ComIQ PDF Table Extractor

Extract structured tables from PDF files with QA-gated delivery.

ComIQ PDF Table Extractor turns PDF tables into structured data you can use in spreadsheets, databases, APIs, automations, and downstream AI workflows.

Upload 1–10 PDF files per run. ComIQ detects table structures, reconstructs rows and columns, evaluates result quality, and returns only customer-ready tables that pass its QA gate.

Why use ComIQ PDF Table Extractor?

• Structured table output — headers, rows, page numbers, table indexes, row/column counts, and more.
• QA-gated delivery — results are quality-scored before they are returned.
• Confidence metadata — inspect table confidence and QA scores with every result.
• Totals and section handling — preserves detected section/group rows and subtotal/total rows where available.
• OCR fallback — supports PDFs that require OCR-based table reconstruction.
• Pay for delivered value — ComIQ custom billing is tied to successfully processed PDFs and QA-approved tables that are delivered.
• Apify-native output — results are written to the Actor's default Dataset and can be consumed through the Apify Console, API, integrations, schedules, or downstream workflows.

Quick start

1. Open the Actor and upload one or more PDF files, or paste direct HTTPS PDF URLs.
2. If you uploaded the files into Apify key-value storage and the run uses Limited permissions, select the same key-value store(s) under Source storage(s) for uploaded PDFs. ComIQ requests READ access only to those selected stores.
3. Set your maximum cost per run high enough for the number of documents/tables you expect.
4. Start the run.
5. When the run finishes, open Dataset / Output to view the extracted tables.

For a simple PDF containing one accepted table, a typical successful run uses:

• 1 × PDF Document Processed
• 1 × PDF Table Delivered
• the small Apify Actor Start event shown by the platform

Input

PDF files

Upload 1–10 PDF documents per run.

ComIQ is best suited to PDFs containing visibly structured tabular data such as:

• business reports
• statements
• order summaries
• operational documents
• schedules
• financial or administrative tables
• text-heavy PDFs containing embedded tables
• scanned or image-based PDFs where OCR reconstruction is needed

Best results

For the most reliable extraction:

• use PDFs with clear table boundaries or consistent column alignment;
• use legible scans when the PDF is image-based;
• avoid heavily rotated, blurred, cropped, or extremely low-resolution pages;
• use documents where columns and rows are visually distinguishable.

Complex layouts may be held back by the QA gate rather than returned as low-confidence customer-ready data.

Search-friendly use cases

ComIQ is suitable when you need a PDF table extractor, PDF to JSON, PDF to CSV, PDF to Excel/XLSX, OCR PDF table extraction, scanned PDF table extraction, or a structured PDF data extraction API with QA metadata.

Output

Each accepted table is returned as a separate item in the default Apify Dataset.

Typical output fields include:

Field	Description
source_file	Original PDF filename
page	Page containing the extracted table
table_index	Table index within the document
headers	Detected column headers
rows	Structured table rows
row_count	Number of extracted rows
column_count	Number of extracted columns
table_confidence	Table extraction confidence score
qa_score	ComIQ QA score
qa_issues	Quality issues detected by the QA gate
section_rows	Preserved section/group rows where detected
summary_rows	Preserved subtotal/total rows where detected
extraction_method	Extraction/reconstruction method used
delivery_status	Delivery state for the result
billing_mode	Preview or Pay Per Event mode
billing_status	Billing state associated with the delivered result
processing_seconds	ComIQ processing time where available

Example

{
  "source_file": "orders.pdf",
  "page": 1,
  "table_index": 1,
  "headers": ["Order No", "Account No", "Account Name", "Order Total"],
  "rows": [["100101", "ABC001", "Customer ABC001", "994.29"]],
  "row_count": 10,
  "column_count": 4,
  "table_confidence": 95,
  "qa_score": 95,
  "delivery_status": "completed",
  "extraction_method": "rapidocr_layout",
  "billing_mode": "apify_pay_per_event",
  "billing_status": "charged"
}

Pricing

ComIQ uses Pay Per Event pricing.

Event	Price
PDF Document Processed	$0.040 per successfully processed PDF
PDF Table Delivered	$0.025 per QA-approved table delivered
Actor Start	Small synthetic Apify event displayed by the platform

The primary pricing event is PDF Table Delivered.

Example pricing

A successful run containing:

• 1 PDF
• 1 QA-approved delivered table

uses approximately:

$0.040 + $0.025 + the Apify Actor Start event

A PDF containing multiple accepted tables can generate multiple PDF Table Delivered events.

Your configured maximum cost per run is respected by the Actor.

When ComIQ custom events are not charged

ComIQ's custom document/table billing is designed around successful delivery:

• failed processing is not treated as successful paid delivery;
• Revenue Guard rejected work is not treated as successful paid delivery;
• QA-held work that produces no customer-ready table is not treated as successful paid delivery;
• a table event is charged only when that table is successfully delivered to the Dataset.

Apify's own synthetic platform events, such as Actor Start, are controlled by the Apify platform and may still apply independently.

QA-gated delivery

ComIQ does not treat every detected table as automatically customer-ready.

The fulfillment engine evaluates extraction quality and can consider factors such as:

• table structure;
• row/column consistency;
• header quality;
• sparse or malformed cells;
• reconstruction quality;
• OCR confidence;
• detected totals and sections.

Results that do not meet the configured quality gate may be held instead of being returned as accepted customer-ready tables.

A QA score is a quality-screening signal, not a guarantee of semantic or ground-truth accuracy. For business-critical workflows, validate extracted data before using it for financial, legal, regulatory, or irreversible decisions.

OCR and scanned PDFs

ComIQ can use OCR-based reconstruction when direct PDF table extraction is not sufficient.

OCR can be useful for:

• scanned documents;
• image-based PDFs;
• PDFs where text is not directly extractable.

OCR accuracy depends on scan quality, font clarity, resolution, rotation, compression, and table complexity.

Limited permissions and scoped file access

Alpha 4.2 is designed to support Limited permissions.

When a PDF is uploaded into Apify key-value storage, select the key-value store containing that upload in the Actor input under Source storage(s) for uploaded PDFs. The input schema requests scoped READ + WRITE permission for the selected store(s). Apify then scopes the run token to those resources instead of granting access to unrelated account data.

If you provide an external HTTPS URL or an already-signed file URL, no Apify source-storage permission is required.

Why WRITE is requested for the selected source store

Apify's key-value-store API only exposes the URL-signing secret to clients that have WRITE permission on that specific store. ComIQ needs that signing capability to create a pre-signed public URL that the external Windows fulfillment worker can download without receiving the user's Apify API token.

The Actor remains Limited permissions overall, and the extra permission is scoped only to the store(s) the user explicitly selects. The current ComIQ code does not intentionally modify or delete the uploaded source records.

The migration code remains compatible with Full-permission runs during validation, but ComIQ's target Store configuration is Limited permissions once the scoped-storage test passes.

Security and processing architecture

• Actor-to-ComIQ communication uses authenticated HTTPS.
• The ComIQ Windows fulfillment worker uses outbound HTTPS; no public inbound port on the processing PC is required for normal fulfillment.
• Results are returned to Apify storage for the requesting Actor run.
• Do not upload documents that you are not authorized to process.

Because PDFs may contain sensitive information, users should evaluate whether this processing model is appropriate for their data and compliance requirements before submitting confidential, regulated, or highly sensitive documents.

API and automation

Because this is an Apify Actor, you can use it through:

• the Apify Console;
• the Apify API;
• Actor Tasks;
• schedules;
• webhooks and integrations supported by Apify;
• downstream automation tools that consume Apify Dataset output.

This makes ComIQ suitable for both one-off extraction and repeatable document-processing workflows.

Common use cases

PDF table to JSON

Upload a PDF and retrieve structured table data through the Dataset/API for downstream processing.

PDF table to spreadsheet workflow

Extract rows and headers, then send Dataset results to a spreadsheet or data-processing workflow.

Document QA pipeline

Use ComIQ's QA and confidence metadata to decide which extracted tables can continue automatically and which should be reviewed.

OCR table extraction

Process scanned or image-based PDFs where conventional text extraction is not enough.

Batch document processing

Upload multiple PDFs in one run and receive one structured Dataset item for each accepted table.

Limitations

ComIQ may not successfully extract every table.

Results can be affected by:

• very low-resolution scans;
• handwriting;
• highly irregular or artistic layouts;
• overlapping text;
• severe page rotation or distortion;
• tables with visually ambiguous columns;
• complex nested tables;
• documents where rows/columns cannot be reconstructed reliably.

When quality is insufficient, the Actor may return fewer tables rather than deliver a low-confidence result.

FAQ

Does ComIQ extract normal PDF text too?

The current Store product is focused on structured table extraction, not general-purpose PDF text extraction.

Does it work with scanned PDFs?

Yes, OCR fallback can be used for image-based PDFs, although quality depends on the source scan.

How many PDFs can I upload?

The current input supports up to 10 PDFs per run.

How is a table billed?

Each QA-approved table successfully delivered to the Dataset triggers the PDF Table Delivered event.

Why is there also a document-processing charge?

The document-processing event covers the successful document-level processing needed before one or more customer-ready tables can be produced.

What happens if ComIQ cannot produce a customer-ready table?

The QA gate can hold the result instead of delivering a low-confidence table. ComIQ's custom document/table events are designed not to treat that outcome as successful delivery.

Why is there a Source storage(s) field?

Under Limited permissions, the Actor cannot read arbitrary user storage. If you upload a PDF into Apify storage, selecting that key-value store explicitly grants the Actor scoped READ + WRITE access only to that selected resource for the run. External/signed HTTPS URLs do not need this selection.

Does ComIQ need Full permissions?

Alpha 4.2 is designed to migrate the Actor to Limited permissions after validation. The compatibility code can still run under Full permissions during the migration test.

Can I use the results through an API?

Yes. Results are stored in the Apify Dataset and can be retrieved using Apify's API and integrations.

Is the QA score an accuracy percentage?

No. It is a heuristic quality-screening score used by ComIQ. It should not be interpreted as guaranteed ground-truth accuracy.

Support

If a run produces an unexpected result:

1. Check the Actor run Log for errors or QA messages.
2. Confirm the source file is a PDF and is readable.
3. Try a clearer or higher-resolution version if the file is scanned.
4. Include the Run ID when reporting a reproducible issue.

Feedback is useful during the early release of ComIQ PDF Table Extractor, especially for unusual table layouts and scanned PDFs.

About ComIQ

ComIQ develops practical AI-assisted document-processing and automation tools focused on structured data extraction, QA validation, and efficient local/cloud workflows.

Current Store product: ComIQ PDF Table Extractor.