docai_parse_codes

Utility scripts for integrating with Google Cloud Document AI API and processing document data.
Built for medtax-ocr-prototype on GCP.

Let us know if we miss something to provide

---

Features

• Extracts document data from PDFs/images using Document AI.
• Processes and finalizes structured JSON output.
• Supports local testing or running in Google Cloud Shell.
• Includes a firestore writing process to store data.

---

How to Test Locally

Note: If you are using Google Cloud Shell, you can skip the setup section.

Use the following link to auto clone the repository to the Google Cloud Shell

Cloud Shell Clone

Note: Google Cloud Shell has a weekly quota and limits of 50 hours a week for usage.

Setup Google Cloud

• Install Google Cloud SDK:
  Download GoogleCloudSDKInstaller.exe
• In your project folder, run:

gcloud init
gcloud auth application-default login

This opens a browser — sign in with the Google account that has project access. Set the project:

gcloud config set project medtax-ocr-prototype

If a browser didn't open, copy the link and paste it on your browser

Install Dependencies:

pip install -r requirements.txt

Testing Extraction

In extractor_caller.py, uncomment and update:

gcs_output_uri = "gs://practice_sample_training/docai/"
gcs_input_uri = "gs://run-sources-medtax-ocr-prototype-us-central1/4 form 2307 pictures.pdf"
input_mime_type = "application/pdf"

gcs_input_uri: path to the document you want to process.

gcs_output_uri: path where processed files will be saved.

To see the results: Uncomment print lines in handle_data.py to see results in your terminal. Or check the output file in your GCS bucket — files ending with _finalized.json contain extracted values.

Documentation links

• Setup Document AI
• Send Process Request Sample Code Used