"Couldn't connect to Weaviate, check your URL/API KEY"

[e-p-armstrong]

Description

I am running into inconsistent errors regarding not being able to connect to weaviate. This is when I am attempting to run locally with Ollama. The error only appears occasionally; often, regenerating the message will be enough to fix it, but it is occurring very often and is quite annoying:

INFO:     127.0.0.1:61907 - "POST /api/get_suggestions HTTP/1.1" 200 OK
✔ Received query: somequery
✔ Connecting new Client
ℹ Connecting to Weaviate Embedded
✘ Couldn't connect to Weaviate, check your URL/API KEY: Embedded DB did
not start because processes are already listening on ports http:8079 and
grpc:50050use weaviate.connect_to_local(port=8079, grpc_port=50050) to connect
to the existing instance
⚠ Query failed: Couldn't connect to Weaviate, check your URL/API KEY:
Embedded DB did not start because processes are already listening on ports
http:8079 and grpc:50050use weaviate.connect_to_local(port=8079,
grpc_port=50050) to connect to the existing instance
INFO:     127.0.0.1:61898 - "POST /api/query HTTP/1.1" 200 OK
✔ Succesfully Connected to Weaviate

I tried checking out what is at port 8079, it seems to be weaviate?

So I don't know why this is failing

^ Example of two failed tries and one success, the only difference between them being the time of the submission

Installation

• [ x ] pip install goldenverba
• pip install from source
• Docker installation

If you installed via pip, please specify the version:

Python version? 3.11
package version? Whatever's latest on pip

Weaviate Deployment

• [x ] Local Deployment
• Docker Deployment
• Cloud Deployment

Configuration

Reader: ?
Chunker: ?
Embedder: Ollama (nomic embed text)
Retriever: Advanced
Generator: Ollama (custom finetuned llama 3)

Steps to Reproduce

Pip install;
Run both embedding and generator with ollama
Upload 47 .md documents
chat a few times
experience issue

Additional context

For some reason Verba does not seem to be working when offline either, even though I assume local deployment is just that -- local. What's going on and how do I force this to 1) run locally and 2) correctly connect to the server so I don't get this issue?

[thomashacker]

Thanks for the issue. When you start Verba in local mode, it will run Weaviate Embedded in the background.(https://weaviate.io/developers/weaviate/installation/embedded).

Looks like a Weaviate Instance is already running on a specific port of yours, and it might be the issue that there is a problem with the way Verba is closing and opening the client everytime you start the application. Are you running any Docker deployed Weaviate instances that make create this conflict?

[thomashacker]

Ok I found out the source of the issue, there was a race condition when creating the client. Multiple requests would try to build the same client with the same credentials, resulting in an error that the client already exists. I've built in an async lock mechanism and it should work in the next version. Thanks a lot for the issue, this was a huge bug!

[thomashacker]

Should be fixed in the newest v2.1 version