genai_contact_form_demo

Generative AI backed Contact Form written in Python/Streamlit.

Setup Instructions

Step 1: Create Corpus via create_google_rag.py

To create a corpus and generate a <project_id>_config_corpus.json file, follow these steps:

1. Ensure you have the necessary dependencies installed. You can install them using:

   pip install -r requirements.txt

2. Run the create_google_rag.py script with the required arguments:

   python create_google_rag.py -d <directory> -p <project_id> -c <credentials_path> -n <corpus_display_name> -g <gcs_source_uri> -l <location>

   • -d, --directory: Directory to upload files from (default: [./upload])
   • -p, --project_id: Your Google Cloud project ID (default: [vsp-genai])
   • -c, --credentials_path: Path to your service account key.json (default: [./key.json])
   • -n, --corpus_display_name: Display name for your dataset
   • -g, --gcs_source_uri: GCS URI for your data file (default: [gs://vision_benefits_docs])
   • -l, --location: Location for the operation (default: [us-central1])

3. This script will:

   • Upload files to Google Cloud Storage (GCS)
   • Create a RAG Corpus
   • Import files into the RAG Corpus
   • Generate a config_corpus.json file for use in the query script

Step 2: Query the Corpus off the comand-line via [google_rag_query.py]")

To query the corpus using the generated config_corpus.json file, follow these steps:

1. Run the [google_rag_query.py] script with the required arguments:

   python google_rag_query.py -c <config_corpus.json> -q <query_text>

   • -c, --corpus_config: Path to the corpus configuration file
   • -q, --query_text: The query text to search in the corpus

2. This script will:

   • Initialize Vertex AI
   • Query the corpus using the provided query text
   • Return the response from the RAG model and Gemini using Vertex AI form google.

Step 3: Configure json files and setup database

See sql_tables.sql for example database layout.
You will need a google service account key as key.json (or change that in code)
The full streamlit demo uses Baserun.AI and OpenAI you get setup free accounts or
Recode these sections as needed.

Additional Scripts

• [contact_helpdesk.py] contact_helpdesk.py"): Handles the contact form submission and response. This is full demo in streamlit and can use OpenAI via Baserun. It includes as google_rag_query.py as an import.
• [contact_utils.py] Contains utility functions for loading configurations and connecting to MySQL.
• [sql_tables.sql] Contains the sql to create the 2 tables used in this demo.

License

This project is licensed under the MIT License.# genai_contact_form_demo Generative AI backed Contact Form written in Python/Streamlit.