PDF & Word support
Upload PDFs and .docx files. More formats on the roadmap.
AI extraction built around how CMS content actually works
Pith isn't a generic PDF reader. It's an extraction pipeline designed to produce CMS-native structured content.
Features
Extraction templates Start building templates
Tell the AI exactly what to find
Define a template with field names, types, and plain-English instructions for the AI. Text, numbers, dates, booleans, arrays — Pith maps each field to your CMS schema before a single page is processed.
Templates are reusable across documents. Set it up once for administrative orders and run every new filing through the same template automatically.
Confidence scoring
Know what the AI is uncertain about — before it ships
Every extracted field comes with a confidence score from 0 to 100. High-confidence fields can auto-approve. Low-confidence fields get flagged for human review.
Color-coded indicators make the review interface fast. Editors only touch what actually needs attention.
CMS connectors
Content arrives as native CMS data — not HTML
When Pith pushes to Sanity, body content becomes proper Portable Text blocks. Images become references. Related documents resolve to typed references in your schema.
No post-processing, no copy-paste, no HTML cleanup. Your CMS gets the structured data it was designed to hold.
Reference resolution
Cross-document relationships — handled automatically
Pith detects when a document references another document in your CMS and resolves it to a proper typed reference. Internal links, related content, parent-child relationships — all wired up at push time.
Batch processing
Hundreds of documents. Same pipeline.
Upload a folder of PDFs and let Pith work through them. Real-time progress tracking shows page counts, extraction status, and review queue depth as each document finishes.
Batch review lets you approve low-variance documents in bulk and focus manual time on exceptions.
Slug generation
Clean URLs configured for your CMS
Define slug rules with prefix support, multi-field composition, and custom separators. Pith generates clean, consistent URLs at push time — no manual slug entry, no duplicates.
More under the hood
Source citations
See which part of the source document each extracted value came from.
Inline editing
Edit any field in the review interface before approving.
Re-extract
Update your template and re-run extraction on any document without re-uploading.
Team review
Multiple reviewers can work through the queue simultaneously.
Usage tracking
See pages used this month, your limit, and a breakdown by project.