Document to Structured Data
Turn PDFs, scans, images, and office files into structured records with Google Gemini on Vertex AI. No account is required.
Document intelligence
Turn documents into records you can check
Upload a file, choose the fields, and review what was extracted before you download it. Missing values stay empty. No account is required.
Supported files
PDF, JPG, JPEG, PNG, WebP, DOCX, TXT, CSV, XLSX
Example uses
Invoices, Contracts, Financial reports, Receipts, Research documents
Exports
JSON keeps nested fields. CSV uses dotted names and record or line-item rows. Excel uses separate sheets.
Privacy
Files go to private storage for this job only. They expire with the job and are not published.
Recent jobs
History in this browser uses the guest access token for each job. Jobs you cannot open are left out.
No jobs in this browser yet.
Extract structured data from documents
Document to Structured Data reads invoices, receipts, contracts, financial reports, and other files, then returns fields you can review. Extraction runs on Google Gemini through Vertex AI. You can export JSON, CSV, or Excel after you check the values.
Supported files
Upload PDF, JPG, JPEG, PNG, WebP, DOCX, TXT, CSV, or XLSX. Scanned PDFs and images are read when OCR is on. The server checks the file signature, size, and page count. Macros are not run.
Guest access and retention
No registration is required. A guest token in this browser reopens the job. Files stay in private storage until the job expires, then cleanup removes them. Read the privacy policy before uploading sensitive documents.
How document extraction works
Step 1
Upload documents
Add PDFs, images, Word, text, CSV, or Excel files. Files stay in private storage for this job.
Step 2
Choose what to extract
Start from an invoice, receipt, contract, financial report, or general schema, or describe your own fields.
Step 3
Review the records
Check missing values, warnings, and model estimates. Correct anything before you export.
Step 4
Download the data
Export JSON, CSV, or Excel. The download link is time-limited and tied to this job.
Frequently Asked Questions
Do I need an account?
No. The tool uses a guest access token stored in this browser. That token is how you reopen the job. It is not an account.
Which files are supported?
PDF, JPG, JPEG, PNG, WebP, DOCX, TXT, CSV, and XLSX. Password-protected files and macro-enabled spreadsheets are rejected. The server checks the file contents, not just the extension.
How long are files kept?
Jobs and their files expire on the date shown in the workspace. Expired files are deleted from private storage. Do not upload documents you would not send to a processor.
Are confidence scores probabilities?
No. When a model estimate is shown, it is the model’s own score, not a calibrated probability. Validation warnings come from deterministic checks such as missing required fields or totals that do not add up.
Can extraction be wrong?
Yes. The tool does not invent missing values, but it can misread a document. Review important fields before you rely on an export.