Mailhook and Knowledge: Disclosure Pack Extraction Template
A ready-made Spojit workflow that turns a multi-document disclosure pack emailed to a Mailhook address into one structured extract of the key terms across every PDF.
What It Builds
When a disclosure pack lands at your workflow's Mailhook address, an Attachment node fetches every PDF in Multiple mode. Each file is embedded into a Transient Knowledge collection for that single run, then a Knowledge Query node reads across all the documents at once and returns the important terms as one structured response. Nothing persists after the run, so each pack stays self-contained.
The Prompt
Paste this into Miraxa and it builds the workflow, connecting the tools for you:
Build a workflow that triggers when a disclosure pack is emailed to a Mailhook address. Fetch all of the PDF attachments, embed each one into a transient knowledge collection for this run, then query across every document to extract the key terms (parties, dates, obligations, amounts, governing law, and termination clauses) and return them as one structured response.Connectors Used
- Mailhook trigger - starts a run whenever a pack arrives at the workflow's dedicated address.
- Attachment node - fetches the PDF bytes in Multiple mode.
- PDF connector - reads the document contents, including scanned pages via OCR.
- Knowledge node - embeds the files into a Transient collection, then Queries across them with a Response Schema for clean JSON.
Customize It
Edit the list of key terms in the prompt to match your document type: add renewal date and liability cap for contracts, or property address and price for property disclosures. You can also ask Miraxa to send the extract to a channel or mailbox after the query.
Tips
- Transient mode keeps each pack isolated; choose a persistent collection only if you want the documents searchable later.
- The Attachment node is available only in Mailhook workflows, so keep the Mailhook trigger in place.
- Define a Response Schema on the Query so downstream nodes get predictable fields instead of free text.