default lancedb error

[yangboz]

it happens at python3.9, macosx13.5.2 , with pip install autollm, FastAPI example code ,after run it, it throw errors at following:

AutoQueryEngine.from_defaults(
  File "/Users/yangboz/anaconda3/envs/py39/lib/python3.9/site-packages/autollm/auto/query_engine.py", line 258, in from_defaults
    return create_query_engine(
  File "/Users/yangboz/anaconda3/envs/py39/lib/python3.9/site-packages/autollm/auto/query_engine.py", line 105, in create_query_engine
    vector_store_index = AutoVectorStoreIndex.from_defaults(
  File "/Users/yangboz/anaconda3/envs/py39/lib/python3.9/site-packages/autollm/auto/vector_store_index.py", line 84, in from_defaults
    lancedb_uri = AutoVectorStoreIndex._validate_and_setup_lancedb_uri(
  File "/Users/yangboz/anaconda3/envs/py39/lib/python3.9/site-packages/autollm/auto/vector_store_index.py", line 143, in _validate_and_setup_lancedb_uri
    raise ValueError(
ValueError: No existing database found at ./.lancedb. Please provide a valid lancedb uri.
Sentry is attempting to send 2 pending events
Waiting up to 2 seconds
Press Ctrl-C to quit

any ideas? thanks.

[yangboz]

even though I manually create a .lancedb file in current project file system.

[yangboz]

from autollm import AutoQueryEngine
 import lancedb
 db = lancedb.connect("~/.lancedb") 

it comes from https://github.com/safevideo/autollm/#create-a-query-engine-in-seconds

[yangboz]

python3.9/pathlib.py", line 1323, in mkdir
self._accessor.mkdir(self, mode)
FileExistsError: [Errno 17] File exists maybe is it python mkdir to overwrite existed file related ?

[AyushExel]

@SeeknnDestroy @fcakyon let me know if this error is related to lancedb's latest release in some way and I'll get it fixed

[fcakyon]

@yangboz can you please share an example code on your use case?

[yangboz]

so , I have tried rm the existed .lancedb file at first , then modified the code to
import lancedb
db = lancedb.connect("./.lancedb"), it works, but after that the whole piece of code is ````query_engine=AutoQueryEngine.from_defaults(

documents='...'

llm_model=llm_model,
llm_api_base=llm_api_base)

but after calling ```query_engine=AutoQueryEngine.from_defaults(
# documents='...'
llm_model=llm_model,
llm_api_base=llm_api_base)
 print("query_engine:",query_engine)

 response = query_engine.query(query_str)``` , it  throws errors as following,

File "/Users/yangboz/anaconda3/envs/py39/lib/python3.9/site-packages/litellm/utils.py", line 6573, in exception_type
raise APIConnectionError(
litellm.exceptions.APIConnectionError: The api_key client option must be set either by passing api_key to the client or by setting the OPENAI_API_KEY environment variable
''' I used
llm_model = "ollama/llama2"
llm_api_base = "http://localhost:11434"
as default configs.
any idea about this? thanks.

[yangboz]

IMBO , ollama is local LLM , should't connect to OPEN_AI, right?

[fcakyon]

In your code, I see you did not set a local embedding model, so you still need openai-key for embedding generation.

[yangboz]

btw, I see the ollama provide a embbeding API here: https://github.com/ollama/ollama/blob/main/docs/api.md#generate-embeddings
wondering without using OPEN_AI_KEY purely in local environments is feasible?

[fcakyon]

your provided code and the error you receive do not match. Are you sure you provided the full code you are having error?

In addition, db = lancedb.connect("./.lancedb") should be deleted from the code.