Sort incoming documents
Invoices, quotes, letters: the Sort files pipeline reads each document, labels it and files it into the right folder
Automatic sorting, folder by folder
Knowledge base building
Your PDFs, emails and files become a queryable base.
OCR, classification and summary pipelines, inside your perimeter
Assemble a pipeline in the console: a trigger, a source, steps, a destination. Six pre-wired templates read your PDFs and attachments, classify them, summarize them and file them into Iceberg tables or folders. OCR and classification run on a model served in your own cluster, or on a third-party API if you prefer.
Specifications
- Pipeline editor
- Trigger, source, steps, destination
- OCR & extraction
- PDF, Word, images into Iceberg tables
- LLM classification
- Labeling and filing into folders
- Hippocampe
- L0/L1 summaries agents can search
- Templates
- Six pre-wired pipelines, or a blank one
- Operations
- Cron schedule, console and hfctl
Use cases
Query your PDFs
Extract documents runs a bucket of PDFs through OCR and lands the result in an Iceberg table, queryable like any other table
From PDF to Iceberg table
Give your agents a memory
Memory distillation summarizes a table into L0 and L1 levels (Hippocampe): your AI agents search the synthesis before diving into the detail
Progressive search
In action
Classify the inbound document flow
A property management team receives invoices, quotes and letters by email every day
- Plug the email connector into the management inbox
- Start from the Sort files template: pick the bucket, the classification model and the labels
- Deploy: on every schedule, the pipeline files each document into the right folder, property by property
- Track runs from the pipelines overview
No more manual sorting: every document lands classified in the right place
Make PDF archives queryable
A data team needs to make years of accumulated PDF archives usable
- Drop the archives into one of the organization's S3 buckets
- Start from the Extract documents template: source bucket, destination Iceberg catalog and schema
- OCR runs on the model served in your cluster, structured data lands in an Iceberg table
- Query the archives in SQL from Petite Requête
Years of archives become a queryable knowledge base
Key benefits
- Your documents become a queryable base, in SQL and for your AI agents
- Six pre-wired pipelines, a visual editor to adapt them, a dashboard to track them
- Fed by your connectors: inbound email, Google Drive, S3 buckets
- Everything runs inside your perimeter: OCR and classification on a model served in your cluster
Ready to make your documents talk?
Discover how document pipelines build your knowledge base.