I have deployed my bot using rasa x and I have connected the tracker store to DynamoDB however the problem is that the conversation related data is getting stored in the disk itself. This is consuming lot of memory. The best solution would be that rasa does not store the conversations to the disk but the as far as I have researched, it is not possible so I want some mechanism by which older conversations are auto deleted.
Hi Swaraj,
You said it’s being stored to disk
and then that it’s also consuming memory
. Can you clarify the full path where you are seeing the disk space being used? How are you monitoring memory?
Can you also post your endpoints.yml
.
Greg
models:
url: ${RASA_MODEL_SERVER}
token: ${RASA_X_TOKEN}
wait_time_between_pulls: ${RASA_MODEL_PULL_INTERVAL}
tracker_store:
type: dynamo
table_name: rasa
region: ap-south-1
event_broker:
type: "pika"
url: ${RABBITMQ_HOST}
username: ${RABBITMQ_USERNAME}
password: ${RABBITMQ_PASSWORD}
queues:
- ${RABBITMQ_QUEUE}
action_endpoint:
url: ${RASA_USER_APP}/webhook
token: ""
All the conversations are being stored in the disk as well as in the dynamodb. I say it is consuming memory because the conversations are visible on rasa x even if I delete them from dynamodb
Rasa X stores it’s own copy of the conversations in it’s own database. You can see the architecture here.