Open dawenxi-007 opened 1 week ago
Thanks! This is a missing functionality in our llama-stack server to implement "delete" for memory_banks. To fix:
We will need to implement unregister function for memory_banks: https://github.com/meta-llama/llama-stack/blob/f04b566c5cfc0d23b59e79103f680fe05ade533d/llama_stack/apis/memory_banks/memory_banks.py#L66
Update AgentStore example to use the unregister api: https://github.com/meta-llama/llama-stack-apps/blob/13f4e83e134855f723b0a1998cb24df5c496fa3a/examples/agent_store/api.py#L265-L273
Thanks. How can I enable the RAG for both memory_bank
and live_bank
?
When I launched the command PYTHONPATH=. python examples/agent_store/app.py [::] 5000 --bank-ids memory_bank live_bank
, I will get the following error with any prompt:
File "/usr/lib/python3.10/asyncio/runners.py", line 44, in run
return loop.run_until_complete(main)
File "/usr/lib/python3.10/asyncio/base_events.py", line 649, in run_until_complete
return future.result()
File "/home/tao/demo_1024/llamastk_metaflow/llama-stack-apps/examples/agent_store/api.py", line 243, in chat
event_type = event.payload.event_type
AttributeError: 'NoneType' object has no attribute 'payload'
Thanks. How can I enable the RAG for both
memory_bank
andlive_bank
?When I launched the command
PYTHONPATH=. python examples/agent_store/app.py [::] 5000 --bank-ids memory_bank live_bank
, I will get the following error with any prompt:File "/usr/lib/python3.10/asyncio/runners.py", line 44, in run return loop.run_until_complete(main) File "/usr/lib/python3.10/asyncio/base_events.py", line 649, in run_until_complete return future.result() File "/home/tao/demo_1024/llamastk_metaflow/llama-stack-apps/examples/agent_store/api.py", line 243, in chat event_type = event.payload.event_type AttributeError: 'NoneType' object has no attribute 'payload'
Unfortunately, the error outputs on client is very general (cc @dineshyv who is making improvements on error handling).
Could you share the error message from the server side?
It shows that the live_bank is not registered. How can I enable the app to support multiple banks?
System Info
Python version: 3.10.12 Pytorch version: llama_models version: 0.0.45 llama_stack version: 0.0.45 llama_stack_client version: 0.0.47 Hardware: 1xH100 (80GB VRAM/GPU)
Information
🐛 Describe the bug
Running the following commands to create the memory bank and launch the app:
llama-stack-client memory_banks list
shows the available banks arememory_bank
andlive_bank
, butlive_bank
can not be cleared with theClear Bank
button.Error logs
The memory banks stay the same after clicking
Clear Bank
button. Basically, thellama-stack-client memory_banks list
shows the same memory banks even after clickingClear Bank
button.Expected behavior
Expect that the
Clear Bank
button will remove/un-register the live_bank (dynamic bank) and keep memory_bank (static bank). Hope this can be fixed, or at least, could you share any command line to manually clear/remove/un-register the memory bank?