Chatbot AI Assistant (Context Knownledge Base) + Vector Database (Weaviate)

Technology: OpenAi, Weaviate

Method: Completions (davinci model)

Description:

Chatbot developed with Python and Flask that features conversation with a virtual assistant. This make use of OpenAi for completions and embeddings, it will also use Weaviate as a vector store to host data. By using Weaviate it allows to properly format data with classes and objects that are previously defined on a schema. One of the advantages of Weaviate is by having great performance for semantic search and long-term memory since it will save all queries that are ran on it.

On this demo is it used a Weaviate Cloud hosted cluster to save the data objects. To create a cluster go to: https://console.weaviate.cloud/dashboard It requires data preparation before running the demo, to do so follow the instructions below.

How to run (commands Windows terminal with Python 2.7):

Part One: Compose data

• Create the dataset objects containing the information you wish to use and place it on folder '/schemas'
• It must follow the example of file 'teamlyzer_companies_dataset.json'

Part Two: Prepare data

• Define necessary parameters (OpenAi API key, ...) on file 'weaviate_prepare_data.py'

• Initialize virtual environment and install dependencies, run:

    virtualenv env
    env\Scripts\activate
    pip install weaviate-client
  

• Set Weaviate Authentication login info for the Weaviate Cloud (https://console.weaviate.cloud/dashboard):

    resource_owner_config = weaviate.AuthClientPassword(
    	username = "",
    	password = ""
    )
  

• Set the URL of the Weaviate cluster created on Weaviate Cloud (https://console.weaviate.cloud/dashboard):

    client = weaviate.Client(
    	url="https://test2-qlps4q84.weaviate.network"
    )
  

Part Three: Run the script

• To run the script:

    python weaviate_prepare_data.py
  

• It will create the schema, import data to the cluster by batchs and cross-reference objects to assign matching IDs for object dependant objects (Example: Company X has Reviews 1 and 2, Company Y has Reviews 3 and 4, ...)

Part Four: Run the chat app

• Define necessary parameters (OpenAi API key, ...) on file 'app.py'

• Install dependencies, run:

    pip install flask python-dotenv
    pip install openai
    flask run
  

• Enter "http://localhost:5000" on browser to interact with app

Changelog

• v0.1
  • initial build