Get started
Document processor
How Doctor Bank reads your PDF and writes your answers back into it.
The document processor turns a conversation into a filled form. Knowing how it works helps you pick the right PDF and spot anything that needs a manual fix.
From upload to download
- 1
Upload
Your browser reads the PDF and attaches it to the chat.
- 2
Analyze
The server sends the PDF and the chat to Google Gemini, guided by the Doctor Bank instructions.
- 3
Save
After every exchange, the chat and PDF are saved to the database so you can resume.
- 4
Extract
When the interview ends, the assistant outputs your answers as structured data.
- 5
Fill
The server matches each answer to a field in the PDF using the pdf-lib library.
- 6
Download
The completed PDF is returned and downloads to your device.
What the extracted data looks like
{
"First Name": "John",
"Annual Income": "600000"
}Supported field types
| Field type | How it is filled |
|---|---|
| Text box | The answer is typed into the box. |
| Checkbox | Checked when the answer is true, yes, or x. |
| Radio buttons | The matching option is selected. |
How answers find their field
Names are compared without capital letters, spaces, or punctuation, so “First Name” matches a field called first_name. If a field's name doesn't match any answer, it is left blank rather than guessed.
Limits to know about
- Only fillable PDFs are filled. A scanned or flat PDF has no form fields to write into.
- Unmatched fields stay empty. Check the finished file for gaps.
- Sensitive fields are skipped by design. Account, NIC, and passport numbers are for you to add.
- Today's date is treated as already known by the assistant.