zylon-ai
Privategpt displays an error when you ask a question before uploading a document
Problem
After initial setting up privategpt, I have an issue where i get a Exception: Collection make_this_parameterizable_per_api_call not found on any input. This appears in the web interface as well as the web ui.
Cause
The cause of this is not having any documents loaded before asking a question.
Proposed Solution
Instead of return an error, privategpt should return some hint or useful feedback to correct the issue.
Hey there Devin! I am currently experiencing the same problem. I've narrowed it down to the /private_gpt/components/vector_store/vector_store_component.py file. In the file, the names of the qdrant and chroma databases are not declared. I will put the relevant code down below (note the commented out "TODO" sections), if anyone has a solution to this, I would be greatly appreciative!:
@singleton
class VectorStoreComponent:
vector_store: VectorStore
@inject
def __init__(self, settings: Settings) -> None:
match settings.vectorstore.database:
case "chroma":
try:
import chromadb # type: ignore
from chromadb.config import ( # type: ignore
Settings as ChromaSettings,
)
except ImportError as e:
raise ImportError(
"'chromadb' is not installed."
"To use PrivateGPT with Chroma, install the 'chroma' extra."
"`poetry install --extras chroma`"
) from e
chroma_settings = ChromaSettings(anonymized_telemetry=False)
chroma_client = chromadb.PersistentClient(
path=str((local_data_path / "chroma_db").absolute()),
settings=chroma_settings,
)
chroma_collection = chroma_client.get_or_create_collection(
"make_this_parameterizable_per_api_call"
) # TODO
self.vector_store = typing.cast(
VectorStore,
BatchedChromaVectorStore(
chroma_client=chroma_client, chroma_collection=chroma_collection
),
)
case "qdrant":
from llama_index.vector_stores.qdrant import QdrantVectorStore
from qdrant_client import QdrantClient
if settings.qdrant is None:
logger.info(
"Qdrant config not found. Using default settings."
"Trying to connect to Qdrant at localhost:6333."
)
client = QdrantClient()
else:
client = QdrantClient(
**settings.qdrant.model_dump(exclude_none=True)
)
self.vector_store = typing.cast(
VectorStore,
QdrantVectorStore(
client=client,
collection_name="make_this_parameterizable_per_api_call",
), # TODO
)
case _:
# Should be unreachable
# The settings validator should have caught this
raise ValueError(
f"Vectorstore database {settings.vectorstore.database} not supported"`
Problem
After initial setting up privategpt, I have an issue where i get a
Exception: Collection make_this_parameterizable_per_api_call not foundon any input. This appears in the web interface as well as the web ui.Cause
The cause of this is not having any documents loaded before asking a question.
Proposed Solution
Instead of return an error, privategpt should return some hint or useful feedback to correct the issue.
Fully agree! Would you open a PR with a better handling of that error? It'd be a great contribution!
this happens when you try to load your old chroma db with the new 0.1.0 version of privategpt, because the default vectorstore changed to qdrant.
go to settings.yaml and change
vectorstore: database: qdrant
to
vectorstore: database: chroma
and it should work again.
https://docs.privategpt.dev/manual/storage/vector-stores#chroma-configuration
this happens when you try to load your old chroma db with the new 0.1.0 version of privategpt, because the default vectorstore changed to qdrant. go to settings.yaml and change
vectorstore: database: qdranttovectorstore: database: chromaand it should work again. https://docs.privategpt.dev/manual/storage/vector-stores#chroma-configuration
This solution worked for me. Thank you.
I had this error on version 0.2.0
I simply didn't check I had installed the commands for the Quick Start .
I particularly found poetry needed to be install with :
curl -sSL https://install.python-poetry.org | python3 -
Just in case your like me don't forget to install these too, I took it all for granted.
sudo apt install git python3.11 python3.11-venv
@Kosmoshn thanks for the note and I hope that the documentation for Windows can reflect that (also poetry install --extras chroma)
this happens when you try to load your old chroma db with the new 0.1.0 version of privategpt, because the default vectorstore changed to qdrant. go to settings.yaml and change
vectorstore: database: qdranttovectorstore: database: chromaand it should work again. https://docs.privategpt.dev/manual/storage/vector-stores#chroma-configuration
Unfortunately this didn't work for me.
What about this paramater in the settings.yaml? Should'nt this be changed to chroma path?
qdrant: path: local_data/private_gpt/qdrant
+1 on this issue. I deleted the local_data and switched to chroma db. It started working. But the issue with qdrant needs to be fixed.
+1 on this issue. I deleted the local_data and switched to chroma db. It started working. But the issue with qdrant needs to be fixed.
How did you do that
Make sure that you have selected the "LLM Chat (no context from files)" option. I had the same error and it narrowed down to selecting the toggle while asking questions before uploading files smh
I made changes to the settings.yaml file to update the database configuration from qdrant to chroma. After making these changes, everything worked as expected.
vectorstore:
database: qdrant
qdrant:
path: local_data/private_gpt/qdrant
to
vectorstore:
database: chroma
chroma:
path: local_data/private_gpt/chroma
