Auto-Tag PDFs at Scale: AI & Template-Based Tagging with PDFix SDK

• Use case: Fastest way to achieve a structured, tagged document
• How it works: The SDK detects text blocks, reading order, tables, and lists automatically
• Pros: Zero setup, immediate results
• Cons: Less predictable structure, may require refinements

from pdfixsdk import *

pdfix = GetPdfix()
doc = pdfix.OpenDoc("weekly_market_commentary_2025_0505.pdf", "")
if not doc:
    raise RuntimeError(pdfix.GetError())

if not doc.AddTags(PdfTagsParams()):
    raise RuntimeError(pdfix.GetError())

doc.Save("weekly_market_commentary_2025_0505_tagged.pdf", kSaveFull)
doc.Close()

#!/bin/bash

INPUT_PDF=weekly_market_commentary_2025_0505.pdf
OUTPUT_PDF=weekly_market_commentary_2025_0505_tagged.pdf

# auto-tag with template
pdfix_app add-tags -i $INPUT_PDF -o $OUTPUT_PDF

• https://github.com/pdfix/pdfix_sdk_example_python
• https://pdfix.github.io/pdfix_sdk_builds
• https://pdfix.net/support/pdfix-command-line

• Use case: Improve document structure by identifying headings, and removing headers or footers automatically.
• How it works: Preflight analyzes the document and generates a template to enhance auto-tagging.
• Pros: Better results than plain auto-tagging.
• Cons: Results vary based on document complexity.

from pdfixsdk import *

pdfix = GetPdfix()
doc = pdfix.OpenDoc("weekly_market_commentary_2025_0505.pdf", "")
if not doc:
    raise RuntimeError(pdfix.GetError())

tmpl = doc.GetTemplate()

for i in range(doc.GetNumPages() - 1):
    tmpl.AddPage(i)
tmpl.Update()

if not doc.AddTags(PdfTagsParams()):
    raise RuntimeError(pdfix.GetError())

doc.Save("weekly_market_commentary_2025_0505_tagged.pdf", kSaveFull)
doc.Close()

#!/bin/bash

INPUT_PDF=weekly_market_commentary_2025_0505.pdf
OUTPUT_PDF=weekly_market_commentary_2025_0505_tagged.pdf

# auto-tag with template
pdfix_app add-tags -i $INPUT_PDF -o $OUTPUT_PDF --preflight

• https://github.com/pdfix/pdfix_sdk_example_python
• https://pdfix.github.io/pdfix_sdk_builds
• https://pdfix.net/support/pdfix-command-line

External AI actions let you integrate AI models and advanced tools directly into your workflow.
For document layout detection, we currently support three external AI models – but you’re not limited to just these. The external actions system allows you to connect any other model, including your own, giving you full flexibility to tailor automation to your needs.

• ☁️ Auto-Tag PDF (Amazon Textract)
• 🖥️ Auto-Tag PDF (Docling IBM)
• 🖥️ Auto-Tag PDF (Paddle)

When dealing with complex documents or diverse layouts, an AI-generated template can deliver the most intelligent, adaptive tagging results.

• Use case: Documents too complex for pre-defined templates.
• How it works: AI models create JSON templates that guide PDFix SDK in tagging.
• Pros: Smarter detection, flexible structure recognition.
• Cons: Requires AI integration setup.

Currenty

from pdfixsdk import *
import os, subprocess

def create_pdfix_ai_template(input_pdf: str, output_json: str, zoom=4.0):
    workdir = os.path.abspath(os.path.dirname(input_pdf))
    cmd = [
        "docker", "run", "--rm",
        "-v", f"{workdir}:/data", "-w", "/data",
        "pdfix/pdf-accessibility-docling:v1.0.4",
        "template", "-i", f"/data/{os.path.basename(input_pdf)}",
        "-o", f"/data/{os.path.basename(output_json)}",
        "--zoom", str(zoom)
    ]
    subprocess.run(cmd, check=True)


pdfix = GetPdfix()
doc = pdfix.OpenDoc("input.pdf", "")
if not doc:
    raise RuntimeError(pdfix.GetError())

create_pdfix_ai_template("weekly_market_commentary_2025_0505.pdf", "ai-template.json")

tmpl = doc.GetTemplate()
stm = pdfix.CreateFileStream("ai-template.json", kPsReadOnly)
tmpl.LoadFromStream(stm, kDataFormatJson)
stm.Destroy()

if not doc.AddTags(PdfTagsParams()):
    raise RuntimeError(pdfix.GetError())

doc.Save("weekly_market_commentary_2025_0505_tagged.pdf", kSaveFull)
doc.Close()

#!/bin/bash

INPUT_PDF=weekly_market_commentary_2025_0505.pdf
OUTPUT_PDF=weekly_market_commentary_2025_0505_tagged.pdf
AI_TEMPLATE=ai_template.json

docker run -v $(pwd):/data -w /data --rm pdfix/pdf-accessibility-docling:v1.0.4 template -i $INPUT_PDF -o $AI_TEMPLATE --zoom 4.0
pdfix_app add-tags -i $INPUT_PDF -o $OUTPUT_PDF -c $AI_TEMPLATE

• https://github.com/pdfix/pdfix_sdk_example_python
• https://pdfix.github.io/pdfix_sdk_builds
• https://pdfix.net/support/pdfix-command-line
• https://pdfix.net/products/actions-marketplace/

For repetitive document layouts such as invoices, statements, or catalogs, a pre-created template ensures predictable and consistent tagging.

• Use case: High-volume workflows with uniform document types.
• How it works: Define structure once in JSON, apply across thousands of PDFs.
• Pros: Consistency, predictability, best accessibility results, ability to define tag properties such as Tag type, or Alt text.
• Cons: Requires upfront template preparation.

from pdfixsdk import *

pdfix = GetPdfix()
doc = pdfix.OpenDoc("weekly_market_commentary_2025_0505.pdf", "")
if not doc:
    raise RuntimeError(pdfix.GetError())

tmpl = doc.GetTemplate()

stm = pdfix.CreateFileStream("template.json", kPsReadOnly)
tmpl.LoadFromStream(stm, kDataFormatJson)
stm.Destroy()

if not doc.AddTags(PdfTagsParams()):
    raise RuntimeError(pdfix.GetError())

doc.Save("weekly_market_commentary_2025_0505_tagged.pdf", kSaveFull)
doc.Close()

#!/bin/bash

INPUT_PDF=weekly_market_commentary_2025_0505.pdf
OUTPUT_PDF=weekly_market_commentary_2025_0505_tagged.pdf
CUSTOM_TEMPLATE=template.json
pdfix_app add-tags -i $INPUT_PDF -o $OUTPUT_PDF -c $CUSTOM_TEMPLATE

• https://github.com/pdfix/pdfix_sdk_example_python
• https://pdfix.github.io/pdfix_sdk_builds
• https://pdfix.net/support/pdfix-command-line
• https://pdfix.net/user-guide-template/

verapdf -f ua1 weekly_market_commentary_2025_0505_tagged.pdf