PromeDocAI · Document Processor
See the structure inside every page.
Document Processor finds every title, paragraph, list, table and figure on a page — with exact coordinates and confidence — and hands it back as clean JSON, tables as CSV and figures as images.
- No templates
- No training
- Scans & photos
- Multi-page PDFs
FY25 Operations Review
Throughput rose 18% year on year, led by the western plants, while unit costs held flat for the third consecutive quarter. Dispatch volumes peaked in March ahead of the monsoon maintenance window.
| Plant | Q2 | Q3 | Q4 |
|---|---|---|---|
| Pune | 412 | 436 | 471 |
| Vadodara | 388 | 401 | 429 |
| Nagpur | 297 | 322 | 341 |
Monthly dispatch volume (kt)
Key risks for FY26
- Coal price volatility
- Rail rake availability
- Monsoon-linked downtime
Outlook: capacity additions at Nagpur come online in Q2 FY26, lifting group output by an estimated 9%. Rail logistics remain the main constraint.
Prepared by Strategy Office · 14 Apr 2026
table-1.csv
Plant,Q2,Q3,Q4 Pune,412,436,471 Vadodara,388,401,429 Nagpur,297,322,341
What it detects
Five element types. Found without a single template.
Every region on a page is located, classified and handed to the right reader — so a table comes back as a table, not a jumble of words.
Tables, rebuilt cell by cell
Rows and columns are reconstructed and returned as HTML you can download as CSV in one click.
| Plant | Q2 | Q3 | Q4 |
|---|---|---|---|
| Pune | 412 | 436 | 471 |
| Vadodara | 388 | 401 | 429 |
| Nagpur | 297 | 322 | 341 |
Figures, cropped out
Charts, photos and diagrams are isolated and saved as their own images.
Titles, pulled out
Titles and section headings are separated from the body text around them.
Text in reading order
Paragraphs in the order a person would read them, across scripts.
Lists as lists
Bulleted and numbered items stay itemised instead of merging into prose.
"items": [ "Coal price volatility", "Rail rake availability", … ]
Inspect the output
Click any element. See exactly what came out.
This is the shape of a real Document Processor result: every element has a type, a box on the page and a confidence score — and tables and figures come with their own downloads.
Elements (7)
FY25 Operations Review
Throughput rose 18% year on year, led by the western plants, while unit costs held flat for the third consecutive quarter. Dispatch volumes peaked in March ahead of the monsoon maintenance window.
| Plant | Q2 | Q3 | Q4 |
|---|---|---|---|
| Pune | 412 | 436 | 471 |
| Vadodara | 388 | 401 | 429 |
| Nagpur | 297 | 322 | 341 |
Monthly dispatch volume (kt)
Key risks for FY26
- Coal price volatility
- Rail rake availability
- Monsoon-linked downtime
Outlook: capacity additions at Nagpur come online in Q2 FY26, lifting group output by an estimated 9%. Rail logistics remain the main constraint.
Prepared by Strategy Office · 14 Apr 2026
4 × 4 table · plant output by quarter
| Plant | Q2 | Q3 | Q4 |
|---|---|---|---|
| Pune | 412 | 436 | 471 |
| Vadodara | 388 | 401 | 429 |
| Nagpur | 297 | 322 | 341 |
How it works
Layout first. Then the reading.
Like an analyst skimming a report: find the structure, then read each part with the right eyes.
- 1
Drop any document
PDFs, scans or phone photos. Multi-page files are split and processed page by page.
- 2
Layout is detected
Every region on the page is located and classified as a title, text, list, table or figure.
- 3
Each element is read
Text keeps its reading order, tables are rebuilt cell by cell, figures are cropped out.
- 4
You get clean output
Coordinates and confidence for every element, as JSON — plus CSV and image downloads.
After parsing
Download it, query it, or pipe it onward.
Structured JSON
Every element with type, bounding box and confidence. Copy it or download the full result.
Tables as CSV
Each detected table is rebuilt as HTML and can be downloaded as a CSV in one click.
Figures as images
Charts, photos and diagrams are cropped out of the page and saved individually.
Annotated pages
Export the page with every bounding box drawn on it — useful for QA and audit.
Pune grew fastest — output rose from 436 in Q3 to 471 in Q4, an 8% jump. Vadodara followed at 7%.
Inputs
Built for the documents you actually have.
Not just clean, born-digital PDFs.
Scanned & skewed
Photocopies, stamps and slightly rotated pages.
Phone photos
Pages captured on site, not on a flatbed.
Long PDFs
Multi-page reports with per-page results.
Many scripts
English, Hindi, Urdu and other Indian languages.
API
Parse pages from your own code.
The processor endpoint returns the same page-level structure you see in the app — ready for search indexes, RAG pipelines or review tools.
Data extraction
Ask for the fields you need and get clean JSON back, or let AI suggest the fields from a sample.
Full page layout
Every text block, table and figure with its position, confidence and reading order.
Bulk processing
Up to 20 files in one request, each counted toward your credits and rate limit.
Secure and tracked
API-key authentication, rate limits and a log of every request with its usage.
POST /api/v1/extract/processor
X-API-Key: ••••••••••••••••
files=@annual-report.pdf{
"documents": [{
"file_name": "annual-report.pdf",
"total_pages": 12,
"pages": [{
"page_number": 1,
"text_blocks": [
{ "text": "FY25 Operations Review", "confidence": 0.99 }
],
"tables": [{ "rows": 4, "columns": 4 }]
}, …]
}]
}Sample request and response.
The platform
The same engine powers everything else.
Document Processor is the foundation. These tools are built on top of it.
8 modules
Procurement Evaluation
Tender evaluation from upload to recommendation — eligibility, comparative statements, price tabulation and clarification queries, with every verdict linked back to its source page.
- Upload Tenders & Bids
- Bid Comparative Statement
- BEC / Eligibility Checker
- Document Checker
- Price Tabulation
- Technical Query Generator
- Procurement Chatbot
- Instrument Validator
Schema Extraction
Define custom target fields and extract structured JSON from unstructured documents.
Bulk Extraction
Batch process multi-file directories with consistent schema outputs.
Compliance Check
Automated clause-by-clause contract comparison against master specifications.
Workflow Configurator
Visual node-based pipelines connecting extraction steps and destinations.
MTC Validator
Material Test Certificate parsing with automated Non-Conformance Reports.
Security & deployment
Your documents stay where they belong.
Run the platform inside your own perimeter, with access scoped per person and per module.
Your infrastructure
Deploy on-premise or inside your own private cloud. Documents never have to leave your network.
Module-level access control
Grant each team member read or write access per module, so evaluators, approvers and admins see only what they should.
Full audit trail
Every API request is logged with status and latency. Every verdict is traceable to its source page.
Encrypted, never trained on
Data is encrypted in transit and at rest. Your documents are not used for model training without explicit consent.
FAQ
Questions we hear most.
Something else on your mind? Ask us on a call.