What it does
PdfKb reads PDF, TXT and Markdown documents, extracts their text, and uses an AI model to create a linked Markdown wiki. It generates a summary page for each document, a glossary of concepts that links across documents, and an index page that ties everything together. The result can be used in Obsidian, a static site generator, or a retrieval‑augmented generation (RAG) pipeline.
Inputs and limits
- Accepted formats: PDF files with selectable text (scanned PDFs need OCR first).
- Upload limit: up to 10 PDF files per run, each up to 100 MB.
- The web service stores the uploaded files and the generated wiki for about one day and then deletes them.
Web app
The page /ai-prod/pdf-kb asks you to upload PDF files (up to 10 files, up to 100 MB each). After uploading, the AI builds a Markdown wiki that contains a summary page for each document, a concepts glossary, and an index page. You can then chat with the knowledge base; answers are cited with the source document and page.
Command line (CLI)
The PdfKb CLI plugin reads input files from a task folder and writes the wiki files (index.md, concepts.md, and one <document>.md per input) to an output folder. The plugin requires a .NET 10 runtime and a Documentize license file. Use --max-concepts to limit the number of concepts extracted per document (default 15).
# Linux / macOS (needs the .NET 10 runtime and documentize.lic next to the DLL)
export PLUGIN_INPUT_DIR=./work/input
export PLUGIN_OUTPUT_DIR=./work/output
mkdir -p "$PLUGIN_INPUT_DIR/job1"
cp sample.pdf cv.pdf "$PLUGIN_INPUT_DIR/job1/"
dotnet Documentize.PdfKb.dll --task-id job1 --max-concepts 5
ls "$PLUGIN_OUTPUT_DIR/job1"
# Windows PowerShell
$env:PLUGIN_INPUT_DIR = "$PWD\work\input"
$env:PLUGIN_OUTPUT_DIR = "$PWD\work\output"
New-Item -ItemType Directory -Force "$env:PLUGIN_INPUT_DIR\job1" | Out-Null
Copy-Item sample.pdf, cv.pdf "$env:PLUGIN_INPUT_DIR\job1\"
dotnet Documentize.PdfKb.dll --task-id job1 --max-concepts 5
Get-ChildItem "$env:PLUGIN_OUTPUT_DIR\job1"
Verified
Every code sample on this page was run on 2026-11-03 and its result was checked:
- PdfKb CLI conversion with text check (CLI plugin): passed in 31 s