Icon

ToxPipe General Flow

There has been no title set for this workflow's metadata.

This workflow walks you through the following steps. This same steps are probably best taken in separate workflows but they are gathered here for convenience.

Step 0: Read a document corpus, in this case documents from politics and facts of 2008

Step 1: Print a PDF with the content of one of the documents. (The PDF is printed automatically in a folder inside the workflow and automatically found by the data app inst Step 5).

Step 2: all the other documents can be used to create a FAISS vector store. This will be used by the AI later to answer questions related to these other facts. To create the vector store OpenAI API keys are necessary. You can create them at platform.openai.com/account/api-keys.

Step 3 (you can use a different workflow from here for deploying the data app): We use again the credentials selecting the same embedding model, the LLM by OpenAI (in this example GPT 3.5 Turbo) and we read the vector store.

Step 4: Provide how the chatbot settings and description of the knowledge base to give more context to the AI for the task at end.

Step 5: The component executes and looks for the PDF generated by the Step 1. If it is not found it will only use the input vector store. The user of the component view (locally) and of the data app (when deployed ) can upload a document and ask specific questions.

URL: AI Extension Example Workflows - KNIME Community Hub Space https://hub.knime.com/-/spaces/-/latest/~CzxTTU_Gi4moY5Gz/

Nodes

Extensions

Links