn't get any response.
any solution for the following case:

• FAQ bot for multiple customers
• each customer have their own set of question answers
• we are able to achieve it by running multiple servers but we are looking for a way to run server on a single port and process request for different customers from their data only.

@tmbo

[vinayver198]

@devashishmamgain
Hi,

Have you achieved the above task?
If yes, Can you please explain how to achieve such scenario.

Thanks,

[streamride]

@vinayver198
hi, did you figure out how to do it?

[gambhirpulkit]

@devashishmamgain Hey Devashish. Did you find a solution for this?

[devashishmamgain]

@vinayver198 @gambhirpulkit no, couldn't find a reliable and scalable solution.

[avinash9008]

@pvagare ---> multi models and multi agent support for rasa in single deployment

Sorry for this issue. I haven't gone through all documentation.
I got the solution : https://rasa-nlu.readthedocs.io/en/stable/http.html#section-http-config
I am closing issue.

hi ... pvagare i am also facing same issue can you please give solution link the
above link i am not able to open that you have mentioned

[vkrntgtm]

Need to know, whether multiple agents or multiple projects running on single rasa server instance feature is enabled in rasa v2.1.2+ or v2.2.x?

[ftarlaci]

I have the same question above by @vkrntgtm

[taimoorkhokhar]

If anybody still stuck on getting this to work then I found a solution which seems to be working with multiple bots for a single server using a flask application.

app = Flask(__name__)
CORS(app)
cors = CORS(app, resources={r”/*”: {“origins”: “*”}})

action_endpoint = “http://localhost:5055/webhook"

def agent_get(orgid):
  orgId= {“agent_one”: ‘./models/20200728–131120.tar.gz’,
  “agent_two”: ‘./models/20200728–140043.tar.gz’,
  “agent_three”: ‘./models/20200728–140043.tar.gz’,
  “agent_four”: ‘./models/20200728–140043.tar.gz’}
  agent = Agent.load(agents_org[orgid],action_endpoint=EndpointConfig(action_endpoint))
  return agent

async def process(agent, msg):
  output = await agent.handle_text(msg)
  print(output)
  return output

@app.route(‘/message’, methods=[‘POST’])
@cross_origin(origin=’*’)
def new_message():
  if not request.json:
    abort(400)
  orgId = request.args.get(‘orgId’)
  current_agent=agent_get(orgId)
  user = request.json[‘sender’]
  message = request.json[‘message’]
  res = (current_agent.handle_text(text_message=message, sender_id=user))
  message= asyncio.run(process(current_agent, message))
  message= json.dumps(message)
  return message
  if __name__ == ‘__main__’:
    app.run(host=’localhost’, port=5000, debug=True, use_reloader=False)

payload: {
“message”: “Hi”,
“sender”: “user”,
“agent”: “agent_one”
}

[{
“recipient_id”: “default”,
“text”: “Hi, I am your virtual assistant\n\n* How can
I help you?
}]

cheers!

[Mushahid2521]

@taimoorkhokhar Thanks for the solution. I am wondering can it retain the states for each bot? I can see each time a message comes and the bot is initialized each time. So can it be used in form filling when multiple users are chatting same time?