Can someone tell me how an agent answer from the attached data source

[AhangarAamir]

Can someone tell me how an agent answer from the attached data source, i uploaded the data but when i asked question related to that document it is not answering

[lemorage]

Can u attach a screenshot of its output if possible? Are u saying the agent is not answering at all, or just gives irrelevant output?

[AhangarAamir]

Agent answers but not considers the data source, screenshot is attached

[lemorage]

I think u can refer to the following code as an example.

ref: memgpt as a rag agent

from letta import create_client 
from letta.schemas.llm_config import LLMConfig

client = create_client()
client.set_default_llm_config(LLMConfig.default_config("gpt-4o-mini"))

source = client.create_source("your_docs")
client.load_file_into_source(
    filename="mydocs.pdf", 
    source_id=source.id
)

agent_state = client.create_agent()
client.attach_source_to_agent(
    agent_id=agent_state.id, 
    source_id=source.id
)


# list if the source is attached
client.list_attached_sources(agent_state.id)

# test if the attached sources work
response = client.send_message(
    agent_id=agent_state.id, 
    message = "...", 
    role = "user"
) 
print(response.messages)

[AhangarAamir]

@lemorage thanks for you reply but still not working below is my code

from letta import create_client
from letta.schemas.embedding_config import EmbeddingConfig
from letta.schemas.llm_config import LLMConfig


client = create_client()
client.set_default_llm_config(LLMConfig.default_config("letta"))
client.set_default_embedding_config(EmbeddingConfig.default_config("letta"))


source = client.create_source("client_data")
print(source.id)
client.load_file_to_source(
    filename="/home/datoin/Downloads/wildfloc/wilffloc_pages/webpage_1.html",
    source_id=source.id
)

agent_state = client.create_agent()
client.attach_source_to_agent(
    agent_id=agent_state.id,
    source_id=source.id
)


# list if the source is attached
client.list_attached_sources(agent_state.id)

# test if the attached sources work
response = client.user_message(
    agent_id=agent_state.id,
    message = "Did you know about wildfloc?"
)
print(response)

[lemorage]

I tried your above code example a couple of times, and worked out the nuances here. Basically, to solve the issue, you should edit your message to be sent like so:

response = client.user_message(
    agent_id=agent_state.id,
    message = "Did you know about wildfloc? Search Archival."
)

Search Archival is the key here. You must include this piece of text to do retrieval. As the following system prompt manifest, memgpt agents will tend to use conversation_search function to search the recall memory by default. But when you attach_source_to_agent, it will be stored in archival memory (call archival_memory_insert func). So to make agents search this area of memory (call archival_memory_search func), you need to explicitly say so.

Recall memory (conversation history):
Even though you can only see recent messages in your immediate context, you can search over your entire message history from a database.
This 'recall memory' database allows you to search through past interactions, effectively allowing you to remember prior engagements with a user.
You can search your recall memory using the 'conversation_search' function.

Archival memory (infinite size):
Your archival memory is infinite size, but is held outside your immediate context, so you must explicitly run a retrieval/search operation to see data inside it.
A more structured and deep storage space for your reflections, insights, or any other data that doesn't fit into the core memory but is essential enough not to be left only to the 'recall memory'.
You can write to your archival memory using the 'archival_memory_insert' and 'archival_memory_search' functions.
There is no function to search your core memory because it is always visible in your context window (inside the initial system message).