"Cannot resolve address {}".format(address)) chatbot | ValueError: Cannot resolve address URL:7687

[miriam-z]

Hi @tomasonjo

Cloned the project:

git clone https://github.com/tomasonjo/NeoGPT-Recommender.git

and do have: .env OPENAI_KEY= NEO4J_URL=bolt://URL:7687 NEO4J_USER="neo4j" NEO4J_PASS=

neo start

Running locally in a Mac OSX host machine

But after docker-compose up

raise ValueError("Cannot resolve address {}".format(address)) chatbot | ValueError: Cannot resolve address URL:7687

1. Are you manually creating the database up front via Cypher commands and then using GPT4 to translate user queries into Cypher commands?

or:

2. Using user queries for both the database creation as well as the queries?

Also don't see anywhere entities are actually created with for example:

CREATE (pfiz: Company {name: 'Pfizer'})

as I would like to create my own dataset

[tomasonjo]

I am not creating the database, only querying it, so its the

Manually creating the database up front via Cypher commands and then using GPT4 to translate user queries into Cypher commands

You will need to instantiate a Neo4j database and then set the correct URL in the .env file.

[miriam-z]

@tomasonjo Thanks, because I already have pharmacological dataset so I only need to populate it in my non-dockerized jupyter notebook entity-relationship creation as I have already done so and change the example training string to something that is more closely aligned with my dataset.

Although it my app it is not dockerized, but your repo is? So is that the reason for the:

Cannot resolve address URL:7687

[tomasonjo]

If you are using jupyter notebook i suggest you take a look at langchain: https://towardsdatascience.com/langchain-has-added-cypher-search-cb9d821120d5

[miriam-z]

Could you point out why ValueError: Cannot resolve address URL:7687 as it is?

[tomasonjo]

Because URL is not valid address. Where is your Neo4j instance anyway?

[miriam-z]

It is local, so to make it work I need to spin out https:// i.e cloud, so is your neo4j local or cloud? So would a local juypter notebook running be able to link to this dockerized app can link to the same cluster.

[miriam-z]

Would a local juypter notebook running be able to link to this dockerized app can link to the same cluster.

[tomasonjo]

I have no idea where does Jupyter Notebook comes into play. I don't have any Jupyter notebook here.

Changing the url to something like:

bolt://localhost:7687

should work and adding

network_mode: host

in the docker-compose.yml file if you have a local instance of Neo4j

[miriam-z]

Thanks @tomasonjo , can populate user defined dataaset into a blank sandbox from neo4j now at:

bolt://MY_IP:7687

But the streamlit UI:

http://localhost:8501/

Screenshot 2023-08-03 at 12.38.45.pdf

In this case change gpt4 to gpt3.5?

Tried:

    completions = openai.ChatCompletion.create(
        model="gpt-4-0613",

Also:

InvalidRequestError: The model gpt-4 does not exist or you do not have access to it. Learn more: https://help.openai.com/en/articles/7102672-how-can-i-access-gpt-4.

Also, Is there any way to anchor user defined entities on the graph such that these default X-Y coordinates are at a preferred location and let other entities flow around them.