search

letta-ai / letta

Letta (formerly MemGPT) is a framework for creating LLM services with memory.
https://letta.com
Apache License 2.0
13.08k stars 1.43k forks source link

Attaching data to agents is not working as expected #1240

Open adarsh-trenser opened 7 months ago

adarsh-trenser commented 7 months ago

Describe the bug Unable attaching a data source to agent's archival memory to access. image

I have attached data using following command: memgpt load directory --name treatment --input-files=treatment.pdf But not able to get source id, after execution.

Please describe your setup

Screenshots image

MemGPT Config [defaults] preset = memgpt_chat persona = sam_simple_pov_gpt35 human = cs_phd

[model] model = gpt-3.5-turbo model_endpoint = https://api.openai.com/v1 model_endpoint_type = openai context_window = 4096

[embedding] embedding_endpoint_type = openai embedding_endpoint = https://api.openai.com/v1 embedding_model = text-embedding-ada-002 embedding_dim = 1536 embedding_chunk_size = 300

[archival_storage] type = chroma path = C:\Users\1267.memgpt\chroma

[recall_storage] type = sqlite path = C:\Users\1267.memgpt

[metadata_storage] type = sqlite path = C:\Users\1267.memgpt

[version] memgpt_version = 0.3.8

[client] anon_clientid = 00000000-0000-0000-0000-000000000000

atroiano-glue commented 4 months ago

Any progress on this Bug?

sarahwooders commented 4 months ago

Thanks for reporting this! The documentation is outdated and the client API should actually be: 

client.attach_source_to_agent(source_id: uuid.UUID, agent_id: uuid.UUID)

You can get available sources and their IDs with client.list_sources().

This is admittedly not the best API, so we might change is so that you can also attach by name.