Data Agent with MCP and Chainlit -Quickstart - in memeory conversation not working

[inigokintana]

Hello,

I was testing this postgreSQL MCP quickstart in my own microk8s env.

I seems the agent will NOT remember all the previous conversation if code is not changed importing "from dapr_agents.memory import ConversationDaprStateMemory" and adding" memory=ConversationDaprStateMemory(store_name="conversationstore", session_id="my-unique-id")" in the agent code part.

...
from dapr_agents.memory import ConversationDaprStateMemory
...
global agent
    agent = Agent(
        name="SQL",
        role="Database Expert",
        instructions=instructions,
        tools=tools,
        memory=ConversationDaprStateMemory(
            store_name="conversationstore", session_id="my-unique-id"
            ), # here we say to use redis
    )
...

Best regards

[CasperGN]

Hi @inigokintana,

If the agent is not instantiated with

memory=ConversationDaprStateMemory(
    store_name="conversationstore", session_id="my-unique-id"
)

The agent has no place for storing it's short/long term memory.
The PostgreSQL example shows how to use a database as an MCP tool so the Agent can interact with it.

If you wish to store the memory of the agent in PostgreSQL you can swap the redis component with a PostgreSQL one instead.

Hope this provides some help!

[sicoyle]

closing as this should be resolved now 🙌